From 153471d8a76df61e6b5b996525b865f9a1b309b0 Mon Sep 17 00:00:00 2001
From: Mohamed ASSOUKTI <mohamed.assoukti@partnre.com>
Date: Tue, 19 Aug 2025 10:56:44 +0200
Subject: [PATCH 1/7] [DERCBOT-1609] Structuring the LLM response

---
 .../model/genai/BotRAGConfigurationDTO.kt     |   6 -
 .../src/test/kotlin/service/RAGServiceTest.kt |   2 -
 .../service/RAGValidationServiceTest.kt       |   1 -
 .../dialogs-list-filters.component.ts         |  20 +-
 .../dialogs-list/dialogs-list.component.ts    |   4 +
 .../models/engines-configurations.ts          | 201 +++++++++++++-
 .../rag/rag-settings/models/rag-settings.ts   |   3 -
 .../rag-settings/rag-settings.component.html  |  74 ------
 .../rag-settings.component.spec.ts            |  34 +--
 .../rag-settings/rag-settings.component.ts    | 100 +------
 .../web/src/app/shared/bot-shared.service.ts  |   1 +
 .../chat-ui-message-debug.component.html      |  21 +-
 .../chat-ui-message-debug.component.scss      |  21 +-
 .../admin/bot/rag/BotRAGConfiguration.kt      |   2 -
 .../kotlin/engine/config/RAGAnswerHandler.kt  | 159 +++++-------
 .../kotlin/BotRAGConfigurationMongoDAOTest.kt |   8 +-
 .../orchestratorclient/responses/Models.kt    |  20 +-
 .../responses/RAGResponse.kt                  |   3 +-
 .../models/llm/OllamaLLMSetting.kt            |   2 -
 .../utils/VectorStoreUtils.kt                 |   2 +-
 .../models/rag/rag_models.py                  |  96 ++++++-
 .../gen_ai_orchestrator/routers/rag_router.py |  28 +-
 .../routers/responses/responses.py            |   7 +-
 .../callback_handlers_factory.py              |   2 -
 .../langfuse_callback_handler_factory.py      |   9 +-
 .../services/langchain/rag_chain.py           | 245 ++++++++++++------
 .../observability/observabilty_service.py     |   4 +-
 .../server/tests/services/test_rag_chain.py   | 181 +++++++++----
 28 files changed, 733 insertions(+), 523 deletions(-)

diff --git a/bot/admin/server/src/main/kotlin/model/genai/BotRAGConfigurationDTO.kt b/bot/admin/server/src/main/kotlin/model/genai/BotRAGConfigurationDTO.kt
index 1214b137ad..5afb90aba9 100644
--- a/bot/admin/server/src/main/kotlin/model/genai/BotRAGConfigurationDTO.kt
+++ b/bot/admin/server/src/main/kotlin/model/genai/BotRAGConfigurationDTO.kt
@@ -42,8 +42,6 @@ data class BotRAGConfigurationDTO(
     val emSetting: EMSettingDTO,
     val indexSessionId: String? = null,
     val indexName: String? = null,
-    val noAnswerSentence: String,
-    val noAnswerStoryId: String? = null,
     val documentsRequired: Boolean = true,
     val debugEnabled: Boolean,
     val maxDocumentsRetrieved: Int,
@@ -63,8 +61,6 @@ data class BotRAGConfigurationDTO(
         emSetting = configuration.emSetting.toDTO(),
         indexSessionId = configuration.indexSessionId,
         indexName = configuration.generateIndexName(),
-        noAnswerSentence = configuration.noAnswerSentence,
-        noAnswerStoryId = configuration.noAnswerStoryId,
         documentsRequired = configuration.documentsRequired,
         debugEnabled = configuration.debugEnabled,
         maxDocumentsRetrieved = configuration.maxDocumentsRetrieved,
@@ -101,8 +97,6 @@ data class BotRAGConfigurationDTO(
                     dto = emSetting,
                 ),
             indexSessionId = indexSessionId,
-            noAnswerSentence = noAnswerSentence,
-            noAnswerStoryId = noAnswerStoryId,
             documentsRequired = documentsRequired,
             debugEnabled = debugEnabled,
             maxDocumentsRetrieved = maxDocumentsRetrieved,
diff --git a/bot/admin/server/src/test/kotlin/service/RAGServiceTest.kt b/bot/admin/server/src/test/kotlin/service/RAGServiceTest.kt
index f2a2e36cde..698f4c5381 100644
--- a/bot/admin/server/src/test/kotlin/service/RAGServiceTest.kt
+++ b/bot/admin/server/src/test/kotlin/service/RAGServiceTest.kt
@@ -97,7 +97,6 @@ class RAGServiceTest : AbstractTest() {
                         model = "model",
                         apiBase = "url",
                     ),
-                noAnswerSentence = "No answer sentence",
                 documentsRequired = true,
                 debugEnabled = false,
                 maxDocumentsRetrieved = 2,
@@ -211,7 +210,6 @@ class RAGServiceTest : AbstractTest() {
             Assertions.assertEquals(PROVIDER, captured.questionAnsweringLlmSetting!!.provider.name)
             Assertions.assertEquals(TEMPERATURE, captured.questionAnsweringLlmSetting!!.temperature)
             Assertions.assertEquals(PROMPT, captured.questionAnsweringPrompt!!.template)
-            Assertions.assertEquals(null, captured.noAnswerStoryId)
         }
 
         TestCase<SaveFnEntry, Unit>("Save valid RAG Configuration that does not exist yet").given(
diff --git a/bot/admin/server/src/test/kotlin/service/RAGValidationServiceTest.kt b/bot/admin/server/src/test/kotlin/service/RAGValidationServiceTest.kt
index 0bdca1b629..edff4ee726 100644
--- a/bot/admin/server/src/test/kotlin/service/RAGValidationServiceTest.kt
+++ b/bot/admin/server/src/test/kotlin/service/RAGValidationServiceTest.kt
@@ -90,7 +90,6 @@ class RAGValidationServiceTest {
             questionAnsweringLlmSetting = openAILLMSetting,
             questionAnsweringPrompt = PromptTemplate(template = "How to bike in the rain"),
             emSetting = azureOpenAIEMSetting,
-            noAnswerSentence = " No answer sentence",
             documentsRequired = true,
             debugEnabled = false,
             maxDocumentsRetrieved = 2,
diff --git a/bot/admin/web/src/app/analytics/dialogs/dialogs-list/dialogs-list-filters/dialogs-list-filters.component.ts b/bot/admin/web/src/app/analytics/dialogs/dialogs-list/dialogs-list-filters/dialogs-list-filters.component.ts
index be812aa671..4272512856 100644
--- a/bot/admin/web/src/app/analytics/dialogs/dialogs-list/dialogs-list-filters/dialogs-list-filters.component.ts
+++ b/bot/admin/web/src/app/analytics/dialogs/dialogs-list/dialogs-list-filters/dialogs-list-filters.component.ts
@@ -14,7 +14,7 @@
  * limitations under the License.
  */
 
-import { Component, EventEmitter, Input, OnInit, Output } from '@angular/core';
+import { Component, EventEmitter, Input, OnDestroy, OnInit, Output } from '@angular/core';
 import { FormControl, FormGroup } from '@angular/forms';
 import { ConnectorType } from '../../../../core/model/configuration';
 import { Subject, debounceTime, take, takeUntil } from 'rxjs';
@@ -67,7 +67,7 @@ export type DialogListFilters = ExtractFormControlTyping<DialogListFiltersForm>;
   templateUrl: './dialogs-list-filters.component.html',
   styleUrl: './dialogs-list-filters.component.scss'
 })
-export class DialogsListFiltersComponent implements OnInit {
+export class DialogsListFiltersComponent implements OnInit, OnDestroy {
   private readonly destroy$: Subject<boolean> = new Subject();
   private lastEmittedValue: Partial<DialogListFilters> | null = null;
 
@@ -109,8 +109,9 @@ export class DialogsListFiltersComponent implements OnInit {
       this.lastEmittedValue = { ...this.form.value };
     }
 
-    this.form.valueChanges.pipe(debounceTime(800), takeUntil(this.destroy$)).subscribe(() => {
+    this.form.valueChanges.pipe(debounceTime(500), takeUntil(this.destroy$)).subscribe(() => {
       this.submitFiltersChange();
+      this.persisteDisplayTests();
     });
   }
 
@@ -143,10 +144,15 @@ export class DialogsListFiltersComponent implements OnInit {
 
   submitFiltersChange(): void {
     const formValue = this.form.value;
-    if (JSON.stringify(formValue) !== JSON.stringify(this.lastEmittedValue)) {
-      this.onFilter.emit(formValue);
-      this.lastEmittedValue = { ...formValue };
-    }
+    this.onFilter.emit(formValue);
+  }
+
+  persisteDisplayTests(): void {
+    const displayTests = this.getFormControl('displayTests')?.value;
+    this.botSharedService.session_storage = {
+      ...this.botSharedService.session_storage,
+      ...{ dialogs: { ...this.botSharedService.session_storage?.dialogs, displayTests } }
+    };
   }
 
   resetControl(ctrl: FormControl, input?: HTMLInputElement): void {
diff --git a/bot/admin/web/src/app/analytics/dialogs/dialogs-list/dialogs-list.component.ts b/bot/admin/web/src/app/analytics/dialogs/dialogs-list/dialogs-list.component.ts
index 28001c0089..71862b0443 100644
--- a/bot/admin/web/src/app/analytics/dialogs/dialogs-list/dialogs-list.component.ts
+++ b/bot/admin/web/src/app/analytics/dialogs/dialogs-list/dialogs-list.component.ts
@@ -73,6 +73,10 @@ export class DialogsListComponent implements OnInit, OnChanges, OnDestroy {
   }
 
   ngOnInit() {
+    if (this.botSharedService.session_storage?.dialogs?.displayTests) {
+      this.filters.displayTests = this.botSharedService.session_storage.dialogs.displayTests;
+    }
+
     this.state.configurationChange.pipe(takeUntil(this.destroy$)).subscribe(() => {
       this.refresh();
     });
diff --git a/bot/admin/web/src/app/rag/rag-settings/models/engines-configurations.ts b/bot/admin/web/src/app/rag/rag-settings/models/engines-configurations.ts
index 925980ad0e..758ababaa3 100644
--- a/bot/admin/web/src/app/rag/rag-settings/models/engines-configurations.ts
+++ b/bot/admin/web/src/app/rag/rag-settings/models/engines-configurations.ts
@@ -27,35 +27,208 @@ import {
   PromptDefinitionFormatter
 } from '../../../shared/model/ai-settings';
 
-export const QuestionCondensingDefaultPrompt: string = `Given a chat history and the latest user question which might reference context in the chat history, formulate a standalone question which can be understood without the chat history. Do NOT answer the question, just reformulate it if needed and otherwise return it as is.`;
+export const QuestionCondensingDefaultPrompt: string = `You are a helpful assistant that reformulates questions.
+
+You are given:
+- The conversation history between the user and the assistant
+- The most recent user question
+
+Your task:
+- Reformulate the user’s latest question into a clear, standalone query.
+- Incorporate relevant context from the conversation history.
+- Do NOT answer the question.
+- If the history does not provide additional context, keep the question as is.
+
+Return only the reformulated question.`;
 
 export const QuestionAnsweringDefaultPrompt: string = `# TOCK (The Open Conversation Kit) chatbot
 
-## General context
+## General Context
 
 You are a chatbot designed to provide short conversational messages in response to user queries.
+Your job is to surface the right information from provided context.
+
+### Forbidden Topics
+
+- **Out of Scope**:
+  - Topics unrelated to the business domain (e.g., personal life advice, unrelated industries).
+  - Requests for unsupported features (e.g., "How do I integrate with [UnsupportedTool]?").
+
+- **Toxic/Offensive Content**:
+  - Hate speech, harassment, or discriminatory language.
+  - Illegal activities or unethical requests (e.g., "How do I bypass security protocols?").
+
+- **Personal/Private Matters**:
+  - User-specific data (e.g., personal identification, private conversations).
+  - Internal or confidential company information (e.g., unreleased product details).
+
+- **Regulated Topics**:
+  - Medical, legal, or financial advice (e.g., "What’s the best treatment for [condition]?").
+  - Speculative or unverified claims (e.g., "Is [Product X] better than competitors?").
+
+### Answer Style
+
+- **Tone**: neutral, kind, “you” address, light humor when appropriate.
+- **Language**: Introduce technical jargon only when strictly necessary and briefly define it.
+- **Structure**: Use short sentences, bold or bullet points for key ideas, headings to separate the main sections, and fenced \`code\` blocks for examples.
+- **Style**: Direct and technical tone, with **bold** for important concepts.
+- **Formatting**: Mandatory Markdown, with line breaks for readability.
+- **Examples**: Include a concrete example (code block or CLI command) for each feature.
+
+### Guidelines
+
+1. If the question is unclear, politely request rephrasing.
+2. If the docs lack or don’t cover the answer, reply with \`"status": "not_found_in_context"\`.
+3. Conclude with:
+   - “Does this help?”
+   - Offer to continue on the same topic, switch topics, or contact support.
+
+### Verification Steps
+
+Before responding, ensure:
+
+- The documentation actually addresses the question.
+- Your answer is consistent with the docs.
+
+## Technical Instructions:
+
+You must respond STRICTLY in valid JSON format (no extra text, no explanations).
+Use only the following context and the rules below to respond the question.
+
+### Rules for JSON output:
+
+- If the answer is found in the context:
+  - "status": "found_in_context"
+  - "answer": the best possible answer in {{ locale }}
+  - "display_answer": "true"
+  - "redirection_intent": null
 
-## Guidelines
+- If the answer is NOT found in the context:
+  - "status": "not_found_in_context"
+  - "answer":
+    - The "answer" must not be a generic refusal. Instead, generate a helpful and intelligent response:
+      - If a similar or related element exists in the context (e.g., another product, service, or regulation with a close name, date, or wording), suggest it naturally in the answer.
+      - If no similar element exists, politely acknowledge the lack of information while encouraging clarification or rephrasing.
+    - Always ensure the response is phrased in a natural and user-friendly way, rather than a dry "not found in context".
+  - "display_answer": "true"
+  - "redirection_intent": null
 
-Incorporate any relevant details from the provided context into your answers, ensuring they are directly related to the user's query.
+- If the question is forbidden or offensive:
+  - "status": "out_of_scope"
+  - "answer":
+    - Generate a polite response explaining why a reponse can't be done.
+  - "topic": "Out of scope or offensive question"
+  - "display_answer": "true"
+  - "redirection_intent": null
 
-## Style and format
+- If the question is small talk:
+  Only to conversational rituals such as greetings (e.g., “hello”, “hi”) and farewells or leave-takings (e.g., “goodbye”, “see you”), you may ignore the context and generate a natural small-talk response in the "answer".
+  - "status": "small_talk"
+  - "topic": "greetings"
+  - "display_answer": "true"
+  - "redirection_intent": null
 
-Your tone is empathetic, informative and polite.
+### Confidence score:
 
-## Additional instructions
+Gives a confidence score between 0 and 1 on the relevance of the answer provided to the user's question:
 
-Use the following pieces of retrieved context to answer the question.
-If you dont know the answer, answer (exactly) with "{{no_answer}}".
-Answer in {{locale}}.
+- "confidence_score": <CONFIDENCE_SCORE>
 
-## Context
+### Users question understanding:
 
-{{context}}
+Explain in one sentence what you understood from the user's question:
 
-## Question
+- "understanding": "<UNDERSTANDING_OF_THE_USER_QUESTION>"
+
+### Context usage tracing requirements (MANDATORY):
+
+- You MUST include **every** chunk from the input context in the "context_usage" array, in the same order they appear. **No chunk may be omitted**.
+- If explicit chunk identifiers are present in the context, use them.
+- For each chunk object:
+  - "chunk": "<chunk_identifier>"
+  - "used_in_response":
+    - "true" if the chunk contributed
+    - "false" if the chunk didn't contributed
+  - "sentences": ["<verbatim sentence(s) from this chunk used to answer the question>"] — leave empty \`[]\` if none.
+  - "reason": "null" if the chunk contributed; otherwise a concise explanation of why this chunk is not relevant to the question (e.g., "general background only", "different product", "no data for the asked period", etc.).
+- If there are zero chunks in the context, return \`"context": []\`.
+
+### Topic Identification & Suggestion Rules (MANDATORY):
+
+#### Rules for Topic Assignment
+
+- If the question explicitly matches a predefined topic, use:
+  - \`"topic": "<EXACT_PREDEFINED_TOPIC>"\`
+  - \`"suggested_topics": []\`
+
+- If the question does not match any predefined topic, use the \`unknown\` topic and provide 1 relevant and concise new topic suggestion in "suggested_topics":
+  - \`"topic": "unknown"\`
+  - \`"suggested_topics": ["<NEW_TOPIC_SUGGESTION>"]\`
+
+#### Predefined topics (use EXACT spelling, no variations):
+
+- \`Concepts and Definitions\`
+- \`Processes and Methods\`
+- \`Tools and Technologies\`
+- \`Rules and Regulations\`
+- \`Examples and Use Cases\`
+- \`Resources and References\`
+
+## Context:
+
+{{ context }}
+
+## User question
+
+You are given the conversation history between a user and an assistant:
+
+- analyze the conversation history to understand the context and the user’s intent
+- use this context to correctly interpret the user’s final question
+- answer only the final user question below in a relevant and contextualized way
+
+Conversation history:
+{{ chat_history }}
+
+User’s final question:
+{{ question }}
+
+## Output format (JSON only):
+
+Return your response in the following format:
+
+\`\`\`json
+{
+    "status": "found_in_context" | "not_found_in_context" | "small_talk" | "out_of_scope",
+    "answer": "<TEXTUAL_ANSWER>",
+    "display_answer": true | false,
+    "confidence_score": "<CONFIDENCE_SCORE>",
+    "topic": "<EXACT_PREDEFINED_TOPIC>" | "greetings" | "Out of scope or offensive question" | "unknown",
+    "suggested_topics": ["<topic_suggestion_1>", "<topic_suggestion_2>"],
+    "understanding": "<UNDERSTANDING_OF_THE_USER_QUESTION>",
+    "redirection_intent": null,
+    "context_usage": [
+        {
+            "chunk": "1",
+            "sentences": ["SENTENCE_1", "SENTENCE_2"],
+            "used_in_response": true | false,
+            "reason": null
+        },
+        {
+            "chunk": "2",
+            "sentences": [],
+            "used_in_response": true | false,
+            "reason": "General description; no details related to the question."
+        },
+        {
+            "chunk": "3",
+            "sentences": ["SENTENCE_X"],
+            "used_in_response": true | false,
+            "reason": null
+        }
+    ]
+}
+\`\`\`
 
-{{question}}
 `;
 
 export const QuestionCondensing_prompt: ProvidersConfigurationParam[] = [
diff --git a/bot/admin/web/src/app/rag/rag-settings/models/rag-settings.ts b/bot/admin/web/src/app/rag/rag-settings/models/rag-settings.ts
index 52b4dbb9fc..e605ec6561 100644
--- a/bot/admin/web/src/app/rag/rag-settings/models/rag-settings.ts
+++ b/bot/admin/web/src/app/rag/rag-settings/models/rag-settings.ts
@@ -24,9 +24,6 @@ export interface RagSettings {
 
   debugEnabled: boolean;
 
-  noAnswerSentence: string;
-  noAnswerStoryId: string | null;
-
   questionCondensingLlmSetting: llmSetting;
   questionCondensingPrompt: PromptDefinition;
   maxMessagesFromHistory: number;
diff --git a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.html b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.html
index 2ecc1b9a2c..cf3527fa9f 100644
--- a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.html
+++ b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.html
@@ -430,80 +430,6 @@ <h1 class="flex-grow-1">Rag settings</h1>
         </nb-card-body>
       </nb-card>
 
-      <nb-card class="mb-3">
-        <nb-card-header>Conversation flow</nb-card-header>
-        <nb-card-body>
-          <div class="row">
-            <div class="col-12 px-3">
-              <tock-form-control
-                label="No answer sentence"
-                name="noAnswerSentence"
-                [controls]="noAnswerSentence"
-                [required]="true"
-                [showError]="isSubmitted"
-                [boldLabel]="false"
-                information="Phrase to be displayed to the user if no response could be given to his request"
-              >
-                <input
-                  nbInput
-                  fullWidth
-                  formControlName="noAnswerSentence"
-                />
-              </tock-form-control>
-            </div>
-            <div class="col-12 px-3">
-              <tock-form-control
-                label="Unanswered story redirection"
-                name="noAnswerStoryId"
-                [controls]="noAnswerStoryId"
-                [showError]="isSubmitted"
-                [boldLabel]="false"
-                information="Story to redirect the user to if no answer is given to the request"
-              >
-                <nb-form-field>
-                  <input
-                    nbInput
-                    fullWidth
-                    autocomplete="off"
-                    type="text"
-                    [value]="getCurrentStoryLabel"
-                    [nbAutocomplete]="storiesautocomplete"
-                    (input)="filterStoriesList($any($event.target).value)"
-                    (change)="onStoryChange($any($event.target).value)"
-                    (focus)="storyInputFocus()"
-                    (blur)="storyInputBlur($event)"
-                  />
-                  <button
-                    nbSuffix
-                    nbButton
-                    ghost
-                    nbTooltip="Delete unanswered story redirection"
-                    (click)="removeNoAnswerStoryId()"
-                    *ngIf="getCurrentStoryLabel !== ''"
-                  >
-                    <nb-icon icon="x-lg"></nb-icon>
-                  </button>
-                </nb-form-field>
-              </tock-form-control>
-
-              <nb-autocomplete
-                #storiesautocomplete
-                optionsListClass="option-list--break-word"
-                (selectedChange)="storySelectedChange($event)"
-              >
-                <nb-option
-                  *ngFor="let option of filteredStories$ | async"
-                  [value]="option.storyId"
-                  [disabled]="!isStoryEnabled(option)"
-                >
-                  {{ option.name }} <span *ngIf="!isStoryEnabled(option)"> &nbsp;(disabled)</span>
-                </nb-option>
-              </nb-autocomplete>
-            </div>
-          </div>
-        </nb-card-body>
-      </nb-card>
-
       <nb-card *ngIf="settingsBackup">
         <nb-card-header>Settings deletion</nb-card-header>
         <nb-card-body>
diff --git a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.spec.ts b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.spec.ts
index 75cebbc001..4a8dca8248 100644
--- a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.spec.ts
+++ b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.spec.ts
@@ -18,8 +18,6 @@ import { NO_ERRORS_SCHEMA } from '@angular/core';
 import { ComponentFixture, TestBed } from '@angular/core/testing';
 import { NbToastrService } from '@nebular/theme';
 import { of } from 'rxjs';
-import { BotService } from '../../bot/bot-service';
-import { StoryDefinitionConfigurationSummary } from '../../bot/model/story';
 import { RestService } from '../../core-nlp/rest/rest.service';
 import { StateService } from '../../core-nlp/state.service';
 import { BotConfigurationService } from '../../core/bot-configuration.service';
@@ -28,22 +26,6 @@ import { RagSettings } from './models';
 
 import { RagSettingsComponent } from './rag-settings.component';
 
-const stories = [
-  {
-    _id: '123456789abcdefghijkl',
-    storyId: 'teststory',
-    botId: 'new_assistant',
-    intent: {
-      name: 'testintent'
-    },
-    currentType: 'simple',
-    name: 'Test story',
-    category: 'faq',
-    description: '',
-    lastEdited: '2023-07-31T14:48:21.291Z'
-  } as unknown as StoryDefinitionConfigurationSummary
-];
-
 const settings = {
   id: 'abcdefghijkl123456789',
   namespace: 'app',
@@ -64,9 +46,7 @@ const settings = {
     embeddingModelName: 'text-embedding-ada-002',
     embeddingApiKey: 'Embedding OpenAI API Key',
     embeddingApiVersion: '2023-03-15-preview'
-  },
-  noAnswerSentence: 'No answer sentence',
-  noAnswerStoryId: 'null'
+  }
 } as unknown as RagSettings;
 
 describe('RagSettingsComponent', () => {
@@ -77,12 +57,6 @@ describe('RagSettingsComponent', () => {
     await TestBed.configureTestingModule({
       declarations: [RagSettingsComponent],
       providers: [
-        {
-          provide: BotService,
-          useValue: {
-            searchStories: () => of(stories)
-          }
-        },
         {
           provide: StateService,
           useValue: {
@@ -118,10 +92,6 @@ describe('RagSettingsComponent', () => {
     expect(component).toBeTruthy();
   });
 
-  it('should load stories', () => {
-    expect(component.availableStories).toEqual(stories);
-  });
-
   it('should load settings', () => {
     expect(component.settingsBackup).toEqual(settings);
 
@@ -130,7 +100,7 @@ describe('RagSettingsComponent', () => {
     delete cleanedSettings['botId'];
 
     const cleanedFormValue = deepCopy(component.form.getRawValue());
-    delete cleanedFormValue.params.apiKey;
+    delete cleanedFormValue.questionAnsweringLlmSetting.apiKey;
 
     expect(cleanedFormValue as unknown).toEqual(cleanedSettings as unknown);
   });
diff --git a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.ts b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.ts
index 0a7f41f8a3..c906ffc48e 100644
--- a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.ts
+++ b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.ts
@@ -16,11 +16,9 @@
 
 import { Component, OnDestroy, OnInit, TemplateRef, ViewChild } from '@angular/core';
 import { FormControl, FormGroup, Validators } from '@angular/forms';
-import { debounceTime, forkJoin, Observable, of, Subject, take, takeUntil, pairwise } from 'rxjs';
+import { debounceTime, forkJoin, Observable, Subject, takeUntil, pairwise } from 'rxjs';
 import { NbDialogRef, NbDialogService, NbToastrService, NbWindowService } from '@nebular/theme';
 
-import { BotService } from '../../bot/bot-service';
-import { StoryDefinitionConfiguration, StorySearchQuery } from '../../bot/model/story';
 import { RestService } from '../../core-nlp/rest/rest.service';
 import { StateService } from '../../core-nlp/state.service';
 import { EnginesConfigurations, QuestionCondensing_prompt, QuestionAnswering_prompt } from './models/engines-configurations';
@@ -47,9 +45,6 @@ interface RagSettingsForm {
 
   debugEnabled: FormControl<boolean>;
 
-  noAnswerSentence: FormControl<string>;
-  noAnswerStoryId: FormControl<string>;
-
   indexSessionId: FormControl<string>;
   indexName: FormControl<string>;
 
@@ -89,10 +84,6 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
 
   questionAnswering_prompt = QuestionAnswering_prompt;
 
-  availableStories: StoryDefinitionConfiguration[];
-
-  filteredStories$: Observable<StoryDefinitionConfiguration[]>;
-
   settingsBackup: RagSettings;
 
   isSubmitted: boolean = false;
@@ -103,7 +94,6 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
   @ViewChild('importModal') importModal: TemplateRef<any>;
 
   constructor(
-    private botService: BotService,
     private state: StateService,
     private rest: RestService,
     private toastrService: NbToastrService,
@@ -176,14 +166,9 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
       this.configurations = confs;
 
       if (confs.length) {
-        forkJoin([this.getStoriesLoader(), this.getRagSettingsLoader()]).subscribe((res) => {
-          this.availableStories = res[0];
-
-          const settings = res[1];
+        forkJoin([this.getRagSettingsLoader()]).subscribe((res) => {
+          const settings = res[0];
           if (settings?.id) {
-            if (!settings.noAnswerStoryId) {
-              settings.noAnswerStoryId = null;
-            }
             this.settingsBackup = deepCopy(settings);
             setTimeout(() => {
               this.initForm(settings);
@@ -209,9 +194,6 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
 
     debugEnabled: new FormControl({ value: undefined, disabled: !this.canRagBeActivated() }),
 
-    noAnswerSentence: new FormControl(undefined, [Validators.required]),
-    noAnswerStoryId: new FormControl(undefined),
-
     indexSessionId: new FormControl(undefined),
     indexName: new FormControl(undefined),
 
@@ -256,13 +238,6 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
     return this.form.get('maxMessagesFromHistory') as FormControl;
   }
 
-  get noAnswerSentence(): FormControl {
-    return this.form.get('noAnswerSentence') as FormControl;
-  }
-  get noAnswerStoryId(): FormControl {
-    return this.form.get('noAnswerStoryId') as FormControl;
-  }
-
   get indexSessionId(): FormControl {
     return this.form.get('indexSessionId') as FormControl;
   }
@@ -283,11 +258,6 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
     return this.isSubmitted ? this.form.valid : this.form.dirty;
   }
 
-  get getCurrentStoryLabel(): string {
-    const currentStory = this.availableStories?.find((story) => story.storyId === this.noAnswerStoryId.value);
-    return currentStory?.name || '';
-  }
-
   accordionItemsExpandedState: Map<string, boolean>;
 
   isAccordionItemsExpanded(itemName: string): boolean {
@@ -455,66 +425,6 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
     return EnginesConfigurations[AiEngineSettingKeyName.emSetting].find((e) => e.key === this.emProvider.value);
   }
 
-  private getStoriesLoader(): Observable<StoryDefinitionConfiguration[]> {
-    return this.botService
-      .getStories(
-        new StorySearchQuery(
-          this.state.currentApplication.namespace,
-          this.state.currentApplication.name,
-          this.state.currentLocale,
-          0,
-          10000,
-          undefined,
-          undefined,
-          false
-        )
-      )
-      .pipe(take(1));
-  }
-
-  isStoryEnabled(story: StoryDefinitionConfiguration): boolean {
-    for (let i = 0; i < story.features.length; i++) {
-      if (!story.features[i].enabled && !story.features[i].switchToStoryId && !story.features[i].endWithStoryId) {
-        return false;
-      }
-    }
-    return true;
-  }
-
-  storySelectedChange(storyId: string): void {
-    this.noAnswerStoryId.patchValue(storyId);
-    this.form.markAsDirty();
-  }
-
-  onStoryChange(value: string): void {
-    if (value?.trim() == '') {
-      this.removeNoAnswerStoryId();
-    }
-  }
-
-  removeNoAnswerStoryId(): void {
-    this.noAnswerStoryId.patchValue(null);
-    this.form.markAsDirty();
-  }
-
-  filterStoriesList(e: string): void {
-    this.filteredStories$ = of(this.availableStories.filter((optionValue) => optionValue.name.toLowerCase().includes(e.toLowerCase())));
-  }
-
-  storyInputFocus(): void {
-    this.filteredStories$ = of(this.availableStories);
-  }
-
-  storyInputBlur(e: FocusEvent): void {
-    setTimeout(() => {
-      // timeout needed to avoid reseting input and filtered stories when clicking on autocomplete suggestions (which fires blur event)
-      const target: HTMLInputElement = e.target as HTMLInputElement;
-      target.value = this.getCurrentStoryLabel;
-
-      this.filteredStories$ = of(this.availableStories);
-    }, 100);
-  }
-
   cancel(): void {
     this.initForm(this.settingsBackup);
   }
@@ -529,14 +439,10 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
       delete formValue['emProvider'];
       formValue.namespace = this.state.currentApplication.namespace;
       formValue.botId = this.state.currentApplication.name;
-      formValue.noAnswerStoryId = this.noAnswerStoryId.value === 'null' ? null : this.noAnswerStoryId.value;
 
       const url = `/gen-ai/bots/${this.state.currentApplication.name}/configuration/rag`;
       this.rest.post(url, formValue, null, null, true).subscribe({
         next: (ragSettings: RagSettings) => {
-          if (!ragSettings.noAnswerStoryId) {
-            ragSettings.noAnswerStoryId = null;
-          }
           this.settingsBackup = ragSettings;
 
           this.indexName.reset();
diff --git a/bot/admin/web/src/app/shared/bot-shared.service.ts b/bot/admin/web/src/app/shared/bot-shared.service.ts
index 5fae05656e..dcf933be48 100644
--- a/bot/admin/web/src/app/shared/bot-shared.service.ts
+++ b/bot/admin/web/src/app/shared/bot-shared.service.ts
@@ -25,6 +25,7 @@ import { AdminConfiguration } from './model/conf';
 
 export interface TockSimpleSessionStorage {
   test: { debug: boolean; sourceWithContent?: boolean };
+  dialogs: { displayTests: boolean };
 }
 
 @Injectable({
diff --git a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
index a9a3c82749..361645b06d 100644
--- a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
+++ b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
@@ -16,8 +16,25 @@
 
 <div
   (click)="showDebug()"
-  class="debug text-center w-100 mx-1"
+  class="debug d-flex gap-1 p-2"
   nbTooltip="Show debug infos"
 >
-  <span>{{ message.text }} DEBUG</span>
+  <div><strong>Status:</strong> {{ message.data.answer?.status }}</div>
+
+  <div *ngIf="message.data.answer?.topic && message.data.answer?.topic !== 'unknown'">
+    <strong>Topic:</strong> {{ message.data.answer?.topic }}
+  </div>
+
+  <div *ngIf="message.data.answer?.suggested_topics?.length">
+    <strong>Suggested topics:</strong>
+    <span *ngFor="let suggestion of message.data.answer.suggested_topics">
+      {{ suggestion }}
+    </span>
+  </div>
+
+  <div *ngIf="message.data.answer?.redirection_intent">
+    <strong>Redirection intent:</strong> {{ message.data.answer?.redirection_intent }}
+  </div>
+
+  <div *ngIf="message.data.answer?.confidence_score"><strong>Confidence score:</strong>: {{ message.data.answer?.confidence_score }}</div>
 </div>
diff --git a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.scss b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.scss
index b0b18a6758..5f59982a64 100644
--- a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.scss
+++ b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.scss
@@ -19,21 +19,26 @@
 
 :host {
   display: flex;
-  justify-content: center;
+  // justify-content: center;
 
   .debug {
+    width: 100%;
     cursor: pointer;
 
+    font-size: 0.75rem;
     color: var(--chat-message-sender-text-color);
+    border-top: 1px dashed var(--chat-message-sender-text-color);
     border-bottom: 1px dashed var(--chat-message-sender-text-color);
+    background-color: var(--background-basic-color-3);
 
-    height: 11px;
-    margin-bottom: 10px;
+    margin-left: 1em;
+    // height: 11px;
+    // margin-bottom: 10px;
 
-    span {
-      background: var(--card-background-color);
-      padding: 0 5px;
-      border-radius: 0.5rem;
-    }
+    // span {
+    //   background: var(--card-background-color);
+    //   padding: 0 5px;
+    //   border-radius: 0.5rem;
+    // }
   }
 }
diff --git a/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt b/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt
index aee3ad8d86..6f2c1af38e 100644
--- a/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt
+++ b/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt
@@ -35,8 +35,6 @@ data class BotRAGConfiguration(
     val llmSetting: LLMSetting? = null,
     val emSetting: EMSetting,
     val indexSessionId: String? = null,
-    val noAnswerSentence: String,
-    val noAnswerStoryId: String? = null,
     val documentsRequired: Boolean = true,
     val debugEnabled: Boolean = false,
     val maxDocumentsRetrieved: Int = 4,
diff --git a/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt b/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt
index cfa550317e..5f8d5c0fa1 100644
--- a/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt
+++ b/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt
@@ -16,11 +16,9 @@
 
 package ai.tock.bot.engine.config
 
-import ai.tock.bot.admin.bot.rag.BotRAGConfiguration
 import ai.tock.bot.admin.indicators.IndicatorValues
 import ai.tock.bot.admin.indicators.Indicators
 import ai.tock.bot.admin.indicators.metric.MetricType
-import ai.tock.bot.definition.RAGStoryDefinition
 import ai.tock.bot.definition.StoryDefinition
 import ai.tock.bot.engine.BotBus
 import ai.tock.bot.engine.BotRepository
@@ -34,9 +32,9 @@ import ai.tock.genai.orchestratorclient.requests.ChatMessage
 import ai.tock.genai.orchestratorclient.requests.ChatMessageType
 import ai.tock.genai.orchestratorclient.requests.DialogDetails
 import ai.tock.genai.orchestratorclient.requests.RAGRequest
+import ai.tock.genai.orchestratorclient.responses.LLMAnswer
 import ai.tock.genai.orchestratorclient.responses.ObservabilityInfo
 import ai.tock.genai.orchestratorclient.responses.RAGResponse
-import ai.tock.genai.orchestratorclient.responses.TextWithFootnotes
 import ai.tock.genai.orchestratorclient.retrofit.GenAIOrchestratorBusinessError
 import ai.tock.genai.orchestratorclient.retrofit.GenAIOrchestratorValidationError
 import ai.tock.genai.orchestratorclient.services.RAGService
@@ -64,7 +62,7 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
             BotRepository.saveMetric(createMetric(MetricType.STORY_HANDLED))
 
             // Call RAG Api - Gen AI Orchestrator
-            val (answer, debug, noAnswerStory, observabilityInfo) = rag(this)
+            val (answer, footnotes, debug, redirectStory, observabilityInfo) = rag(this)
 
             // Add debug data if available and if debugging is enabled
             if (debug != null) {
@@ -72,37 +70,42 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
                 sendDebugData("RAG", debug)
             }
 
-            // Handle the RAG answer
-            if (noAnswerStory == null && answer != null) {
-                logger.info { "Send RAG answer." }
+            val modifiedObservabilityInfo = observabilityInfo?.let { updateObservabilityInfo(this, it) }
 
-                val modifiedObservabilityInfo = observabilityInfo?.let { updateObservabilityInfo(this, it) }
+            // Footnotes building
+            val preparedFootnotes =
+                footnotes?.map {
+                    Footnote(
+                        it.identifier,
+                        it.title,
+                        it.url,
+                        if (action.metadata.sourceWithContent) it.content else null,
+                        it.score,
+                    )
+                }?.toMutableList() ?: mutableListOf()
+
+            // Identifying text to be sent
+            val textToSend = if (answer?.displayAnswer == true) answer.answer.orEmpty() else ""
 
-                send(
+            // Send SendSentenceWithFootnotes
+            logger.info { "Send RAG answer." }
+            send(
+                action =
                     SendSentenceWithFootnotes(
-                        botId,
-                        connectorId,
-                        userId,
-                        text = answer.text,
-                        footnotes =
-                            answer.footnotes.map {
-                                Footnote(
-                                    it.identifier,
-                                    it.title,
-                                    it.url,
-                                    if (action.metadata.sourceWithContent) it.content else null,
-                                    it.score,
-                                )
-                            }.toMutableList(),
-                        // modifiedObservabilityInfo includes the public langfuse URL if filled.
-                        metadata = ActionMetadata(isGenAiRagAnswer = true, observabilityInfo = modifiedObservabilityInfo),
+                        playerId = botId,
+                        applicationId = connectorId,
+                        recipientId = userId,
+                        text = textToSend,
+                        footnotes = preparedFootnotes,
+                        metadata =
+                            ActionMetadata(
+                                isGenAiRagAnswer = true,
+                                observabilityInfo = modifiedObservabilityInfo,
+                            ),
                     ),
-                )
-            } else {
-                logger.info { "No RAG answer to send, because a noAnswerStory is returned." }
-            }
+            )
 
-            noAnswerStory
+            redirectStory
         }
     }
 
@@ -122,7 +125,7 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
     }
 
     /**
-     * Manage story redirection when no answer redirection is filled
+     * Manage story redirection
      * Use the handler of the configured story otherwise launch default unknown story
      * @param botBus the bot Bus
      * @param response the RAG response
@@ -131,54 +134,9 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
         botBus: BotBus,
         response: RAGResponse?,
     ): StoryDefinition? {
-        return with(botBus) {
-            botDefinition.ragConfiguration?.let { ragConfig ->
-                if (response?.answer?.text.equals(ragConfig.noAnswerSentence, ignoreCase = true)) {
-                    // Save no answer metric
-                    saveRagMetric(IndicatorValues.NO_ANSWER)
-
-                    // Switch to no answer story if configured
-                    if (!ragConfig.noAnswerStoryId.isNullOrBlank()) {
-                        logger.info { "The RAG response is equal to the configured no-answer sentence, so switch to the no-answer story." }
-                        getNoAnswerRAGStory(ragConfig)
-                    } else {
-                        null
-                    }
-                } else {
-                    // Save success metric
-                    saveRagMetric(IndicatorValues.SUCCESS)
-                    null
-                }
-            }
-        }
-    }
-
-    /**
-     * Switch to the configured no-answer story if exists.
-     * Switch to the default unknown story otherwise.
-     * @param ragConfig: The RAG configuration
-     */
-    private fun BotBus.getNoAnswerRAGStory(ragConfig: BotRAGConfiguration): StoryDefinition {
-        val noAnswerStory: StoryDefinition
-        val noAnswerStoryId = ragConfig.noAnswerStoryId
-        if (!noAnswerStoryId.isNullOrBlank()) {
-            logger.info { "A no-answer story $noAnswerStoryId is configured, so run it." }
-            noAnswerStory =
-                botDefinition.findStoryDefinitionById(noAnswerStoryId, connectorId).let {
-                    // Prevent infinite loop when the noAnswerStory is removed or disabled
-                    if (it.id == RAGStoryDefinition.RAG_STORY_NAME) {
-                        logger.info { "The no-answer story is removed or disabled, so run the default unknown story." }
-                        botDefinition.unknownStory
-                    } else {
-                        it
-                    }
-                }
-        } else {
-            logger.info { "No no-answer story is configured, so run the default unknown story." }
-            noAnswerStory = botDefinition.unknownStory
+        return response?.answer?.redirectionIntent?.let {
+            botBus.botDefinition.findStoryDefinition(it, "")
         }
-
-        return noAnswerStory
     }
 
     /**
@@ -235,7 +193,6 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
                                             mapOf(
                                                 "question" to action.toString(),
                                                 "locale" to userPreferences.locale.displayLanguage,
-                                                "no_answer" to ragConfiguration.noAnswerSentence,
                                             ),
                                     ),
                                 embeddingQuestionEmSetting = ragConfiguration.emSetting,
@@ -249,27 +206,36 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
                         debug = action.metadata.debugEnabled || ragConfiguration.debugEnabled,
                     )
 
+                if (response?.answer?.status.equals("not_found_in_context", ignoreCase = true)) {
+                    // Save no answer metric
+                    saveRagMetric(IndicatorValues.NO_ANSWER)
+                } else {
+                    // Save success metric
+                    saveRagMetric(IndicatorValues.SUCCESS)
+                }
+
                 // Handle RAG response
-                return RAGResult(response?.answer, response?.debug, ragStoryRedirection(this, response), response?.observabilityInfo)
+                return RAGResult(
+                    response?.answer,
+                    response?.footnotes,
+                    response?.debug,
+                    ragStoryRedirection(this, response),
+                    response?.observabilityInfo,
+                )
             } catch (exc: Exception) {
                 logger.error { exc }
                 // Save failure metric
                 saveRagMetric(IndicatorValues.FAILURE)
 
-                return if (exc is GenAIOrchestratorBusinessError && exc.error.info.error == "APITimeoutError") {
-                    logger.info { "The APITimeoutError is raised, so switch to the no-answer story." }
-                    RAGResult(noAnswerStory = getNoAnswerRAGStory(ragConfiguration))
-                } else {
-                    RAGResult(
-                        answer = TextWithFootnotes(text = technicalErrorMessage),
-                        debug =
-                            when (exc) {
-                                is GenAIOrchestratorBusinessError -> RAGError(exc.message, exc.error)
-                                is GenAIOrchestratorValidationError -> RAGError(exc.message, exc.detail)
-                                else -> RAGError(errorMessage = exc.message)
-                            },
-                    )
-                }
+                return RAGResult(
+                    answer = LLMAnswer(status = "error", answer = technicalErrorMessage),
+                    debug =
+                        when (exc) {
+                            is GenAIOrchestratorBusinessError -> RAGError(exc.message, exc.error)
+                            is GenAIOrchestratorValidationError -> RAGError(exc.message, exc.detail)
+                            else -> RAGError(errorMessage = exc.message)
+                        },
+                )
             }
         }
     }
@@ -330,9 +296,10 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
  * Aggregation of RAG answer, debug and the no answer Story.
  */
 data class RAGResult(
-    val answer: TextWithFootnotes? = null,
+    val answer: LLMAnswer? = null,
+    val footnotes: List<ai.tock.genai.orchestratorclient.responses.Footnote>? = null,
     val debug: Any? = null,
-    val noAnswerStory: StoryDefinition? = null,
+    val redirectStory: StoryDefinition? = null,
     val observabilityInfo: ObservabilityInfo? = null,
 )
 
diff --git a/bot/storage-mongo/src/test/kotlin/BotRAGConfigurationMongoDAOTest.kt b/bot/storage-mongo/src/test/kotlin/BotRAGConfigurationMongoDAOTest.kt
index 9e4de4a6c1..bad7061d6c 100644
--- a/bot/storage-mongo/src/test/kotlin/BotRAGConfigurationMongoDAOTest.kt
+++ b/bot/storage-mongo/src/test/kotlin/BotRAGConfigurationMongoDAOTest.kt
@@ -59,7 +59,6 @@ internal class BotRAGConfigurationMongoDAOTest : AbstractTest() {
                         model = "modelName1",
                         baseUrl = "https://api.openai.com/v1",
                     ),
-                noAnswerSentence = "no answer sentence",
             )
 
         BotRAGConfigurationMongoDAO.save(config)
@@ -90,7 +89,6 @@ internal class BotRAGConfigurationMongoDAOTest : AbstractTest() {
                         baseUrl = "https://api.openai.com/v1",
                         model = "modelName1",
                     ),
-                noAnswerSentence = "no answer sentence1",
             )
 
         val config2 =
@@ -113,7 +111,6 @@ internal class BotRAGConfigurationMongoDAOTest : AbstractTest() {
                         baseUrl = "https://api.openai.com/v1",
                         model = "modelName1",
                     ),
-                noAnswerSentence = "no answer sentence1",
             )
 
         assertNotEquals(config1, config2)
@@ -121,11 +118,11 @@ internal class BotRAGConfigurationMongoDAOTest : AbstractTest() {
         BotRAGConfigurationMongoDAO.save(config1)
         BotRAGConfigurationMongoDAO.save(config2)
 
-        BotRAGConfigurationMongoDAO.save(config1.copy(noAnswerSentence = "New no answer sentence"))
+        BotRAGConfigurationMongoDAO.save(config1.copy(documentsRequired = true))
 
         val configBDD = BotRAGConfigurationMongoDAO.findByNamespaceAndBotId("namespace1", "botId1")
 
-        assertEquals(config1.copy(noAnswerSentence = "New no answer sentence"), configBDD)
+        assertEquals(config1.copy(documentsRequired = true), configBDD)
     }
 
     @Test
@@ -150,7 +147,6 @@ internal class BotRAGConfigurationMongoDAOTest : AbstractTest() {
                         baseUrl = "https://api.openai.com/v1",
                         model = "modelName1",
                     ),
-                noAnswerSentence = "no answer sentence",
             )
 
         BotRAGConfigurationMongoDAO.save(config)
diff --git a/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/Models.kt b/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/Models.kt
index 09017b9523..95f765d00f 100644
--- a/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/Models.kt
+++ b/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/Models.kt
@@ -16,9 +16,23 @@
 
 package ai.tock.genai.orchestratorclient.responses
 
-data class TextWithFootnotes(
-    val text: String,
-    val footnotes: List<Footnote> = emptyList(),
+data class ChunkSentences(
+    val chunk: String? = null,
+    val sentences: List<String>? = emptyList(),
+    val usedInResponse: Boolean? = false,
+    val reason: String? = null,
+)
+
+data class LLMAnswer(
+    val status: String?,
+    val answer: String?,
+    val confidenceScore: Double? = 0.0,
+    val displayAnswer: Boolean? = true,
+    val topic: String? = null,
+    val suggestedTopics: List<String>? = null,
+    val understanding: String? = null,
+    val redirectionIntent: String? = null,
+    val contextUsage: List<ChunkSentences>? = null,
 )
 
 data class Footnote(
diff --git a/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/RAGResponse.kt b/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/RAGResponse.kt
index a4dfde4c3e..fb699fd7f3 100644
--- a/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/RAGResponse.kt
+++ b/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/RAGResponse.kt
@@ -17,7 +17,8 @@
 package ai.tock.genai.orchestratorclient.responses
 
 data class RAGResponse(
-    val answer: TextWithFootnotes,
+    val answer: LLMAnswer,
+    val footnotes: List<Footnote> = emptyList(),
     val debug: Any? = null,
     val observabilityInfo: ObservabilityInfo? = null,
 )
diff --git a/gen-ai/orchestrator-core/src/main/kotlin/ai/tock/genai/orchestratorcore/models/llm/OllamaLLMSetting.kt b/gen-ai/orchestrator-core/src/main/kotlin/ai/tock/genai/orchestratorcore/models/llm/OllamaLLMSetting.kt
index 430db23ede..295da45356 100644
--- a/gen-ai/orchestrator-core/src/main/kotlin/ai/tock/genai/orchestratorcore/models/llm/OllamaLLMSetting.kt
+++ b/gen-ai/orchestrator-core/src/main/kotlin/ai/tock/genai/orchestratorcore/models/llm/OllamaLLMSetting.kt
@@ -27,5 +27,3 @@ data class OllamaLLMSetting<T>(
         return this.copy(temperature = temperature)
     }
 }
-
-// TODO MASS : Check Compile + TU (car dernier commit)
diff --git a/gen-ai/orchestrator-core/src/main/kotlin/ai/tock/genai/orchestratorcore/utils/VectorStoreUtils.kt b/gen-ai/orchestrator-core/src/main/kotlin/ai/tock/genai/orchestratorcore/utils/VectorStoreUtils.kt
index 379d3a9c4a..0c25fa1aae 100644
--- a/gen-ai/orchestrator-core/src/main/kotlin/ai/tock/genai/orchestratorcore/utils/VectorStoreUtils.kt
+++ b/gen-ai/orchestrator-core/src/main/kotlin/ai/tock/genai/orchestratorcore/utils/VectorStoreUtils.kt
@@ -26,7 +26,7 @@ import ai.tock.shared.property
 private val vectorStore =
     property(
         name = "tock_gen_ai_orchestrator_vector_store_provider",
-        defaultValue = VectorStoreProvider.OpenSearch.name,
+        defaultValue = VectorStoreProvider.PGVector.name,
     )
 
 typealias DocumentIndexName = String
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/models/rag/rag_models.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/models/rag/rag_models.py
index 4ab0519edd..5682c99ee9 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/models/rag/rag_models.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/models/rag/rag_models.py
@@ -28,9 +28,15 @@ class Source(BaseModel):
     """A source model, used to associate document sources with the QA response"""
 
     title: str = Field(description='Source title', examples=['Tock Documentation'])
-    url: Optional[AnyUrl] = Field(description='Source url', examples=['https://doc.tock.ai/tock/'], default=None)
-    content: str = Field(description='Source content', examples=['Tock: The Open Conversation Kit'])
-    score: Optional[float] = Field(description='The compressor score', examples=[0.9149009585380554], default=None)
+    url: Optional[AnyUrl] = Field(
+        description='Source url', examples=['https://doc.tock.ai/tock/'], default=None
+    )
+    content: str = Field(
+        description='Source content', examples=['Tock: The Open Conversation Kit']
+    )
+    score: Optional[float] = Field(
+        description='The compressor score', examples=[0.9149009585380554], default=None
+    )
 
     def __eq__(self, other):
         """
@@ -55,14 +61,77 @@ class Footnote(Source):
     identifier: str = Field(description='Footnote identifier', examples=['1'])
 
 
-class TextWithFootnotes(BaseModel):
-    """Text with its footnotes. Used for RAG response"""
+class ChunkInfos(BaseModel):
+    """A model representing information about a chunk used in the RAG context."""
 
-    text: str = Field(
-        description='Text with footnotes used to list outside sources',
-        examples=['This is page content [1], and this is more content [2]'],
+    chunk: Optional[str] = Field(
+        description='Unique identifier of the chunk.',
+        examples=['cd6d8221-ba9f-44da-86ee-0e25a3c9a5c7'],
+        default=None,
+    )
+    sentences: Optional[List[str]] = Field(
+        description='List of verbatim sentences from the chunk that were used by the LLM.',
+        default=None,
+    )
+    used_in_response: Optional[bool] = Field(
+        description='It indicates whether the chunk is actually used to generate the response.',
+        examples=[True],
+        default=False,
+    )
+    reason: Optional[str] = Field(
+        description='Reason why the chunk was not used (e.g., irrelevant, general background).',
+        default=None,
+    )
+
+
+class LLMAnswer(BaseModel):
+    """
+    A model representing the structured answer generated by the LLM
+    in response to a user query, based on the provided RAG context.
+    """
+
+    status: Optional[str] = Field(
+        description='The status of the answer generation. '
+        "Possible values: 'found_in_context', 'not_found_in_context', 'small_talk', "
+        'or other case-specific codes.',
+        default=None,
+    )
+    answer: Optional[str] = Field(
+        description="The textual answer generated by the LLM, in the user's locale.",
+        default=None,
+    )
+    confidence_score: Optional[float] = Field(
+        description="Confidence score assigned to the answer provided to the user's question",
+        examples=[0.93],
+        default=None,
+    )
+    display_answer: Optional[bool] = Field(
+        description='It indicates if we should display answer or not.',
+        examples=[True],
+        default=False,
+    )
+    topic: Optional[str] = Field(
+        description='The main topic assigned to the answer. Must be one of the predefined list '
+        "of topics, or 'unknown' if no match is possible.",
+        default=None,
+    )
+    suggested_topics: Optional[List[str]] = Field(
+        description='A list of suggested alternative or related topics, '
+        "used when the main topic is 'unknown'.",
+        default=None,
+    )
+    understanding: Optional[str] = Field(
+        description='LLM understanding of user query.',
+        default=None,
+    )
+    redirection_intent: Optional[str] = Field(
+        description='The intent to use to redirect the conversation flow.', default=None
+    )
+    context_usage: Optional[List[ChunkInfos]] = Field(
+        description='The list of chunks from the context that contributed to or were considered '
+        "in the LLM's answer. Each entry contains identifiers, sentences, and reasons.",
+        default=None,
     )
-    footnotes: set[Footnote] = Field(description='Set of footnotes')
 
 
 @unique
@@ -142,10 +211,13 @@ class RAGDebugData(QADebugData):
 
     question_condensing_prompt: Optional[str] = Field(
         description='The prompt of the question rephrased with the history of the conversation.',
-        examples=['Given the following conversation, rephrase the follow up question to be a standalone question.'],
+        examples=[
+            'Given the following conversation, rephrase the follow up question to be a standalone question.'
+        ],
     )
     question_condensing_history: list[ChatMessage] = Field(
-        description="Conversation history, used to reformulate the user's question.")
+        description="Conversation history, used to reformulate the user's question."
+    )
     condensed_question: Optional[str] = Field(
         description='The question rephrased with the history of the conversation.',
         examples=['Hello, how to plan a trip to Morocco ?'],
@@ -156,4 +228,4 @@ class RAGDebugData(QADebugData):
             'Question: Hello, how to plan a trip to Morocco ?. Answer in French.'
         ],
     )
-    answer: str = Field(description='The RAG answer.')
+    answer: LLMAnswer = Field(description='The RAG answer.')
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/rag_router.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/rag_router.py
index 4cc8f961a8..ecc8f500eb 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/rag_router.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/rag_router.py
@@ -40,21 +40,27 @@
 
 
 @rag_router.post('')
-async def ask_rag(http_request: Request, request: RAGRequest, debug: bool = False) -> RAGResponse:
+async def ask_rag(
+    http_request: Request, request: RAGRequest, debug: bool = False
+) -> RAGResponse:
     """
     ## Ask a RAG System
     Ask question to a RAG System, and return answer by using a knowledge base (documents)
     """
     # Check the consistency of the Vector Store Provider with the request body
-    validate_vector_store_rag_query(http_request, request.vector_store_setting, request.document_search_params)
+    validate_vector_store_rag_query(
+        http_request, request.vector_store_setting, request.document_search_params
+    )
 
     # execute RAG
     return await rag(request, debug)
 
+
 def validate_vector_store_rag_query(
-        http_request: Request,
-        vector_store_setting: VectorStoreSetting,
-        vector_store_search_params: DocumentSearchParams):
+    http_request: Request,
+    vector_store_setting: VectorStoreSetting,
+    vector_store_search_params: DocumentSearchParams,
+):
     """
     Check the consistency of the Vector Store Provider with the request body
     Args:
@@ -73,11 +79,15 @@ def validate_vector_store_rag_query(
         vector_store_provider = vector_store_setting.provider
 
     if vector_store_provider != vector_store_search_params.provider:
-        logger.error('Inconsistency between vector store provider and document search parameters (%s Vs %s)',
-                     vector_store_provider.value, vector_store_search_params.provider.value)
+        logger.error(
+            'Inconsistency between vector store provider and document search parameters (%s Vs %s)',
+            vector_store_provider.value,
+            vector_store_search_params.provider.value,
+        )
         raise AIProviderBadRequestException(
             create_error_info_bad_request(
-                request=http_request,
+                http_request=http_request,
                 provider=vector_store_search_params.provider,
-                cause='Inconsistency between vector store provider and document search parameters')
+                cause='Inconsistency between vector store provider and document search parameters',
+            )
         )
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/responses/responses.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/responses/responses.py
index 50d52e57ef..1b7b2ab60b 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/responses/responses.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/responses/responses.py
@@ -30,7 +30,7 @@
 from gen_ai_orchestrator.models.observability.observability_provider import (
     ObservabilityProvider,
 )
-from gen_ai_orchestrator.models.rag.rag_models import Source, TextWithFootnotes
+from gen_ai_orchestrator.models.rag.rag_models import Source, LLMAnswer, Footnote
 from gen_ai_orchestrator.models.vector_stores.vectore_store_provider import (
     VectorStoreProvider,
 )
@@ -127,9 +127,10 @@ class ObservabilityInfo(BaseModel):
 class RAGResponse(BaseModel):
     """The RAG response model"""
 
-    answer: TextWithFootnotes = Field(
-        description='The RAG answer, with outside sources.'
+    answer: Optional[LLMAnswer] = Field(
+        description='The RAG answer'
     )
+    footnotes: set[Footnote] = Field(description='Set of footnotes')
     debug: Optional[Any] = Field(
         description='Debug data',
         examples=[{'action': 'retrieve', 'result': 'OK', 'errors': []}],
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/factories/callback_handlers/callback_handlers_factory.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/factories/callback_handlers/callback_handlers_factory.py
index 45bb6c5293..81e1dd40b1 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/factories/callback_handlers/callback_handlers_factory.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/factories/callback_handlers/callback_handlers_factory.py
@@ -49,5 +49,3 @@ def check_observability_setting(self) -> bool:
         :raises BusinessException: For incorrect setting
         """
         pass
-
-
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/factories/callback_handlers/langfuse_callback_handler_factory.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/factories/callback_handlers/langfuse_callback_handler_factory.py
index 737f391e9d..f2c0900e22 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/factories/callback_handlers/langfuse_callback_handler_factory.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/factories/callback_handlers/langfuse_callback_handler_factory.py
@@ -120,9 +120,7 @@ def check_observability_setting(self) -> bool:
 
         except ApiError as exc:
             logger.error(exc)
-            raise GenAIObservabilityErrorException(
-                create_error_info_langfuse(exc)
-            )
+            raise GenAIObservabilityErrorException(create_error_info_langfuse(exc))
         return True
 
     def _fetch_settings(self) -> dict:
@@ -140,7 +138,10 @@ def _fetch_settings(self) -> dict:
 
     def _get_httpx_client(self) -> Optional[Client]:
         langfuse_settings = self._fetch_settings()
-        if ProxyServerType.AWS_LAMBDA == application_settings.observability_proxy_server:
+        if (
+            ProxyServerType.AWS_LAMBDA
+            == application_settings.observability_proxy_server
+        ):
             """
             This AWSLambda proxy is used when the architecture implemented for the Langfuse
             observability tool places it behind an API Gateway which requires its
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py
index 82f8b55246..df5ac562ce 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py
@@ -17,10 +17,12 @@
 It uses LangChain to perform a Conversational Retrieval Chain
 """
 
+import json
 import logging
 import time
 from functools import partial
 from logging import ERROR, WARNING
+from operator import itemgetter
 from typing import List, Optional
 
 from langchain.retrievers.contextual_compression import (
@@ -29,10 +31,13 @@
 from langchain_community.chat_message_histories import ChatMessageHistory
 from langchain_core.callbacks import BaseCallbackHandler
 from langchain_core.documents import Document
-from langchain_core.output_parsers import StrOutputParser
+from langchain_core.messages import AIMessage, HumanMessage
+from langchain_core.output_parsers import JsonOutputParser, StrOutputParser
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 from langchain_core.prompts import PromptTemplate as LangChainPromptTemplate
 from langchain_core.runnables import (
+    RunnableConfig,
+    RunnableLambda,
     RunnableParallel,
     RunnablePassthrough,
     RunnableSerializable,
@@ -63,10 +68,10 @@
 from gen_ai_orchestrator.models.rag.rag_models import (
     ChatMessageType,
     Footnote,
+    LLMAnswer,
     RAGDebugData,
     RAGDocument,
     RAGDocumentMetadata,
-    TextWithFootnotes,
 )
 from gen_ai_orchestrator.routers.requests.requests import RAGRequest
 from gen_ai_orchestrator.routers.responses.responses import RAGResponse
@@ -104,7 +109,7 @@ async def execute_rag_chain(
     Args:
         request: The RAG request
         debug: True if RAG data debug should be returned with the response.
-        custom_observability_handler: Custom observability handler
+        custom_observability_handler: Custom observability handler (Used in the tooling run_experiment.py script)
     Returns:
         The RAG response (Answer and document sources)
     """
@@ -112,7 +117,9 @@ async def execute_rag_chain(
     logger.info('RAG chain - Start of execution...')
     start_time = time.time()
 
-    conversational_retrieval_chain = create_rag_chain(request=request)
+    conversational_retrieval_chain = create_rag_chain(
+        request=request, vector_db_async_mode=False  # TODO MASS
+    )
 
     message_history = ChatMessageHistory()
     session_id = None
@@ -173,39 +180,44 @@ async def execute_rag_chain(
             metadata=metadata,
         ),
     )
+    llm_answer = LLMAnswer(**response['answer'])
 
     # RAG Guard
-    rag_guard(inputs, response, request.documents_required)
+    rag_guard(inputs, llm_answer, response, request.documents_required)
 
     # Guardrail
     if request.guardrail_setting:
         guardrail = get_guardrail_factory(
             setting=request.guardrail_setting
         ).get_parser()
-        guardrail_output = guardrail.parse(response['answer'])
+        guardrail_output = guardrail.parse(llm_answer.answer)
         check_guardrail_output(guardrail_output)
 
     # Calculation of RAG processing time
     rag_duration = '{:.2f}'.format(time.time() - start_time)
     logger.info('RAG chain - End of execution. (Duration : %s seconds)', rag_duration)
 
+    # Group contexts by chunk id
+    contexts_by_chunk = {
+        ctx.chunk: ctx
+        for ctx in (llm_answer.context_usage or [])
+        if ctx.used_in_response
+    }
+
     # Returning RAG response
     return RAGResponse(
-        answer=TextWithFootnotes(
-            text=response['answer'],
-            footnotes=set(
-                map(
-                    lambda doc: Footnote(
-                        identifier=doc.metadata['id'],
-                        title=doc.metadata['title'],
-                        url=doc.metadata['source'],
-                        content=get_source_content(doc),
-                        score=doc.metadata.get('retriever_score', None),
-                    ),
-                    response['documents'],
-                )
-            ),
-        ),
+        answer=llm_answer,
+        footnotes={
+            Footnote(
+                identifier=doc.metadata['id'],
+                title=doc.metadata['title'],
+                url=doc.metadata['source'],
+                content=get_source_content(doc),
+                score=doc.metadata.get('retriever_score', None),
+            )
+            for doc in response['documents']
+            if doc.metadata['id'] in contexts_by_chunk
+        },
         observability_info=get_observability_info(
             observability_handler,
             ObservabilityTrace.RAG.value if observability_handler is not None else None,
@@ -284,31 +296,75 @@ def create_rag_chain(
     if question_condensing_llm_factory is not None:
         question_condensing_llm = question_condensing_llm_factory.get_language_model()
     question_answering_llm = question_answering_llm_factory.get_language_model()
-    rag_prompt = build_rag_prompt(request)
 
-    # Construct the RAG chain using the prompt and LLM,
-    # This chain will consume the documents retrieved by the retriever as input.
-    rag_chain = construct_rag_chain(question_answering_llm, rag_prompt)
+    # Fallback in case of missing condensing LLM setting using the answering LLM setting.
+    if question_condensing_llm is not None:
+        condensing_llm = question_condensing_llm
+    else:
+        condensing_llm = question_answering_llm
 
     # Build the chat chain for question contextualization
     chat_chain = build_question_condensation_chain(
-        question_condensing_llm
-        if question_condensing_llm is not None
-        else question_answering_llm,
-        request.question_condensing_prompt,
+        condensing_llm, request.question_condensing_prompt
     )
+    rag_prompt = build_rag_prompt(request)
 
     # Function to contextualize the question based on chat history
     contextualize_question_fn = partial(contextualize_question, chat_chain=chat_chain)
 
-    # Final RAG chain with retriever and source documents
-    rag_chain_with_retriever = (
-        contextualize_question_fn
-        | RunnableParallel({'documents': retriever, 'question': RunnablePassthrough()})
-        | RunnablePassthrough.assign(answer=rag_chain)
+    # Calculate the condensed question
+    with_condensed_question = RunnableParallel(
+        {
+            'condensed_question': contextualize_question_fn,
+            'question': itemgetter('question'),
+            'chat_history': itemgetter('chat_history'),
+        }
+    )
+
+    def retrieve_with_variants(inputs):
+        variants = [
+            # inputs["question"], Deactivated. It's an example to prove the multi retriever process
+            inputs['condensed_question']
+        ]
+        docs = []
+        for v in variants:
+            docs.extend(retriever.invoke(v))
+        # Deduplicate docs
+        unique_docs = {d.metadata['id']: d for d in docs}
+
+        # TODO [DERCBOT-1649] Apply the RRF Algo on unique_docs.
+        return list(unique_docs.values())
+
+    # Build the RAG inputs
+    rag_inputs = with_condensed_question | RunnableParallel(
+        {
+            'question': itemgetter('condensed_question'),
+            'chat_history': itemgetter('chat_history'),
+            'documents': RunnableLambda(retrieve_with_variants),
+        }
     )
 
-    return rag_chain_with_retriever
+    return rag_inputs | RunnablePassthrough.assign(
+        answer=(
+            {
+                'context': lambda x: json.dumps(
+                    [
+                        {
+                            'chunk_id': doc.metadata['id'],
+                            'chunk_text': doc.page_content,
+                        }
+                        for doc in x['documents']
+                    ],
+                    ensure_ascii=False,
+                    indent=2,
+                ),
+                'chat_history': format_chat_history,
+            }
+            | rag_prompt
+            | question_answering_llm
+            | JsonOutputParser(pydantic_object=LLMAnswer, name='rag_chain_output')
+        )
+    )
 
 
 def build_rag_prompt(request: RAGRequest) -> LangChainPromptTemplate:
@@ -322,23 +378,14 @@ def build_rag_prompt(request: RAGRequest) -> LangChainPromptTemplate:
     )
 
 
-def construct_rag_chain(llm, rag_prompt):
-    """
-    Construct the RAG chain from LLM and prompt.
-    """
-    return (
-        {
-            'context': lambda inputs: '\n\n'.join(
-                doc.page_content for doc in inputs['documents']
-            ),
-            'question': lambda inputs: inputs[
-                'question'
-            ],  # Override the user's original question with the condensed one
-        }
-        | rag_prompt
-        | llm
-        | StrOutputParser(name='rag_chain_output')
-    )
+def format_chat_history(x):
+    messages = []
+    for msg in x['chat_history']:
+        if isinstance(msg, HumanMessage):
+            messages.append({'user': msg.content})
+        elif isinstance(msg, AIMessage):
+            messages.append({'assistant': msg.content})
+    return json.dumps(messages, ensure_ascii=False, indent=2)
 
 
 def build_question_condensation_chain(
@@ -347,14 +394,27 @@ def build_question_condensation_chain(
     """
     Build the chat chain for contextualizing questions.
     """
+    # TODO deprecated : All Gen configurations are supposed to have this prompt now. It is mandatory in the RAG configuration.
     if prompt is None:
         # Default prompt
         prompt = PromptTemplate(
             formatter=PromptFormatter.F_STRING,
             inputs={},
-            template='Given a chat history and the latest user question which might reference context in \
-the chat history, formulate a standalone question which can be understood without the chat history. \
-Do NOT answer the question, just reformulate it if needed and otherwise return it as is.',
+            template="""
+You are a helpful assistant that reformulates questions.
+
+You are given:
+- The conversation history between the user and the assistant
+- The most recent user question
+
+Your task:
+- Reformulate the user’s latest question into a clear, standalone query.
+- Incorporate relevant context from the conversation history.
+- Do NOT answer the question.
+- If the history does not provide additional context, keep the question as is.
+
+Return only the reformulated question.
+""",
         )
 
     return (
@@ -379,50 +439,55 @@ def contextualize_question(inputs: dict, chat_chain) -> str:
     return inputs['question']
 
 
-def rag_guard(inputs, response, documents_required):
+def rag_guard(question, answer, response, documents_required):
     """
     Validates the RAG system's response based on the presence or absence of source documents
     and the `documentsRequired` setting.
 
     Args:
-        inputs: question answering prompt inputs
+        question: user question
+        answer: the LLM answer
         response: the RAG response
         documents_required (bool): Specifies whether documents are mandatory for the response.
     """
 
-    no_docs_retrieved = response['documents'] == []
-    no_docs_but_required = no_docs_retrieved and documents_required
-    chain_can_give_no_answer_reply = 'no_answer' in inputs
-    chain_reply_no_answer = False
-
-    if chain_can_give_no_answer_reply:
-        chain_reply_no_answer = response['answer'] == inputs['no_answer']
-
-    if no_docs_but_required:
-        if chain_can_give_no_answer_reply and chain_reply_no_answer:
-            # We expect the chain to use its non-response value, and it has done so, which is the expected behavior.
-            return
-        # Everything else isn't expected
-        message = 'The RAG system cannot provide an answer when no documents are found and documents are required'
-        rag_log(level=ERROR, message=message, inputs=inputs, response=response)
+    if (
+        documents_required
+        and answer.status == 'found_in_context'
+        and len(response['documents']) == 0
+    ):
+        message = 'No documents were retrieved, yet an answer was attempted.'
+        rag_log(
+            level=ERROR,
+            message=message,
+            question=question,
+            answer=answer.answer,
+            response=response,
+        )
         raise GenAIGuardCheckException(ErrorInfo(cause=message))
 
-    if chain_reply_no_answer and not no_docs_retrieved:
-        # If the chain responds with its non-response value and the documents are retrieved,
-        # so we remove them from the RAG response.
-        message = 'The RAG gives no answer for user question, but some documents has been found!'
-        rag_log(level=WARNING, message=message, inputs=inputs, response=response)
+    if answer.status == 'not_found_in_context' and len(response['documents']) > 0:
+        # If the answer is not found in context and some documents are retrieved, so we remove them from the RAG response.
+        message = 'No answer found in the retrieved context. The documents are therefore removed from the RAG response.'
+        rag_log(
+            level=WARNING,
+            message=message,
+            question=question,
+            answer=answer.answer,
+            response=response,
+        )
         response['documents'] = []
 
 
-def rag_log(level, message, inputs, response):
+def rag_log(level, message, question, answer, response):
     """
     RAG logging
 
     Args:
         level: logging level
         message: message to log
-        inputs: question answering prompt inputs
+        question: question answering prompt inputs
+        answer: LLM answer
         response: the RAG response
     """
 
@@ -432,9 +497,9 @@ def rag_log(level, message, inputs, response):
         'RAG chain - question="%(question)s", answer="%(answer)s", documents="%(documents)s"',
         {
             'message': message,
-            'question': inputs['question'],
-            'answer': response['answer'],
-            'documents': response['documents'],
+            'question': question,
+            'answer': answer,
+            'documents': len(response['documents']),
         },
     )
 
@@ -447,6 +512,9 @@ def get_rag_documents(handler: RAGCallbackHandler) -> List[RAGDocument]:
         handler: the RAG Callback Handler
     """
 
+    if handler.records['documents'] is None:
+        return []
+
     return [
         # Get first 100 char of content
         RAGDocument(
@@ -457,6 +525,17 @@ def get_rag_documents(handler: RAGCallbackHandler) -> List[RAGDocument]:
     ]
 
 
+def get_llm_answer(rag_chain_output) -> LLMAnswer:
+    if rag_chain_output is None:
+        return LLMAnswer()
+
+    return LLMAnswer(
+        **json.loads(
+            rag_chain_output.strip().removeprefix('```json').removesuffix('```').strip()
+        )
+    )
+
+
 def get_rag_debug_data(
     request: RAGRequest, records_callback_handler: RAGCallbackHandler, rag_duration
 ) -> RAGDebugData:
@@ -475,7 +554,7 @@ def get_rag_debug_data(
         documents=get_rag_documents(records_callback_handler),
         document_index_name=request.document_index_name,
         document_search_params=request.document_search_params,
-        answer=records_callback_handler.records['rag_chain_output'],
+        answer=get_llm_answer(records_callback_handler.records['rag_chain_output']),
         duration=rag_duration,
     )
 
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/observability/observabilty_service.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/observability/observabilty_service.py
index ceb33e0c1b..f81bcd57a9 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/observability/observabilty_service.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/observability/observabilty_service.py
@@ -42,7 +42,9 @@ def check_observability_setting(setting: ObservabilitySetting) -> bool:
          True for a valid Observability setting. Raise exception otherwise.
     """
 
-    logger.info('Get the Callback handler Factory, then check the Observability setting.')
+    logger.info(
+        'Get the Callback handler Factory, then check the Observability setting.'
+    )
     return get_callback_handler_factory(setting).check_observability_setting()
 
 
diff --git a/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py b/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py
index 6aeedf86a6..000e483226 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py
@@ -29,6 +29,7 @@
 from gen_ai_orchestrator.models.guardrail.bloomz.bloomz_guardrail_setting import (
     BloomzGuardrailSetting,
 )
+from gen_ai_orchestrator.models.rag.rag_models import LLMAnswer
 from gen_ai_orchestrator.routers.requests.requests import RAGRequest
 from gen_ai_orchestrator.services.langchain import rag_chain
 from gen_ai_orchestrator.services.langchain.factories.langchain_factory import (
@@ -43,19 +44,25 @@
 )
 
 
-@patch('gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post')
-@patch('gen_ai_orchestrator.services.langchain.factories.langchain_factory.get_compressor_factory')
-@patch('gen_ai_orchestrator.services.langchain.factories.langchain_factory.get_callback_handler_factory')
+@patch(
+    'gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post'
+)
+@patch(
+    'gen_ai_orchestrator.services.langchain.factories.langchain_factory.get_compressor_factory'
+)
+@patch(
+    'gen_ai_orchestrator.services.langchain.factories.langchain_factory.get_callback_handler_factory'
+)
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.create_rag_chain')
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.RAGCallbackHandler')
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_guard')
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.RAGResponse')
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.TextWithFootnotes')
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.RAGDebugData')
+@patch('gen_ai_orchestrator.services.langchain.rag_chain.get_llm_answer')
 @pytest.mark.asyncio
 async def test_rag_chain(
+    mocked_get_llm_answer,
     mocked_rag_debug_data,
-    mocked_text_with_footnotes,
     mocked_rag_response,
     mocked_rag_guard,
     mocked_callback_init,
@@ -70,13 +77,19 @@ async def test_rag_chain(
         'dialog': {
             'history': [
                 {'text': 'Hello, how can I do this?', 'type': 'HUMAN'},
-                {'text': 'you can do this with the following method ....', 'type': 'AI'}
+                {
+                    'text': 'you can do this with the following method ....',
+                    'type': 'AI',
+                },
             ],
-            'tags': []
+            'tags': [],
         },
         'question_answering_llm_setting': {
             'provider': 'OpenAI',
-            'api_key': {'type': 'Raw', 'secret': 'ab7***************************A1IV4B'},
+            'api_key': {
+                'type': 'Raw',
+                'secret': 'ab7***************************A1IV4B',
+            },
             'temperature': 1.2,
             'model': 'gpt-3.5-turbo',
         },
@@ -92,15 +105,18 @@ async def test_rag_chain(
 {question}
 
 Answer in {locale}:""",
-            'inputs' : {
+            'inputs': {
                 'question': 'How to get started playing guitar ?',
                 'no_answer': 'Sorry, I don t know.',
                 'locale': 'French',
-            }
+            },
         },
         'embedding_question_em_setting': {
             'provider': 'OpenAI',
-            'api_key': {'type': 'Raw', 'secret': 'ab7***************************A1IV4B'},
+            'api_key': {
+                'type': 'Raw',
+                'secret': 'ab7***************************A1IV4B',
+            },
             'model': 'text-embedding-ada-002',
         },
         'document_index_name': 'my-index-name',
@@ -154,11 +170,23 @@ async def test_rag_chain(
             AIMessage(content='you can do this with the following method ....'),
         ],
     }
-    docs = [Document(
-        page_content='some page content',
-        metadata={'id':'123-abc', 'title':'my-title', 'source': None},
-    )]
-    response = {'answer': 'an answer from llm', 'documents': docs}
+    docs = [
+        Document(
+            page_content='some page content',
+            metadata={'id': '123-abc', 'title': 'my-title', 'source': None},
+        )
+    ]
+    response = {
+        'answer': {
+            'status': '',
+            'answer': 'an answer from llm',
+            'topic': None,
+            'suggested_topics': None,
+            'context': [],
+        },
+        'documents': docs,
+    }
+    llm_answer = LLMAnswer(**response['answer'])
 
     # Setup mock factories/init return value
     observability_factory_instance = mocked_get_callback_handler_factory.return_value
@@ -191,27 +219,26 @@ async def test_rag_chain(
     )
     # Assert the response is build using the expected settings
     mocked_rag_response.assert_called_once_with(
-        # TextWithFootnotes must be mocked or mapping the footnotes will fail
-        answer=mocked_text_with_footnotes(
-            text=mocked_rag_answer['answer'], footnotes=[]
-        ),
+        answer=llm_answer,
+        footnotes=set(),
         debug=mocked_rag_debug_data(request, mocked_rag_answer, mocked_callback, 1),
-        observability_info=None
-    )
-    mocked_get_document_compressor_factory(
-        setting=request.compressor_setting
+        observability_info=None,
     )
+    mocked_get_document_compressor_factory(setting=request.compressor_setting)
     # Assert the rag guardrail is called
     mocked_guardrail_parse.assert_called_once_with(
         os.path.join(request.guardrail_setting.api_base, 'guardrail'),
-        json={'text': [mocked_rag_answer['answer']]},
+        json={'text': [mocked_rag_answer['answer']['answer']]},
     )
     # Assert the rag guard is called
     mocked_rag_guard.assert_called_once_with(
-        inputs, response, request.documents_required
+        inputs, llm_answer, response, request.documents_required
     )
 
-@patch('gen_ai_orchestrator.services.langchain.impls.guardrail.bloomz_guardrail.requests.post')
+
+@patch(
+    'gen_ai_orchestrator.services.langchain.impls.guardrail.bloomz_guardrail.requests.post'
+)
 def test_guardrail_parse_succeed_with_toxicities_encountered(
     mocked_guardrail_response,
 ):
@@ -220,7 +247,15 @@ def test_guardrail_parse_succeed_with_toxicities_encountered(
             provider='BloomzGuardrail', max_score=0.5, api_base='http://test-guard.com'
         )
     ).get_parser()
-    rag_response = {'answer': 'This is a sample text.'}
+    rag_response = {
+        'answer': {
+            'status': '',
+            'answer': 'This is a sample text.',
+            'topic': None,
+            'suggested_topics': None,
+            'context': [],
+        }
+    }
 
     mocked_response = MagicMock()
     mocked_response.status_code = 200
@@ -236,11 +271,11 @@ def test_guardrail_parse_succeed_with_toxicities_encountered(
     }
 
     mocked_guardrail_response.return_value = mocked_response
-    guardrail_output = guardrail.parse(rag_response['answer'])
+    guardrail_output = guardrail.parse(rag_response['answer']['answer'])
 
     mocked_guardrail_response.assert_called_once_with(
         os.path.join(guardrail.endpoint, 'guardrail'),
-        json={'text': [rag_response['answer']]},
+        json={'text': [rag_response['answer']['answer']]},
     )
     assert guardrail_output == {
         'content': 'This is a sample text.',
@@ -249,14 +284,24 @@ def test_guardrail_parse_succeed_with_toxicities_encountered(
     }
 
 
-@patch('gen_ai_orchestrator.services.langchain.impls.guardrail.bloomz_guardrail.requests.post')
+@patch(
+    'gen_ai_orchestrator.services.langchain.impls.guardrail.bloomz_guardrail.requests.post'
+)
 def test_guardrail_parse_fail(mocked_guardrail_response):
     guardrail = get_guardrail_factory(
         BloomzGuardrailSetting(
             provider='BloomzGuardrail', max_score=0.5, api_base='http://test-guard.com'
         )
     ).get_parser()
-    rag_response = {'answer': 'This is a sample text.'}
+    rag_response = {
+        'answer': {
+            'status': '',
+            'answer': 'This is a sample text.',
+            'topic': None,
+            'suggested_topics': None,
+            'context': [],
+        }
+    }
 
     mocked_response = MagicMock()
     mocked_response.status_code = 500
@@ -266,15 +311,17 @@ def test_guardrail_parse_fail(mocked_guardrail_response):
         HTTPError,
         match=f"Error {mocked_response.status_code}. Bloomz guardrail didn't respond as expected.",
     ):
-        guardrail.parse(rag_response['answer'])
+        guardrail.parse(rag_response['answer']['answer'])
 
     mocked_guardrail_response.assert_called_once_with(
         os.path.join(guardrail.endpoint, 'guardrail'),
-        json={'text': [rag_response['answer']]},
+        json={'text': [rag_response['answer']['answer']]},
     )
 
 
-@patch('gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post')
+@patch(
+    'gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post'
+)
 def test_compress_documents_should_succeed(mocked_rerank):
     bloomz_reranker = BloomzRerank(label='entailement', endpoint='http://example.com')
     documents = [
@@ -351,7 +398,9 @@ def test_compress_documents_should_succeed(mocked_rerank):
     ]
 
 
-@patch('gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post')
+@patch(
+    'gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post'
+)
 def test_compress_documents_with_unknown_label(mocked_rerank):
     bloomz_reranker = BloomzRerank(label='unknown_label', endpoint='http://example.com')
     documents = [
@@ -414,59 +463,87 @@ def test_check_guardrail_output_is_ok():
 
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
 def test_rag_guard_fails_if_no_docs_in_valid_answer(mocked_log):
-    inputs = {'no_answer': "Sorry, I don't know."}
+    question = 'Hi!'
     response = {
-        'answer': 'a valid answer',
+        'answer': {'status': 'found_in_context', 'answer': 'a valid answer'},
         'documents': [],
     }
     try:
-        rag_chain.rag_guard(inputs, response,documents_required=True)
+        rag_chain.rag_guard(
+            question, LLMAnswer(**response['answer']), response, documents_required=True
+        )
     except Exception as e:
         assert isinstance(e, GenAIGuardCheckException)
 
 
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
 def test_rag_guard_accepts_no_answer_even_with_docs(mocked_log):
-    inputs = {'no_answer': "Sorry, I don't know."}
+    question = 'Hi!'
     response = {
-        'answer': "Sorry, I don't know.",
+        'answer': {
+            'status': 'not_found_in_context',
+            'answer': 'Sorry, I don t know.',
+            'context': [
+                {
+                    'chunk': 1,
+                    'sentences': ['str1'],
+                }
+            ],
+        },
         'documents': ['a doc as a string'],
     }
-    rag_chain.rag_guard(inputs, response, documents_required=True)
+    rag_chain.rag_guard(
+        question, LLMAnswer(**response['answer']), response, documents_required=True
+    )
+    # No answer found in the retrieved context. The documents are therefore removed from the RAG response.
     assert response['documents'] == []
 
 
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
 def test_rag_guard_valid_answer_with_docs(mocked_log):
-    inputs = {'no_answer': "Sorry, I don't know."}
+    question = 'Hi!'
     response = {
-        'answer': 'a valid answer',
+        'answer': {
+            'status': 'found_in_context',
+            'answer': 'a valid answer',
+        },
         'documents': ['doc1', 'doc2'],
     }
-    rag_chain.rag_guard(inputs, response, documents_required=True)
+    rag_chain.rag_guard(
+        question, LLMAnswer(**response['answer']), response, documents_required=True
+    )
     assert response['documents'] == ['doc1', 'doc2']
 
+
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
 def test_rag_guard_no_answer_with_no_docs(mocked_log):
-    inputs = {'no_answer': "Sorry, I don't know."}
+    question = 'Hi!'
     response = {
-        'answer': "Sorry, I don't know.",
+        'answer': {'status': 'not_found_in_context', 'answer': 'Sorry, I don t know.'},
         'documents': [],
     }
-    rag_chain.rag_guard(inputs, response, documents_required=True)
+    rag_chain.rag_guard(
+        question, LLMAnswer(**response['answer']), response, documents_required=True
+    )
     assert response['documents'] == []
 
+
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
 def test_rag_guard_without_no_answer_input(mocked_log):
     """Test that __rag_guard handles missing no_answer input correctly."""
-    inputs = {}  # No 'no_answer' key
+    question = 'Hi!'
     response = {
-        'answer': 'some answer',
+        'answer': {
+            'status': 'found_in_context',
+            'answer': 'a valid answer',
+        },
         'documents': [],
     }
     with pytest.raises(GenAIGuardCheckException) as exc:
-        rag_chain.rag_guard(inputs, response, documents_required=True)
+        rag_chain.rag_guard(
+            question, LLMAnswer(**response['answer']), response, documents_required=True
+        )
 
     mocked_log.assert_called_once()
 
-    assert isinstance(exc.value, GenAIGuardCheckException)
\ No newline at end of file
+    assert isinstance(exc.value, GenAIGuardCheckException)

From b4facf748b42b37c7a8677ebb555dcfc2a897ef3 Mon Sep 17 00:00:00 2001
From: Mohamed ASSOUKTI <mohamed.assoukti@partnre.com>
Date: Tue, 3 Feb 2026 17:08:21 +0100
Subject: [PATCH 2/7] [DERCBOT-1609] w

---
 .../model/genai/BotRAGConfigurationDTO.kt     |   3 -
 .../src/test/kotlin/service/RAGServiceTest.kt |   1 -
 .../service/RAGValidationServiceTest.kt       |   1 -
 .../admin/bot/rag/BotRAGConfiguration.kt      |   1 -
 .../kotlin/engine/config/RAGAnswerHandler.kt  |   1 -
 .../kotlin/BotRAGConfigurationMongoDAOTest.kt |   4 +-
 .../orchestratorclient/requests/RAGRequest.kt |   1 -
 .../routers/requests/requests.py              |  39 +++---
 .../services/langchain/rag_chain.py           | 104 +++++----------
 .../server/tests/services/test_rag_chain.py   | 119 +++---------------
 10 files changed, 71 insertions(+), 203 deletions(-)

diff --git a/bot/admin/server/src/main/kotlin/model/genai/BotRAGConfigurationDTO.kt b/bot/admin/server/src/main/kotlin/model/genai/BotRAGConfigurationDTO.kt
index 5afb90aba9..adef95304d 100644
--- a/bot/admin/server/src/main/kotlin/model/genai/BotRAGConfigurationDTO.kt
+++ b/bot/admin/server/src/main/kotlin/model/genai/BotRAGConfigurationDTO.kt
@@ -42,7 +42,6 @@ data class BotRAGConfigurationDTO(
     val emSetting: EMSettingDTO,
     val indexSessionId: String? = null,
     val indexName: String? = null,
-    val documentsRequired: Boolean = true,
     val debugEnabled: Boolean,
     val maxDocumentsRetrieved: Int,
     val maxMessagesFromHistory: Int,
@@ -61,7 +60,6 @@ data class BotRAGConfigurationDTO(
         emSetting = configuration.emSetting.toDTO(),
         indexSessionId = configuration.indexSessionId,
         indexName = configuration.generateIndexName(),
-        documentsRequired = configuration.documentsRequired,
         debugEnabled = configuration.debugEnabled,
         maxDocumentsRetrieved = configuration.maxDocumentsRetrieved,
         maxMessagesFromHistory = configuration.maxMessagesFromHistory,
@@ -97,7 +95,6 @@ data class BotRAGConfigurationDTO(
                     dto = emSetting,
                 ),
             indexSessionId = indexSessionId,
-            documentsRequired = documentsRequired,
             debugEnabled = debugEnabled,
             maxDocumentsRetrieved = maxDocumentsRetrieved,
             maxMessagesFromHistory = maxMessagesFromHistory,
diff --git a/bot/admin/server/src/test/kotlin/service/RAGServiceTest.kt b/bot/admin/server/src/test/kotlin/service/RAGServiceTest.kt
index 698f4c5381..d34d8ac5c0 100644
--- a/bot/admin/server/src/test/kotlin/service/RAGServiceTest.kt
+++ b/bot/admin/server/src/test/kotlin/service/RAGServiceTest.kt
@@ -97,7 +97,6 @@ class RAGServiceTest : AbstractTest() {
                         model = "model",
                         apiBase = "url",
                     ),
-                documentsRequired = true,
                 debugEnabled = false,
                 maxDocumentsRetrieved = 2,
                 maxMessagesFromHistory = 2,
diff --git a/bot/admin/server/src/test/kotlin/service/RAGValidationServiceTest.kt b/bot/admin/server/src/test/kotlin/service/RAGValidationServiceTest.kt
index edff4ee726..bfa0dd9631 100644
--- a/bot/admin/server/src/test/kotlin/service/RAGValidationServiceTest.kt
+++ b/bot/admin/server/src/test/kotlin/service/RAGValidationServiceTest.kt
@@ -90,7 +90,6 @@ class RAGValidationServiceTest {
             questionAnsweringLlmSetting = openAILLMSetting,
             questionAnsweringPrompt = PromptTemplate(template = "How to bike in the rain"),
             emSetting = azureOpenAIEMSetting,
-            documentsRequired = true,
             debugEnabled = false,
             maxDocumentsRetrieved = 2,
             maxMessagesFromHistory = 2,
diff --git a/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt b/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt
index 6f2c1af38e..18b6552f48 100644
--- a/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt
+++ b/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt
@@ -35,7 +35,6 @@ data class BotRAGConfiguration(
     val llmSetting: LLMSetting? = null,
     val emSetting: EMSetting,
     val indexSessionId: String? = null,
-    val documentsRequired: Boolean = true,
     val debugEnabled: Boolean = false,
     val maxDocumentsRetrieved: Int = 4,
     val maxMessagesFromHistory: Int = 5,
diff --git a/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt b/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt
index 5f8d5c0fa1..d37fa6b0f9 100644
--- a/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt
+++ b/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt
@@ -201,7 +201,6 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
                                 compressorSetting = botDefinition.documentCompressorConfiguration?.setting,
                                 vectorStoreSetting = vectorStoreSetting,
                                 observabilitySetting = botDefinition.observabilityConfiguration?.setting,
-                                documentsRequired = ragConfiguration.documentsRequired,
                             ),
                         debug = action.metadata.debugEnabled || ragConfiguration.debugEnabled,
                     )
diff --git a/bot/storage-mongo/src/test/kotlin/BotRAGConfigurationMongoDAOTest.kt b/bot/storage-mongo/src/test/kotlin/BotRAGConfigurationMongoDAOTest.kt
index bad7061d6c..bc55f21568 100644
--- a/bot/storage-mongo/src/test/kotlin/BotRAGConfigurationMongoDAOTest.kt
+++ b/bot/storage-mongo/src/test/kotlin/BotRAGConfigurationMongoDAOTest.kt
@@ -118,11 +118,11 @@ internal class BotRAGConfigurationMongoDAOTest : AbstractTest() {
         BotRAGConfigurationMongoDAO.save(config1)
         BotRAGConfigurationMongoDAO.save(config2)
 
-        BotRAGConfigurationMongoDAO.save(config1.copy(documentsRequired = true))
+        BotRAGConfigurationMongoDAO.save(config1.copy(debugEnabled = true))
 
         val configBDD = BotRAGConfigurationMongoDAO.findByNamespaceAndBotId("namespace1", "botId1")
 
-        assertEquals(config1.copy(documentsRequired = true), configBDD)
+        assertEquals(config1.copy(debugEnabled = true), configBDD)
     }
 
     @Test
diff --git a/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/requests/RAGRequest.kt b/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/requests/RAGRequest.kt
index 4ae3eb2749..43a47de2dc 100644
--- a/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/requests/RAGRequest.kt
+++ b/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/requests/RAGRequest.kt
@@ -34,5 +34,4 @@ data class RAGRequest(
     val compressorSetting: DocumentCompressorSetting?,
     val vectorStoreSetting: VectorStoreSetting?,
     val observabilitySetting: ObservabilitySetting?,
-    val documentsRequired: Boolean = true,
 )
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/requests/requests.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/requests/requests.py
index 238d74da2a..5a25f40ed1 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/requests/requests.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/requests/requests.py
@@ -66,7 +66,9 @@ class ObservabilityProviderSettingStatusRequest(BaseModel):
 class DocumentCompressorProviderSettingStatusRequest(BaseModel):
     """The request for the Document Compressor Provider Setting Status"""
 
-    setting: DocumentCompressorSetting = Field(description='The Document Compressor Provider setting to be checked.')
+    setting: DocumentCompressorSetting = Field(
+        description='The Document Compressor Provider setting to be checked.'
+    )
 
 
 class BaseRequest(BaseModel):
@@ -144,21 +146,28 @@ class VectorStoreProviderSettingStatusRequest(BaseModel):
         default=None,
     )
 
+
 class DialogDetails(BaseModel):
     """The dialog details model"""
 
     dialog_id: Optional[str] = Field(
         description='The dialog/session ID, attached to the observability traces if '
-                    'the observability provider support it.',
-        default=None, examples=['uuid-0123'])
+        'the observability provider support it.',
+        default=None,
+        examples=['uuid-0123'],
+    )
     user_id: Optional[str] = Field(
         description='The user ID, attached to the observability traces if the observability provider support it',
-        default=None, examples=['address@mail.com'])
+        default=None,
+        examples=['address@mail.com'],
+    )
     history: list[ChatMessage] = Field(
-        description="Conversation history, used to reformulate the user's question.")
+        description="Conversation history, used to reformulate the user's question."
+    )
     tags: list[str] = Field(
         description='List of tags, attached to the observability trace, if the observability provider support it.',
-        examples=[['my-Tag']])
+        examples=[['my-Tag']],
+    )
 
 
 class RAGRequest(BaseRequest):
@@ -166,15 +175,16 @@ class RAGRequest(BaseRequest):
 
     dialog: Optional[DialogDetails] = Field(description='The user dialog details.')
     question_condensing_llm_setting: Optional[LLMSetting] = Field(
-        description="LLM setting, used to condense the user's question.", default=None)
+        description="LLM setting, used to condense the user's question.", default=None
+    )
     question_condensing_prompt: Optional[PromptTemplate] = Field(
         description='Prompt template, used to create a prompt with inputs for jinja and fstring format',
-        default = None
+        default=None,
     )
     question_answering_llm_setting: LLMSetting = Field(
         description='LLM setting, used to perform a QA Prompt.'
     )
-    question_answering_prompt : PromptTemplate = Field(
+    question_answering_prompt: PromptTemplate = Field(
         description='Prompt template, used to create a prompt with inputs for jinja and fstring format'
     )
     guardrail_setting: Optional[GuardrailSetting] = Field(
@@ -182,8 +192,8 @@ class RAGRequest(BaseRequest):
     )
     documents_required: Optional[bool] = Field(
         description='Specifies whether the presence of documents is mandatory for generating answers. '
-                    'If set to True, the system will only provide answers when relevant documents are found. '
-                    'If set to False, the system can respond without requiring document sources. Default is True.',
+        'If set to True, the system will only provide answers when relevant documents are found. '
+        'If set to False, the system can respond without requiring document sources. Default is True.',
         default=True,
     )
 
@@ -191,7 +201,7 @@ class RAGRequest(BaseRequest):
         'json_schema_extra': {
             'examples': [
                 {
-                    'dialog' : {
+                    'dialog': {
                         'history': [
                             {'text': 'Hello, how can I do this?', 'type': 'HUMAN'},
                             {
@@ -212,7 +222,7 @@ class RAGRequest(BaseRequest):
                     'question_answering_prompt': {
                         'formatter': 'f-string',
                         'template': """Use the following context to answer the question at the end.
-If you don't know the answer, just say {no_answer}.
+If you don't know the answer, just say I don't know.
 
 Context:
 {context}
@@ -223,9 +233,8 @@ class RAGRequest(BaseRequest):
 Answer in {locale}:""",
                         'inputs': {
                             'question': 'How to get started playing guitar ?',
-                            'no_answer': 'Sorry, I don t know.',
                             'locale': 'French',
-                        }
+                        },
                     },
                     'embedding_question_em_setting': {
                         'provider': 'OpenAI',
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py
index df5ac562ce..e585db67f4 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py
@@ -21,7 +21,6 @@
 import logging
 import time
 from functools import partial
-from logging import ERROR, WARNING
 from operator import itemgetter
 from typing import List, Optional
 
@@ -32,11 +31,14 @@
 from langchain_core.callbacks import BaseCallbackHandler
 from langchain_core.documents import Document
 from langchain_core.messages import AIMessage, HumanMessage
-from langchain_core.output_parsers import JsonOutputParser, StrOutputParser
+from langchain_core.output_parsers import (
+    JsonOutputParser,
+    PydanticOutputParser,
+    StrOutputParser,
+)
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 from langchain_core.prompts import PromptTemplate as LangChainPromptTemplate
 from langchain_core.runnables import (
-    RunnableConfig,
     RunnableLambda,
     RunnableParallel,
     RunnablePassthrough,
@@ -44,7 +46,6 @@
 )
 from langchain_core.runnables.config import RunnableConfig
 from langchain_core.vectorstores import VectorStoreRetriever
-from langfuse import get_client, propagate_attributes
 from typing_extensions import Any
 
 from gen_ai_orchestrator.errors.exceptions.exceptions import (
@@ -122,18 +123,21 @@ async def execute_rag_chain(
     )
 
     message_history = ChatMessageHistory()
-    session_id = None
-    user_id = None
-    tags = []
+    metadata = {}
+
     if request.dialog:
         for msg in request.dialog.history:
             if ChatMessageType.HUMAN == msg.type:
                 message_history.add_user_message(msg.text)
             else:
                 message_history.add_ai_message(msg.text)
-        session_id = request.dialog.dialog_id
-        user_id = request.dialog.user_id
-        tags = request.dialog.tags or []
+
+        if request.dialog.user_id is not None:
+            metadata['langfuse_user_id'] = request.dialog.user_id
+        if request.dialog.dialog_id is not None:
+            metadata['langfuse_session_id'] = request.dialog.dialog_id
+        if request.dialog.tags:
+            metadata['langfuse_tags'] = request.dialog.tags
 
     logger.debug(
         'RAG chain - Use chat history: %s',
@@ -165,14 +169,6 @@ async def execute_rag_chain(
         )
         callback_handlers.append(observability_handler)
 
-    metadata = {}
-    if user_id is not None:
-        metadata['langfuse_user_id'] = user_id
-    if session_id is not None:
-        metadata['langfuse_session_id'] = session_id
-    if tags:
-        metadata['langfuse_tags'] = tags
-
     response = await conversational_retrieval_chain.ainvoke(
         input=inputs,
         config=RunnableConfig(
@@ -180,10 +176,8 @@ async def execute_rag_chain(
             metadata=metadata,
         ),
     )
-    llm_answer = LLMAnswer(**response['answer'])
 
-    # RAG Guard
-    rag_guard(inputs, llm_answer, response, request.documents_required)
+    llm_answer: LLMAnswer = response['answer']
 
     # Guardrail
     if request.guardrail_setting:
@@ -204,20 +198,22 @@ async def execute_rag_chain(
         if ctx.used_in_response
     }
 
+    footnotes = {
+        Footnote(
+            identifier=doc.metadata['id'],
+            title=doc.metadata['title'],
+            url=doc.metadata['source'],
+            content=get_source_content(doc),
+            score=doc.metadata.get('retriever_score', None),
+        )
+        for doc in response['documents']
+        if doc.metadata['id'] in contexts_by_chunk
+    }
+
     # Returning RAG response
     return RAGResponse(
         answer=llm_answer,
-        footnotes={
-            Footnote(
-                identifier=doc.metadata['id'],
-                title=doc.metadata['title'],
-                url=doc.metadata['source'],
-                content=get_source_content(doc),
-                score=doc.metadata.get('retriever_score', None),
-            )
-            for doc in response['documents']
-            if doc.metadata['id'] in contexts_by_chunk
-        },
+        footnotes=footnotes,
         observability_info=get_observability_info(
             observability_handler,
             ObservabilityTrace.RAG.value if observability_handler is not None else None,
@@ -344,6 +340,8 @@ def retrieve_with_variants(inputs):
         }
     )
 
+    parser = PydanticOutputParser(pydantic_object=LLMAnswer, name='rag_chain_output')
+
     return rag_inputs | RunnablePassthrough.assign(
         answer=(
             {
@@ -362,7 +360,7 @@ def retrieve_with_variants(inputs):
             }
             | rag_prompt
             | question_answering_llm
-            | JsonOutputParser(pydantic_object=LLMAnswer, name='rag_chain_output')
+            | parser
         )
     )
 
@@ -439,46 +437,6 @@ def contextualize_question(inputs: dict, chat_chain) -> str:
     return inputs['question']
 
 
-def rag_guard(question, answer, response, documents_required):
-    """
-    Validates the RAG system's response based on the presence or absence of source documents
-    and the `documentsRequired` setting.
-
-    Args:
-        question: user question
-        answer: the LLM answer
-        response: the RAG response
-        documents_required (bool): Specifies whether documents are mandatory for the response.
-    """
-
-    if (
-        documents_required
-        and answer.status == 'found_in_context'
-        and len(response['documents']) == 0
-    ):
-        message = 'No documents were retrieved, yet an answer was attempted.'
-        rag_log(
-            level=ERROR,
-            message=message,
-            question=question,
-            answer=answer.answer,
-            response=response,
-        )
-        raise GenAIGuardCheckException(ErrorInfo(cause=message))
-
-    if answer.status == 'not_found_in_context' and len(response['documents']) > 0:
-        # If the answer is not found in context and some documents are retrieved, so we remove them from the RAG response.
-        message = 'No answer found in the retrieved context. The documents are therefore removed from the RAG response.'
-        rag_log(
-            level=WARNING,
-            message=message,
-            question=question,
-            answer=answer.answer,
-            response=response,
-        )
-        response['documents'] = []
-
-
 def rag_log(level, message, question, answer, response):
     """
     RAG logging
diff --git a/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py b/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py
index 000e483226..ef16f03eb9 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py
@@ -55,7 +55,6 @@
 )
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.create_rag_chain')
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.RAGCallbackHandler')
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_guard')
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.RAGResponse')
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.RAGDebugData')
 @patch('gen_ai_orchestrator.services.langchain.rag_chain.get_llm_answer')
@@ -64,7 +63,6 @@ async def test_rag_chain(
     mocked_get_llm_answer,
     mocked_rag_debug_data,
     mocked_rag_response,
-    mocked_rag_guard,
     mocked_callback_init,
     mocked_create_rag_chain,
     mocked_get_callback_handler_factory,
@@ -96,7 +94,7 @@ async def test_rag_chain(
         'question_answering_prompt': {
             'formatter': 'f-string',
             'template': """Use the following context to answer the question at the end.
-If you don't know the answer, just say {no_answer}.
+If you don't know the answer, just say I don't know.
 
 Context:
 {context}
@@ -107,7 +105,6 @@ async def test_rag_chain(
 Answer in {locale}:""",
             'inputs': {
                 'question': 'How to get started playing guitar ?',
-                'no_answer': 'Sorry, I don t know.',
                 'locale': 'French',
             },
         },
@@ -177,16 +174,18 @@ async def test_rag_chain(
         )
     ]
     response = {
-        'answer': {
-            'status': '',
-            'answer': 'an answer from llm',
-            'topic': None,
-            'suggested_topics': None,
-            'context': [],
-        },
+        'answer': LLMAnswer(
+            **{
+                'status': '',
+                'answer': 'an answer from llm',
+                'topic': None,
+                'suggested_topics': None,
+                'context': [],
+            }
+        ),
         'documents': docs,
     }
-    llm_answer = LLMAnswer(**response['answer'])
+    llm_answer = response['answer']
 
     # Setup mock factories/init return value
     observability_factory_instance = mocked_get_callback_handler_factory.return_value
@@ -228,11 +227,9 @@ async def test_rag_chain(
     # Assert the rag guardrail is called
     mocked_guardrail_parse.assert_called_once_with(
         os.path.join(request.guardrail_setting.api_base, 'guardrail'),
-        json={'text': [mocked_rag_answer['answer']['answer']]},
-    )
-    # Assert the rag guard is called
-    mocked_rag_guard.assert_called_once_with(
-        inputs, llm_answer, response, request.documents_required
+        json={
+            'text': [mocked_rag_answer['answer'].answer]
+        },  # TODO MASS : answer.answer pas top, à renomer
     )
 
 
@@ -459,91 +456,3 @@ def test_check_guardrail_output_is_ok():
     }
 
     assert check_guardrail_output(guardrail_output) is True
-
-
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
-def test_rag_guard_fails_if_no_docs_in_valid_answer(mocked_log):
-    question = 'Hi!'
-    response = {
-        'answer': {'status': 'found_in_context', 'answer': 'a valid answer'},
-        'documents': [],
-    }
-    try:
-        rag_chain.rag_guard(
-            question, LLMAnswer(**response['answer']), response, documents_required=True
-        )
-    except Exception as e:
-        assert isinstance(e, GenAIGuardCheckException)
-
-
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
-def test_rag_guard_accepts_no_answer_even_with_docs(mocked_log):
-    question = 'Hi!'
-    response = {
-        'answer': {
-            'status': 'not_found_in_context',
-            'answer': 'Sorry, I don t know.',
-            'context': [
-                {
-                    'chunk': 1,
-                    'sentences': ['str1'],
-                }
-            ],
-        },
-        'documents': ['a doc as a string'],
-    }
-    rag_chain.rag_guard(
-        question, LLMAnswer(**response['answer']), response, documents_required=True
-    )
-    # No answer found in the retrieved context. The documents are therefore removed from the RAG response.
-    assert response['documents'] == []
-
-
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
-def test_rag_guard_valid_answer_with_docs(mocked_log):
-    question = 'Hi!'
-    response = {
-        'answer': {
-            'status': 'found_in_context',
-            'answer': 'a valid answer',
-        },
-        'documents': ['doc1', 'doc2'],
-    }
-    rag_chain.rag_guard(
-        question, LLMAnswer(**response['answer']), response, documents_required=True
-    )
-    assert response['documents'] == ['doc1', 'doc2']
-
-
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
-def test_rag_guard_no_answer_with_no_docs(mocked_log):
-    question = 'Hi!'
-    response = {
-        'answer': {'status': 'not_found_in_context', 'answer': 'Sorry, I don t know.'},
-        'documents': [],
-    }
-    rag_chain.rag_guard(
-        question, LLMAnswer(**response['answer']), response, documents_required=True
-    )
-    assert response['documents'] == []
-
-
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.rag_log')
-def test_rag_guard_without_no_answer_input(mocked_log):
-    """Test that __rag_guard handles missing no_answer input correctly."""
-    question = 'Hi!'
-    response = {
-        'answer': {
-            'status': 'found_in_context',
-            'answer': 'a valid answer',
-        },
-        'documents': [],
-    }
-    with pytest.raises(GenAIGuardCheckException) as exc:
-        rag_chain.rag_guard(
-            question, LLMAnswer(**response['answer']), response, documents_required=True
-        )
-
-    mocked_log.assert_called_once()
-
-    assert isinstance(exc.value, GenAIGuardCheckException)

From df02044f14208f9ac6ccea6fb44c2f7b55c0a31a Mon Sep 17 00:00:00 2001
From: Mohamed ASSOUKTI <mohamed.assoukti@partnre.com>
Date: Tue, 19 Aug 2025 10:56:44 +0200
Subject: [PATCH 3/7] [DERCBOT-1609] Structuring the LLM response

---
 .../admin/bot/rag/BotRAGConfiguration.kt      |   4 +
 .../kotlin/engine/config/RAGAnswerHandler.kt  |  84 ++--
 .../orchestratorclient/responses/Models.kt    |  16 +-
 .../models/rag/rag_models.py                  | 130 ++---
 .../routers/responses/responses.py            |  87 ++--
 .../services/langchain/rag_chain.py           | 297 ++++++------
 .../server/tests/services/test_rag_chain.py   | 459 ++++++++++--------
 7 files changed, 534 insertions(+), 543 deletions(-)

diff --git a/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt b/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt
index 18b6552f48..e769149c7d 100644
--- a/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt
+++ b/bot/engine/src/main/kotlin/admin/bot/rag/BotRAGConfiguration.kt
@@ -35,6 +35,10 @@ data class BotRAGConfiguration(
     val llmSetting: LLMSetting? = null,
     val emSetting: EMSetting,
     val indexSessionId: String? = null,
+    @Deprecated("Replaced by LLM answer status")
+    val noAnswerSentence: String,
+    val noAnswerStoryId: String? = null,
+    val documentsRequired: Boolean = true,
     val debugEnabled: Boolean = false,
     val maxDocumentsRetrieved: Int = 4,
     val maxMessagesFromHistory: Int = 5,
diff --git a/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt b/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt
index d37fa6b0f9..e2a3dee712 100644
--- a/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt
+++ b/bot/engine/src/main/kotlin/engine/config/RAGAnswerHandler.kt
@@ -28,10 +28,7 @@ import ai.tock.bot.engine.action.SendSentence
 import ai.tock.bot.engine.action.SendSentenceWithFootnotes
 import ai.tock.bot.engine.dialog.Dialog
 import ai.tock.bot.engine.user.PlayerType
-import ai.tock.genai.orchestratorclient.requests.ChatMessage
-import ai.tock.genai.orchestratorclient.requests.ChatMessageType
-import ai.tock.genai.orchestratorclient.requests.DialogDetails
-import ai.tock.genai.orchestratorclient.requests.RAGRequest
+import ai.tock.genai.orchestratorclient.requests.*
 import ai.tock.genai.orchestratorclient.responses.LLMAnswer
 import ai.tock.genai.orchestratorclient.responses.ObservabilityInfo
 import ai.tock.genai.orchestratorclient.responses.RAGResponse
@@ -62,7 +59,7 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
             BotRepository.saveMetric(createMetric(MetricType.STORY_HANDLED))
 
             // Call RAG Api - Gen AI Orchestrator
-            val (answer, footnotes, debug, redirectStory, observabilityInfo) = rag(this)
+            val (answer, footnotes, debug, noAnswerStory, observabilityInfo) = rag(this)
 
             // Add debug data if available and if debugging is enabled
             if (debug != null) {
@@ -72,15 +69,23 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
 
             val modifiedObservabilityInfo = observabilityInfo?.let { updateObservabilityInfo(this, it) }
 
-            // Footnotes building
-            val preparedFootnotes =
-                footnotes?.map {
-                    Footnote(
-                        it.identifier,
-                        it.title,
-                        it.url,
-                        if (action.metadata.sourceWithContent) it.content else null,
-                        it.score,
+                val modifiedObservabilityInfo = observabilityInfo?.let { updateObservabilityInfo(this, it) }
+
+                send(
+                    action = SendSentenceWithFootnotes(
+                        playerId = botId,
+                        applicationId = connectorId,
+                        recipientId = userId,
+                        text = answer.answer,
+                        footnotes = footnotes?.map {
+                            Footnote(
+                                it.identifier, it.title, it.url,
+                                if(action.metadata.sourceWithContent) it.content else null,
+                                it.score
+                            )
+                        }?.toMutableList() ?: mutableListOf<Footnote>(),
+                        // modifiedObservabilityInfo includes the public langfuse URL if filled.
+                        metadata = ActionMetadata(isGenAiRagAnswer = true, observabilityInfo = modifiedObservabilityInfo)
                     )
                 }?.toMutableList() ?: mutableListOf()
 
@@ -130,12 +135,24 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
      * @param botBus the bot Bus
      * @param response the RAG response
      */
-    private fun ragStoryRedirection(
-        botBus: BotBus,
-        response: RAGResponse?,
-    ): StoryDefinition? {
-        return response?.answer?.redirectionIntent?.let {
-            botBus.botDefinition.findStoryDefinition(it, "")
+    private fun ragStoryRedirection(botBus: BotBus, response: RAGResponse?): StoryDefinition? {
+        return with(botBus) {
+            botDefinition.ragConfiguration?.let { ragConfig ->
+                if (response?.answer?.status.equals("not_found_in_context", ignoreCase = true)) {
+                    // Save no answer metric
+                    saveRagMetric(IndicatorValues.NO_ANSWER)
+
+                    // Switch to no answer story if configured
+                    if (!ragConfig.noAnswerStoryId.isNullOrBlank()) {
+                        logger.info { "Switch to the no-answer RAG story." }
+                        getNoAnswerRAGStory(ragConfig)
+                    } else null
+                } else {
+                    // Save success metric
+                    saveRagMetric(IndicatorValues.SUCCESS)
+                    null
+                }
+            }
         }
     }
 
@@ -214,26 +231,23 @@ object RAGAnswerHandler : AbstractProactiveAnswerHandler {
                 }
 
                 // Handle RAG response
-                return RAGResult(
-                    response?.answer,
-                    response?.footnotes,
-                    response?.debug,
-                    ragStoryRedirection(this, response),
-                    response?.observabilityInfo,
-                )
+                return RAGResult(response?.answer, response?.footnotes, response?.debug, ragStoryRedirection(this, response), response?.observabilityInfo)
             } catch (exc: Exception) {
                 logger.error { exc }
                 // Save failure metric
                 saveRagMetric(IndicatorValues.FAILURE)
 
-                return RAGResult(
-                    answer = LLMAnswer(status = "error", answer = technicalErrorMessage),
-                    debug =
-                        when (exc) {
-                            is GenAIOrchestratorBusinessError -> RAGError(exc.message, exc.error)
-                            is GenAIOrchestratorValidationError -> RAGError(exc.message, exc.detail)
-                            else -> RAGError(errorMessage = exc.message)
-                        },
+                return if (exc is GenAIOrchestratorBusinessError && exc.error.info.error == "APITimeoutError") {
+                    logger.info { "The APITimeoutError is raised, so switch to the no-answer story." }
+                    RAGResult(noAnswerStory = getNoAnswerRAGStory(ragConfiguration))
+                }
+                else RAGResult(
+                    answer = LLMAnswer(status="error", answer = technicalErrorMessage),
+                    debug = when(exc) {
+                        is GenAIOrchestratorBusinessError -> RAGError(exc.message, exc.error)
+                        is GenAIOrchestratorValidationError -> RAGError(exc.message, exc.detail)
+                        else -> RAGError(errorMessage = exc.message)
+                    }
                 )
             }
         }
diff --git a/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/Models.kt b/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/Models.kt
index 95f765d00f..4cd5ceaaf8 100644
--- a/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/Models.kt
+++ b/gen-ai/orchestrator-client/src/main/kotlin/ai/tock/genai/orchestratorclient/responses/Models.kt
@@ -23,16 +23,18 @@ data class ChunkSentences(
     val reason: String? = null,
 )
 
+data class ChunkSentences(
+    val chunk: String? = null,
+    val sentences: List<String>? = emptyList(),
+    val reason: String? = null,
+)
+
 data class LLMAnswer(
-    val status: String?,
-    val answer: String?,
-    val confidenceScore: Double? = 0.0,
-    val displayAnswer: Boolean? = true,
+    val status: String,
+    val answer: String,
     val topic: String? = null,
     val suggestedTopics: List<String>? = null,
-    val understanding: String? = null,
-    val redirectionIntent: String? = null,
-    val contextUsage: List<ChunkSentences>? = null,
+    val context: List<ChunkSentences>? = null,
 )
 
 data class Footnote(
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/models/rag/rag_models.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/models/rag/rag_models.py
index 5682c99ee9..1e2ab63a1b 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/models/rag/rag_models.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/models/rag/rag_models.py
@@ -27,16 +27,10 @@
 class Source(BaseModel):
     """A source model, used to associate document sources with the QA response"""
 
-    title: str = Field(description='Source title', examples=['Tock Documentation'])
-    url: Optional[AnyUrl] = Field(
-        description='Source url', examples=['https://doc.tock.ai/tock/'], default=None
-    )
-    content: str = Field(
-        description='Source content', examples=['Tock: The Open Conversation Kit']
-    )
-    score: Optional[float] = Field(
-        description='The compressor score', examples=[0.9149009585380554], default=None
-    )
+    title: str = Field(description="Source title", examples=["Tock Documentation"])
+    url: Optional[AnyUrl] = Field(description="Source url", examples=["https://doc.tock.ai/tock/"], default=None)
+    content: str = Field(description="Source content", examples=["Tock: The Open Conversation Kit"])
+    score: Optional[float] = Field(description="The compressor score", examples=[0.9149009585380554], default=None)
 
     def __eq__(self, other):
         """
@@ -52,35 +46,26 @@ def __eq__(self, other):
         )
 
     def __hash__(self):
-        return hash((self.title, str(self.url or ''), self.content))
+        return hash((self.title, str(self.url or ""), self.content))
 
 
 class Footnote(Source):
     """A footnote model, used to associate document sources with the RAG answer"""
 
-    identifier: str = Field(description='Footnote identifier', examples=['1'])
+    identifier: str = Field(description="Footnote identifier", examples=["1"])
 
 
 class ChunkInfos(BaseModel):
     """A model representing information about a chunk used in the RAG context."""
 
     chunk: Optional[str] = Field(
-        description='Unique identifier of the chunk.',
-        examples=['cd6d8221-ba9f-44da-86ee-0e25a3c9a5c7'],
-        default=None,
+        description="Unique identifier of the chunk.", examples=["cd6d8221-ba9f-44da-86ee-0e25a3c9a5c7"], default=None
     )
     sentences: Optional[List[str]] = Field(
-        description='List of verbatim sentences from the chunk that were used by the LLM.',
-        default=None,
-    )
-    used_in_response: Optional[bool] = Field(
-        description='It indicates whether the chunk is actually used to generate the response.',
-        examples=[True],
-        default=False,
+        description="List of verbatim sentences from the chunk that were used by the LLM.", default=None
     )
     reason: Optional[str] = Field(
-        description='Reason why the chunk was not used (e.g., irrelevant, general background).',
-        default=None,
+        description="Reason why the chunk was not used (e.g., irrelevant, general background).", default=None
     )
 
 
@@ -91,44 +76,25 @@ class LLMAnswer(BaseModel):
     """
 
     status: Optional[str] = Field(
-        description='The status of the answer generation. '
+        description="The status of the answer generation. "
         "Possible values: 'found_in_context', 'not_found_in_context', 'small_talk', "
-        'or other case-specific codes.',
+        "or other case-specific codes.",
         default=None,
     )
     answer: Optional[str] = Field(
-        description="The textual answer generated by the LLM, in the user's locale.",
-        default=None,
-    )
-    confidence_score: Optional[float] = Field(
-        description="Confidence score assigned to the answer provided to the user's question",
-        examples=[0.93],
-        default=None,
-    )
-    display_answer: Optional[bool] = Field(
-        description='It indicates if we should display answer or not.',
-        examples=[True],
-        default=False,
+        description="The textual answer generated by the LLM, in the user's locale.", default=None
     )
     topic: Optional[str] = Field(
-        description='The main topic assigned to the answer. Must be one of the predefined list '
+        description="The main topic assigned to the answer. Must be one of the predefined list "
         "of topics, or 'unknown' if no match is possible.",
         default=None,
     )
     suggested_topics: Optional[List[str]] = Field(
-        description='A list of suggested alternative or related topics, '
-        "used when the main topic is 'unknown'.",
+        description="A list of suggested alternative or related topics, " "used when the main topic is 'unknown'.",
         default=None,
     )
-    understanding: Optional[str] = Field(
-        description='LLM understanding of user query.',
-        default=None,
-    )
-    redirection_intent: Optional[str] = Field(
-        description='The intent to use to redirect the conversation flow.', default=None
-    )
-    context_usage: Optional[List[ChunkInfos]] = Field(
-        description='The list of chunks from the context that contributed to or were considered '
+    context: Optional[List[ChunkInfos]] = Field(
+        description="The list of chunks from the context that contributed to or were considered "
         "in the LLM's answer. Each entry contains identifiers, sentences, and reasons.",
         default=None,
     )
@@ -138,35 +104,31 @@ class LLMAnswer(BaseModel):
 class ChatMessageType(str, Enum):
     """Enumeration to list a chat message type"""
 
-    HUMAN = 'HUMAN'
-    AI = 'AI'
+    HUMAN = "HUMAN"
+    AI = "AI"
 
 
 class ChatMessage(BaseModel):
     """A conversation chat message"""
 
-    text: str = Field(
-        description='Conversation message text', examples=['Hello, how can I do this?']
-    )
-    type: ChatMessageType = Field(description='The message origin (Human or AI)')
+    text: str = Field(description="Conversation message text", examples=["Hello, how can I do this?"])
+    type: ChatMessageType = Field(description="The message origin (Human or AI)")
 
 
 class RAGDocumentMetadata(BaseModel):
     """The RAG document metadata"""
 
-    index_session_id: str = Field(
-        description='The indexing session id.', examples=['123f-ed01-gt21-gg08']
-    )
-    id: str = Field(description='The document id.', examples=['e014-g24-0f11-1g3e'])
-    title: str = Field(description='The document title.', examples=['Tracking shot'])
+    index_session_id: str = Field(description="The indexing session id.", examples=["123f-ed01-gt21-gg08"])
+    id: str = Field(description="The document id.", examples=["e014-g24-0f11-1g3e"])
+    title: str = Field(description="The document title.", examples=["Tracking shot"])
     url: Optional[HttpUrl] = Field(
-        description='The document url.',
-        examples=['https://en.wikipedia.org/wiki/Tracking_shot'],
+        description="The document url.",
+        examples=["https://en.wikipedia.org/wiki/Tracking_shot"],
         default=None,
     )
-    chunk: str = Field(description='The document chunk.', examples=['1/3'])
+    chunk: str = Field(description="The document chunk.", examples=["1/3"])
     retriever_score: Optional[float] = Field(
-        description='The compressor score', examples=[0.9149009585380554], default=None
+        description="The compressor score", examples=[0.9149009585380554], default=None
     )
 
 
@@ -174,14 +136,14 @@ class RAGDocument(BaseModel):
     """The definition of RAG document"""
 
     content: str = Field(
-        description='The document content.',
+        description="The document content.",
         examples=[
-            'In cinematography, a tracking shot is any shot where the camera follows backward, '
-            'forward or moves alongside the subject being recorded.'
+            "In cinematography, a tracking shot is any shot where the camera follows backward, "
+            "forward or moves alongside the subject being recorded."
         ],
     )
     metadata: RAGDocumentMetadata = Field(
-        description='The document metadata.',
+        description="The document metadata.",
     )
 
 
@@ -192,40 +154,32 @@ class QADebugData(BaseModel):
         description="The user's initial question.",
         examples=["I'm interested in going to Morocco"],
     )
-    documents: List[RAGDocument] = Field(
-        description='Documents retrieved from the vector store.'
-    )
+    documents: List[RAGDocument] = Field(description="Documents retrieved from the vector store.")
     document_index_name: str = Field(
-        description='Index name corresponding to a document collection in the vector database.',
+        description="Index name corresponding to a document collection in the vector database.",
     )
     document_search_params: DocumentSearchParams = Field(
-        description='The document search parameters. Ex: number of documents, metadata filter',
-    )
-    duration: float = Field(
-        description='The duration of RAG in seconds.', examples=['7.2']
+        description="The document search parameters. Ex: number of documents, metadata filter",
     )
+    duration: float = Field(description="The duration of RAG in seconds.", examples=["7.2"])
 
 
 class RAGDebugData(QADebugData):
     """A RAG debug data"""
 
     question_condensing_prompt: Optional[str] = Field(
-        description='The prompt of the question rephrased with the history of the conversation.',
-        examples=[
-            'Given the following conversation, rephrase the follow up question to be a standalone question.'
-        ],
+        description="The prompt of the question rephrased with the history of the conversation.",
+        examples=["Given the following conversation, rephrase the follow up question to be a standalone question."],
     )
     question_condensing_history: list[ChatMessage] = Field(
         description="Conversation history, used to reformulate the user's question."
     )
     condensed_question: Optional[str] = Field(
-        description='The question rephrased with the history of the conversation.',
-        examples=['Hello, how to plan a trip to Morocco ?'],
+        description="The question rephrased with the history of the conversation.",
+        examples=["Hello, how to plan a trip to Morocco ?"],
     )
     question_answering_prompt: Optional[str] = Field(
-        description='The question answering prompt.',
-        examples=[
-            'Question: Hello, how to plan a trip to Morocco ?. Answer in French.'
-        ],
+        description="The question answering prompt.",
+        examples=["Question: Hello, how to plan a trip to Morocco ?. Answer in French."],
     )
-    answer: LLMAnswer = Field(description='The RAG answer.')
+    answer: LLMAnswer = Field(description="The RAG answer.")
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/responses/responses.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/responses/responses.py
index 1b7b2ab60b..e638100a8a 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/responses/responses.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/routers/responses/responses.py
@@ -27,62 +27,52 @@
     ErrorInfo,
 )
 from gen_ai_orchestrator.models.llm.llm_provider import LLMProvider
-from gen_ai_orchestrator.models.observability.observability_provider import (
-    ObservabilityProvider,
-)
 from gen_ai_orchestrator.models.rag.rag_models import Source, LLMAnswer, Footnote
-from gen_ai_orchestrator.models.vector_stores.vectore_store_provider import (
-    VectorStoreProvider,
-)
+from gen_ai_orchestrator.models.observability.observability_provider import ObservabilityProvider
+from gen_ai_orchestrator.models.vector_stores.vectore_store_provider import VectorStoreProvider
 
 
 class ErrorResponse(BaseModel):
     """The error response model"""
 
     code: ErrorCode = Field(
-        description='The AI orchestrator error code.',
+        description="The AI orchestrator error code.",
         examples=[ErrorCode.GEN_AI_AUTHENTICATION_ERROR],
     )
     message: str = Field(
-        description='The AI orchestrator error message.',
-        examples=['Authentication error to the AI Provider API.'],
+        description="The AI orchestrator error message.",
+        examples=["Authentication error to the AI Provider API."],
     )
     detail: Optional[str] = Field(
-        description='The AI orchestrator error detail. It provides help or a solution.',
-        examples=[
-            'Check your API key or token and make sure it is correct and active.'
-        ],
+        description="The AI orchestrator error detail. It provides help or a solution.",
+        examples=["Check your API key or token and make sure it is correct and active."],
         default=None,
     )
-    info: ErrorInfo = Field(
-        description='The AI orchestrator error info. It exposes the raised error cause.'
-    )
+    info: ErrorInfo = Field(description="The AI orchestrator error info. It exposes the raised error cause.")
 
 
 class ProviderSettingStatusResponse(BaseModel):
     """The response model of the provider setting status"""
 
     valid: bool = Field(
-        description='It indicates the setting validity.',
+        description="It indicates the setting validity.",
         examples=[True],
         default=False,
     )
-    errors: list[ErrorResponse] = Field(description='The list of errors.', default=[])
+    errors: list[ErrorResponse] = Field(description="The list of errors.", default=[])
 
 
 class LLMProviderResponse(BaseModel):
     """The response model of the LLM provider"""
 
-    provider: LLMProvider = Field(
-        description='The LLM Provider ID', default=[LLMProvider.OPEN_AI]
-    )
+    provider: LLMProvider = Field(description="The LLM Provider ID", default=[LLMProvider.OPEN_AI])
 
 
 class VectorStoreProviderResponse(BaseModel):
     """The response model of the Vector Store provider"""
 
     provider: VectorStoreProvider = Field(
-        description='The Vector Store Provider ID', default=[VectorStoreProvider.OPEN_SEARCH]
+        description="The Vector Store Provider ID", default=[VectorStoreProvider.OPEN_SEARCH]
     )
 
 
@@ -90,7 +80,7 @@ class ObservabilityProviderResponse(BaseModel):
     """The response model of the Observability provider"""
 
     provider: ObservabilityProvider = Field(
-        description='The Observability Provider ID', default=[ObservabilityProvider.LANGFUSE]
+        description="The Observability Provider ID", default=[ObservabilityProvider.LANGFUSE]
     )
 
 
@@ -98,7 +88,7 @@ class DocumentCompressorProviderResponse(BaseModel):
     """The response model of the Document Compressor provider"""
 
     provider: DocumentCompressorProvider = Field(
-        description='The Document Compressor Provider ID', default=[DocumentCompressorProvider.BLOOMZ]
+        description="The Document Compressor Provider ID", default=[DocumentCompressorProvider.BLOOMZ]
     )
 
 
@@ -106,7 +96,7 @@ class EMProviderResponse(BaseModel):
     """The response model of the EM provider"""
 
     provider: EMProvider = Field(
-        description='The Embedding Model Provider ID',
+        description="The Embedding Model Provider ID",
         default=[EMProvider.AZURE_OPEN_AI_SERVICE],
     )
 
@@ -114,55 +104,38 @@ class EMProviderResponse(BaseModel):
 class ObservabilityInfo(BaseModel):
     """The Observability Info model"""
 
-    trace_id: str = Field(
-        description='The observability trace id.'
-    )
-    trace_name: str = Field(
-        description='The observability trace name.'
-    )
-    trace_url: str = Field(
-        description='The observability trace url.'
-    )
+    trace_id: str = Field(description="The observability trace id.")
+    trace_name: str = Field(description="The observability trace name.")
+    trace_url: str = Field(description="The observability trace url.")
+
 
 class RAGResponse(BaseModel):
     """The RAG response model"""
 
-    answer: Optional[LLMAnswer] = Field(
-        description='The RAG answer'
-    )
-    footnotes: set[Footnote] = Field(description='Set of footnotes')
+    answer: Optional[LLMAnswer] = Field(description="The RAG answer")
+    footnotes: set[Footnote] = Field(description="Set of footnotes")
     debug: Optional[Any] = Field(
-        description='Debug data',
-        examples=[{'action': 'retrieve', 'result': 'OK', 'errors': []}],
+        description="Debug data",
+        examples=[{"action": "retrieve", "result": "OK", "errors": []}],
         default=None,
     )
-    observability_info: Optional[ObservabilityInfo] = Field(
-        description='The observability info.',
-        default=None
-    )
+    observability_info: Optional[ObservabilityInfo] = Field(description="The observability info.", default=None)
+
 
 class QAResponse(BaseModel):
     """The QA response model"""
 
-    documents: set[Source] = Field(
-        description='The sources corresponding to the QA request.'
-    )
+    documents: set[Source] = Field(description="The sources corresponding to the QA request.")
 
 
 class SentenceGenerationResponse(BaseModel):
     """The sentence generation response model"""
 
-    sentences: list[str] = Field(
-        description='The list of generated sentences.', default=[]
-    )
+    sentences: list[str] = Field(description="The list of generated sentences.", default=[])
+
 
 class PlaygroundResponse(BaseModel):
     """The playground response model"""
 
-    answer: str = Field(
-        description='The playground answer.'
-    )
-    observability_info: Optional[ObservabilityInfo] = Field(
-        description='The observability info.',
-        default=None
-    )
+    answer: str = Field(description="The playground answer.")
+    observability_info: Optional[ObservabilityInfo] = Field(description="The observability info.", default=None)
diff --git a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py
index e585db67f4..e4531327f1 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/src/gen_ai_orchestrator/services/langchain/rag_chain.py
@@ -21,8 +21,9 @@
 import logging
 import time
 from functools import partial
+from logging import ERROR, WARNING
 from operator import itemgetter
-from typing import List, Optional
+from typing import List, Optional, Tuple
 
 from langchain.retrievers.contextual_compression import (
     ContextualCompressionRetriever,
@@ -30,12 +31,8 @@
 from langchain_community.chat_message_histories import ChatMessageHistory
 from langchain_core.callbacks import BaseCallbackHandler
 from langchain_core.documents import Document
-from langchain_core.messages import AIMessage, HumanMessage
-from langchain_core.output_parsers import (
-    JsonOutputParser,
-    PydanticOutputParser,
-    StrOutputParser,
-)
+from langchain_core.messages import HumanMessage, AIMessage
+from langchain_core.output_parsers import StrOutputParser, JsonOutputParser
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 from langchain_core.prompts import PromptTemplate as LangChainPromptTemplate
 from langchain_core.runnables import (
@@ -43,6 +40,8 @@
     RunnableParallel,
     RunnablePassthrough,
     RunnableSerializable,
+    RunnableConfig,
+    RunnableLambda,
 )
 from langchain_core.runnables.config import RunnableConfig
 from langchain_core.vectorstores import VectorStoreRetriever
@@ -73,9 +72,12 @@
     RAGDebugData,
     RAGDocument,
     RAGDocumentMetadata,
+    LLMAnswer,
 )
 from gen_ai_orchestrator.routers.requests.requests import RAGRequest
-from gen_ai_orchestrator.routers.responses.responses import RAGResponse
+from gen_ai_orchestrator.routers.responses.responses import (
+    RAGResponse,
+)
 from gen_ai_orchestrator.services.langchain.callbacks.rag_callback_handler import (
     RAGCallbackHandler,
 )
@@ -98,7 +100,7 @@
 
 
 @opensearch_exception_handler
-@openai_exception_handler(provider='OpenAI or AzureOpenAIService')
+@openai_exception_handler(provider="OpenAI or AzureOpenAIService")
 async def execute_rag_chain(
     request: RAGRequest,
     debug: bool,
@@ -115,16 +117,12 @@ async def execute_rag_chain(
         The RAG response (Answer and document sources)
     """
 
-    logger.info('RAG chain - Start of execution...')
+    logger.info("RAG chain - Start of execution...")
     start_time = time.time()
 
-    conversational_retrieval_chain = create_rag_chain(
-        request=request, vector_db_async_mode=False  # TODO MASS
-    )
+    conversational_retrieval_chain = create_rag_chain(request=request, vector_db_async_mode=False)  # TODO MASS
 
     message_history = ChatMessageHistory()
-    metadata = {}
-
     if request.dialog:
         for msg in request.dialog.history:
             if ChatMessageType.HUMAN == msg.type:
@@ -132,95 +130,83 @@ async def execute_rag_chain(
             else:
                 message_history.add_ai_message(msg.text)
 
-        if request.dialog.user_id is not None:
-            metadata['langfuse_user_id'] = request.dialog.user_id
-        if request.dialog.dialog_id is not None:
-            metadata['langfuse_session_id'] = request.dialog.dialog_id
-        if request.dialog.tags:
-            metadata['langfuse_tags'] = request.dialog.tags
+    logger.debug("RAG chain - Use chat history: %s", len(message_history.messages) > 0)
+    logger.debug("RAG chain - Use RAGCallbackHandler for debugging : %s", debug)
 
-    logger.debug(
-        'RAG chain - Use chat history: %s',
-        'Yes' if len(message_history.messages) > 0 else 'No',
-    )
+    records_handler, observability_handler = get_callback_handlers(request, debug)
+
+    callbacks = [
+        handler
+        for handler in (records_handler, observability_handler, custom_observability_handler)
+        if handler is not None
+    ]
 
     inputs = {
         **request.question_answering_prompt.inputs,
-        'chat_history': message_history.messages,
+        "chat_history": message_history.messages,
     }
 
-    logger.debug(
-        'RAG chain - Use RAGCallbackHandler for debugging : %s',
-        debug,
-    )
+    response = await conversational_retrieval_chain.ainvoke(input=inputs, config=RunnableConfig(callbacks=callbacks))
+    llm_answer = LLMAnswer(**response["answer"])
 
-    callback_handlers = []
-    records_callback_handler = RAGCallbackHandler()
-    observability_handler = None
-    if debug:
-        # Debug callback handler
-        callback_handlers.append(records_callback_handler)
-    if custom_observability_handler is not None:
-        callback_handlers.append(custom_observability_handler)
-    if request.observability_setting is not None:
-        # Langfuse callback handler
-        observability_handler = create_observability_callback_handler(
-            observability_setting=request.observability_setting,
-        )
-        callback_handlers.append(observability_handler)
-
-    response = await conversational_retrieval_chain.ainvoke(
-        input=inputs,
-        config=RunnableConfig(
-            callbacks=callback_handlers,
-            metadata=metadata,
-        ),
-    )
-
-    llm_answer: LLMAnswer = response['answer']
+    # RAG Guard
+    rag_guard(inputs, llm_answer, response, request.documents_required)
 
     # Guardrail
     if request.guardrail_setting:
-        guardrail = get_guardrail_factory(
-            setting=request.guardrail_setting
-        ).get_parser()
+        guardrail = get_guardrail_factory(setting=request.guardrail_setting).get_parser()
         guardrail_output = guardrail.parse(llm_answer.answer)
         check_guardrail_output(guardrail_output)
 
     # Calculation of RAG processing time
-    rag_duration = '{:.2f}'.format(time.time() - start_time)
-    logger.info('RAG chain - End of execution. (Duration : %s seconds)', rag_duration)
+    rag_duration = "{:.2f}".format(time.time() - start_time)
+    logger.info("RAG chain - End of execution. (Duration : %s seconds)", rag_duration)
 
     # Group contexts by chunk id
-    contexts_by_chunk = {
-        ctx.chunk: ctx
-        for ctx in (llm_answer.context_usage or [])
-        if ctx.used_in_response
-    }
-
-    footnotes = {
-        Footnote(
-            identifier=doc.metadata['id'],
-            title=doc.metadata['title'],
-            url=doc.metadata['source'],
-            content=get_source_content(doc),
-            score=doc.metadata.get('retriever_score', None),
-        )
-        for doc in response['documents']
-        if doc.metadata['id'] in contexts_by_chunk
-    }
+    contexts_by_chunk = {ctx.chunk: ctx for ctx in (llm_answer.context or []) if ctx.sentences}
 
     # Returning RAG response
     return RAGResponse(
         answer=llm_answer,
-        footnotes=footnotes,
-        observability_info=get_observability_info(
-            observability_handler,
-            ObservabilityTrace.RAG.value if observability_handler is not None else None,
-        ),
-        debug=get_rag_debug_data(request, records_callback_handler, rag_duration)
-        if debug
-        else None,
+        footnotes={
+            Footnote(
+                identifier=doc.metadata["id"],
+                title=doc.metadata["title"],
+                url=doc.metadata["source"],
+                content=get_source_content(doc),
+                score=doc.metadata.get("retriever_score", None),
+            )
+            for doc in response["documents"]
+            if doc.metadata["id"] in contexts_by_chunk
+        },
+        observability_info=get_observability_info(observability_handler),
+        debug=get_rag_debug_data(request, records_handler, rag_duration) if debug else None,
+    )
+
+
+def get_callback_handlers(request, debug) -> Tuple[
+    Optional[RAGCallbackHandler],
+    Optional[object],
+]:
+    records_handler = RAGCallbackHandler() if debug else None
+    observability_handler = None
+
+    if request.observability_setting is not None:
+        if request.dialog:
+            session_id = request.dialog.dialog_id
+            user_id = request.dialog.user_id
+            tags = request.dialog.tags
+        else:
+            session_id = None
+            user_id = None
+            tags = None
+        observability_handler = create_observability_callback_handler(
+            observability_setting=request.observability_setting,
+        )
+
+    return (
+        records_handler,
+        observability_handler,
     )
 
 
@@ -254,23 +240,15 @@ def create_rag_chain(
     """
 
     # Log progress and validate prompt template
-    logger.info('RAG chain - Validating LLM prompt template')
-    validate_prompt_template(
-        request.question_answering_prompt, 'Question answering prompt'
-    )
+    logger.info("RAG chain - Validating LLM prompt template")
+    validate_prompt_template(request.question_answering_prompt, "Question answering prompt")
     if request.question_condensing_prompt is not None:
-        validate_prompt_template(
-            request.question_condensing_prompt, 'Question condensing prompt'
-        )
+        validate_prompt_template(request.question_condensing_prompt, "Question condensing prompt")
 
     question_condensing_llm_factory = None
     if request.question_condensing_llm_setting is not None:
-        question_condensing_llm_factory = get_llm_factory(
-            setting=request.question_condensing_llm_setting
-        )
-    question_answering_llm_factory = get_llm_factory(
-        setting=request.question_answering_llm_setting
-    )
+        question_condensing_llm_factory = get_llm_factory(setting=request.question_condensing_llm_setting)
+    question_answering_llm_factory = get_llm_factory(setting=request.question_answering_llm_setting)
     em_factory = get_em_factory(setting=request.embedding_question_em_setting)
     vector_store_factory = get_vector_store_factory(
         setting=request.vector_store_setting,
@@ -285,7 +263,7 @@ def create_rag_chain(
     if request.compressor_setting:
         retriever = add_document_compressor(retriever, request.compressor_setting)
 
-    logger.debug('RAG chain - Document index name: %s', request.document_index_name)
+    logger.debug("RAG chain - Document index name: %s", request.document_index_name)
 
     # Build LLM and prompt templates
     question_condensing_llm = None
@@ -300,9 +278,7 @@ def create_rag_chain(
         condensing_llm = question_answering_llm
 
     # Build the chat chain for question contextualization
-    chat_chain = build_question_condensation_chain(
-        condensing_llm, request.question_condensing_prompt
-    )
+    chat_chain = build_question_condensation_chain(condensing_llm, request.question_condensing_prompt)
     rag_prompt = build_rag_prompt(request)
 
     # Function to contextualize the question based on chat history
@@ -311,22 +287,22 @@ def create_rag_chain(
     # Calculate the condensed question
     with_condensed_question = RunnableParallel(
         {
-            'condensed_question': contextualize_question_fn,
-            'question': itemgetter('question'),
-            'chat_history': itemgetter('chat_history'),
+            "condensed_question": contextualize_question_fn,
+            "question": itemgetter("question"),
+            "chat_history": itemgetter("chat_history"),
         }
     )
 
     def retrieve_with_variants(inputs):
         variants = [
             # inputs["question"], Deactivated. It's an example to prove the multi retriever process
-            inputs['condensed_question']
+            inputs["condensed_question"]
         ]
         docs = []
         for v in variants:
             docs.extend(retriever.invoke(v))
         # Deduplicate docs
-        unique_docs = {d.metadata['id']: d for d in docs}
+        unique_docs = {d.metadata["id"]: d for d in docs}
 
         # TODO [DERCBOT-1649] Apply the RRF Algo on unique_docs.
         return list(unique_docs.values())
@@ -334,33 +310,31 @@ def retrieve_with_variants(inputs):
     # Build the RAG inputs
     rag_inputs = with_condensed_question | RunnableParallel(
         {
-            'question': itemgetter('condensed_question'),
-            'chat_history': itemgetter('chat_history'),
-            'documents': RunnableLambda(retrieve_with_variants),
+            "question": itemgetter("condensed_question"),
+            "chat_history": itemgetter("chat_history"),
+            "documents": RunnableLambda(retrieve_with_variants),
         }
     )
 
-    parser = PydanticOutputParser(pydantic_object=LLMAnswer, name='rag_chain_output')
-
     return rag_inputs | RunnablePassthrough.assign(
         answer=(
             {
-                'context': lambda x: json.dumps(
+                "context": lambda x: json.dumps(
                     [
                         {
-                            'chunk_id': doc.metadata['id'],
-                            'chunk_text': doc.page_content,
+                            "chunk_id": doc.metadata["id"],
+                            "chunk_text": doc.page_content,
                         }
-                        for doc in x['documents']
+                        for doc in x["documents"]
                     ],
                     ensure_ascii=False,
                     indent=2,
                 ),
-                'chat_history': format_chat_history,
+                "chat_history": format_chat_history,
             }
             | rag_prompt
             | question_answering_llm
-            | parser
+            | JsonOutputParser(pydantic_object=LLMAnswer, name="rag_chain_output")
         )
     )
 
@@ -378,17 +352,15 @@ def build_rag_prompt(request: RAGRequest) -> LangChainPromptTemplate:
 
 def format_chat_history(x):
     messages = []
-    for msg in x['chat_history']:
+    for msg in x["chat_history"]:
         if isinstance(msg, HumanMessage):
-            messages.append({'user': msg.content})
+            messages.append({"user": msg.content})
         elif isinstance(msg, AIMessage):
-            messages.append({'assistant': msg.content})
+            messages.append({"assistant": msg.content})
     return json.dumps(messages, ensure_ascii=False, indent=2)
 
 
-def build_question_condensation_chain(
-    llm, prompt: Optional[PromptTemplate]
-) -> ChatPromptTemplate:
+def build_question_condensation_chain(llm, prompt: Optional[PromptTemplate]) -> ChatPromptTemplate:
     """
     Build the chat chain for contextualizing questions.
     """
@@ -418,13 +390,13 @@ def build_question_condensation_chain(
     return (
         ChatPromptTemplate.from_messages(
             [
-                ('system', prompt.template),
-                MessagesPlaceholder(variable_name='chat_history'),
-                ('human', '{question}'),
+                ("system", prompt.template),
+                MessagesPlaceholder(variable_name="chat_history"),
+                ("human", "{question}"),
             ]
         ).partial(**prompt.inputs)
         | llm
-        | StrOutputParser(name='chat_chain_output')
+        | StrOutputParser(name="chat_chain_output")
     )
 
 
@@ -432,9 +404,33 @@ def contextualize_question(inputs: dict, chat_chain) -> str:
     """
     Contextualize the question based on the chat history.
     """
-    if inputs.get('chat_history') and len(inputs['chat_history']) > 0:
+    if inputs.get("chat_history") and len(inputs["chat_history"]) > 0:
         return chat_chain
-    return inputs['question']
+    return inputs["question"]
+
+
+def rag_guard(question, answer, response, documents_required):
+    """
+    Validates the RAG system's response based on the presence or absence of source documents
+    and the `documentsRequired` setting.
+
+    Args:
+        question: user question
+        answer: the LLM answer
+        response: the RAG response
+        documents_required (bool): Specifies whether documents are mandatory for the response.
+    """
+
+    if documents_required and answer.status == "found_in_context" and len(response["documents"]) == 0:
+        message = "No documents were retrieved, yet an answer was attempted."
+        rag_log(level=ERROR, message=message, question=question, answer=answer.answer, response=response)
+        raise GenAIGuardCheckException(ErrorInfo(cause=message))
+
+    if answer.status == "not_found_in_context" and len(response["documents"]) > 0:
+        # If the answer is not found in context and some documents are retrieved, so we remove them from the RAG response.
+        message = "No answer found in the retrieved context. The documents are therefore removed from the RAG response."
+        rag_log(level=WARNING, message=message, question=question, answer=answer.answer, response=response)
+        response["documents"] = []
 
 
 def rag_log(level, message, question, answer, response):
@@ -451,13 +447,12 @@ def rag_log(level, message, question, answer, response):
 
     logger.log(
         level,
-        '%(message)s \n'
-        'RAG chain - question="%(question)s", answer="%(answer)s", documents="%(documents)s"',
+        "%(message)s \n" 'RAG chain - question="%(question)s", answer="%(answer)s", documents="%(documents)s"',
         {
-            'message': message,
-            'question': question,
-            'answer': answer,
-            'documents': len(response['documents']),
+            "message": message,
+            "question": question,
+            "answer": answer,
+            "documents": len(response["documents"]),
         },
     )
 
@@ -470,33 +465,31 @@ def get_rag_documents(handler: RAGCallbackHandler) -> List[RAGDocument]:
         handler: the RAG Callback Handler
     """
 
-    if handler.records['documents'] is None:
+    if handler.records["documents"] is None:
         return []
 
     return [
         # Get first 100 char of content
         RAGDocument(
-            content=doc.page_content[0 : len(doc.metadata['title']) + 100] + '...',
+            content=doc.page_content[0 : len(doc.metadata["title"]) + 100] + "...",
             metadata=RAGDocumentMetadata(**doc.metadata),
         )
-        for doc in handler.records['documents']
+        for doc in handler.records["documents"]
     ]
 
 
+def get_llm_answer(rag_chain_output) -> LLMAnswer:
+    return LLMAnswer(**json.loads(rag_chain_output.strip().removeprefix("```json").removesuffix("```").strip()))
+
+
 def get_llm_answer(rag_chain_output) -> LLMAnswer:
     if rag_chain_output is None:
         return LLMAnswer()
 
-    return LLMAnswer(
-        **json.loads(
-            rag_chain_output.strip().removeprefix('```json').removesuffix('```').strip()
-        )
-    )
+    return LLMAnswer(**json.loads(rag_chain_output.strip().removeprefix("```json").removesuffix("```").strip()))
 
 
-def get_rag_debug_data(
-    request: RAGRequest, records_callback_handler: RAGCallbackHandler, rag_duration
-) -> RAGDebugData:
+def get_rag_debug_data(request: RAGRequest, records_callback_handler: RAGCallbackHandler, rag_duration) -> RAGDebugData:
     """RAG debug data assembly"""
 
     history = []
@@ -504,15 +497,15 @@ def get_rag_debug_data(
         history = request.dialog.history
 
     return RAGDebugData(
-        user_question=request.question_answering_prompt.inputs['question'],
-        question_condensing_prompt=records_callback_handler.records['chat_prompt'],
+        user_question=request.question_answering_prompt.inputs["question"],
+        question_condensing_prompt=records_callback_handler.records["chat_prompt"],
         question_condensing_history=history,
-        condensed_question=records_callback_handler.records['chat_chain_output'],
-        question_answering_prompt=records_callback_handler.records['rag_prompt'],
+        condensed_question=records_callback_handler.records["chat_chain_output"],
+        question_answering_prompt=records_callback_handler.records["rag_prompt"],
         documents=get_rag_documents(records_callback_handler),
         document_index_name=request.document_index_name,
         document_search_params=request.document_search_params,
-        answer=get_llm_answer(records_callback_handler.records['rag_chain_output']),
+        answer=get_llm_answer(records_callback_handler.records["rag_chain_output"]),
         duration=rag_duration,
     )
 
@@ -524,7 +517,7 @@ def check_guardrail_output(guardrail_output: dict) -> bool:
     Returns:
         Returns True if nothing is detected, raises an exception otherwise.
     """
-    if guardrail_output['output_toxicity']:
+    if guardrail_output["output_toxicity"]:
         message = f"Toxicity detected in LLM output ({','.join(guardrail_output['output_toxicity_reason'])})"
         raise GenAIGuardCheckException(ErrorInfo(cause=message))
     return True
diff --git a/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py b/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py
index ef16f03eb9..46589feb0e 100644
--- a/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py
+++ b/gen-ai/orchestrator-server/src/main/python/server/tests/services/test_rag_chain.py
@@ -44,25 +44,20 @@
 )
 
 
-@patch(
-    'gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post'
-)
-@patch(
-    'gen_ai_orchestrator.services.langchain.factories.langchain_factory.get_compressor_factory'
-)
-@patch(
-    'gen_ai_orchestrator.services.langchain.factories.langchain_factory.get_callback_handler_factory'
-)
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.create_rag_chain')
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.RAGCallbackHandler')
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.RAGResponse')
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.RAGDebugData')
-@patch('gen_ai_orchestrator.services.langchain.rag_chain.get_llm_answer')
+@patch("gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post")
+@patch("gen_ai_orchestrator.services.langchain.factories.langchain_factory.get_compressor_factory")
+@patch("gen_ai_orchestrator.services.langchain.factories.langchain_factory.get_callback_handler_factory")
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.create_rag_chain")
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.RAGCallbackHandler")
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.RAGResponse")
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.RAGDebugData")
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.get_llm_answer")
 @pytest.mark.asyncio
 async def test_rag_chain(
     mocked_get_llm_answer,
     mocked_rag_debug_data,
     mocked_rag_response,
+    mocked_rag_guard,
     mocked_callback_init,
     mocked_create_rag_chain,
     mocked_get_callback_handler_factory,
@@ -72,28 +67,28 @@ async def test_rag_chain(
     """Test the full execute_qa_chain method by mocking all external calls."""
     # Build a test RAGRequest
     query_dict = {
-        'dialog': {
-            'history': [
-                {'text': 'Hello, how can I do this?', 'type': 'HUMAN'},
+        "dialog": {
+            "history": [
+                {"text": "Hello, how can I do this?", "type": "HUMAN"},
                 {
-                    'text': 'you can do this with the following method ....',
-                    'type': 'AI',
+                    "text": "you can do this with the following method ....",
+                    "type": "AI",
                 },
             ],
-            'tags': [],
+            "tags": [],
         },
-        'question_answering_llm_setting': {
-            'provider': 'OpenAI',
-            'api_key': {
-                'type': 'Raw',
-                'secret': 'ab7***************************A1IV4B',
+        "question_answering_llm_setting": {
+            "provider": "OpenAI",
+            "api_key": {
+                "type": "Raw",
+                "secret": "ab7***************************A1IV4B",
             },
-            'temperature': 1.2,
-            'model': 'gpt-3.5-turbo',
+            "temperature": 1.2,
+            "model": "gpt-3.5-turbo",
         },
-        'question_answering_prompt': {
-            'formatter': 'f-string',
-            'template': """Use the following context to answer the question at the end.
+        "question_answering_prompt": {
+            "formatter": "f-string",
+            "template": """Use the following context to answer the question at the end.
 If you don't know the answer, just say I don't know.
 
 Context:
@@ -103,89 +98,81 @@ async def test_rag_chain(
 {question}
 
 Answer in {locale}:""",
-            'inputs': {
-                'question': 'How to get started playing guitar ?',
-                'locale': 'French',
+            "inputs": {
+                "question": "How to get started playing guitar ?",
+                "locale": "French",
             },
         },
-        'embedding_question_em_setting': {
-            'provider': 'OpenAI',
-            'api_key': {
-                'type': 'Raw',
-                'secret': 'ab7***************************A1IV4B',
+        "embedding_question_em_setting": {
+            "provider": "OpenAI",
+            "api_key": {
+                "type": "Raw",
+                "secret": "ab7***************************A1IV4B",
             },
-            'model': 'text-embedding-ada-002',
+            "model": "text-embedding-ada-002",
         },
-        'document_index_name': 'my-index-name',
-        'document_search_params': {
-            'provider': 'OpenSearch',
-            'filter': [
-                {
-                    'term': {
-                        'metadata.index_session_id.keyword': '352d2466-17c5-4250-ab20-d7c823daf035'
-                    }
-                }
-            ],
-            'k': 4,
+        "document_index_name": "my-index-name",
+        "document_search_params": {
+            "provider": "OpenSearch",
+            "filter": [{"term": {"metadata.index_session_id.keyword": "352d2466-17c5-4250-ab20-d7c823daf035"}}],
+            "k": 4,
         },
-        'vector_store_setting': {
-            'provider': 'OpenSearch',
-            'host': 'localhost',
-            'port': 9200,
-            'username': 'admin',
-            'password': {
-                'type': 'Raw',
-                'secret': 'admin',
+        "vector_store_setting": {
+            "provider": "OpenSearch",
+            "host": "localhost",
+            "port": 9200,
+            "username": "admin",
+            "password": {
+                "type": "Raw",
+                "secret": "admin",
             },
         },
-        'observability_setting': {
-            'provider': 'Langfuse',
-            'url': 'http://localhost:3000',
-            'secret_key': {
-                'type': 'Raw',
-                'secret': 'sk-********************be8f',
+        "observability_setting": {
+            "provider": "Langfuse",
+            "url": "http://localhost:3000",
+            "secret_key": {
+                "type": "Raw",
+                "secret": "sk-********************be8f",
             },
-            'public_key': 'pk-lf-5e374dc6-e194-4b37-9c07-b77e68ef7d2c',
+            "public_key": "pk-lf-5e374dc6-e194-4b37-9c07-b77e68ef7d2c",
         },
-        'guardrail_setting': {
-            'provider': 'BloomzGuardrail',
-            'api_base': 'http://test-guard.com',
-            'max_score': 0.5,
+        "guardrail_setting": {
+            "provider": "BloomzGuardrail",
+            "api_base": "http://test-guard.com",
+            "max_score": 0.5,
         },
-        'compressor_setting': {
-            'provider': 'BloomzRerank',
-            'min_score': 0.7,
-            'endpoint': 'http://test-rerank.com',
+        "compressor_setting": {
+            "provider": "BloomzRerank",
+            "min_score": 0.7,
+            "endpoint": "http://test-rerank.com",
         },
-        'documents_required': True,
+        "documents_required": True,
     }
     request = RAGRequest(**query_dict)
     inputs = {
         **request.question_answering_prompt.inputs,
-        'chat_history': [
-            HumanMessage(content='Hello, how can I do this?'),
-            AIMessage(content='you can do this with the following method ....'),
+        "chat_history": [
+            HumanMessage(content="Hello, how can I do this?"),
+            AIMessage(content="you can do this with the following method ...."),
         ],
     }
     docs = [
         Document(
-            page_content='some page content',
-            metadata={'id': '123-abc', 'title': 'my-title', 'source': None},
+            page_content="some page content",
+            metadata={"id": "123-abc", "title": "my-title", "source": None},
         )
     ]
     response = {
-        'answer': LLMAnswer(
-            **{
-                'status': '',
-                'answer': 'an answer from llm',
-                'topic': None,
-                'suggested_topics': None,
-                'context': [],
-            }
-        ),
-        'documents': docs,
+        "answer": {
+            "status": "",
+            "answer": "an answer from llm",
+            "topic": None,
+            "suggested_topics": None,
+            "context": [],
+        },
+        "documents": docs,
     }
-    llm_answer = response['answer']
+    llm_answer = LLMAnswer(**response["answer"])
 
     # Setup mock factories/init return value
     observability_factory_instance = mocked_get_callback_handler_factory.return_value
@@ -197,7 +184,7 @@ async def test_rag_chain(
 
     mocked_response = MagicMock()
     mocked_response.status_code = 200
-    mocked_response.content = {'response': []}
+    mocked_response.content = {"response": []}
 
     mocked_guardrail_parse.return_value = mocked_response
 
@@ -205,15 +192,13 @@ async def test_rag_chain(
     await execute_rag_chain(request, debug=True)
 
     # Assert that the given observability_setting is used
-    mocked_get_callback_handler_factory.assert_called_once_with(
-        setting=request.observability_setting
-    )
+    mocked_get_callback_handler_factory.assert_called_once_with(setting=request.observability_setting)
     # Assert qa chain is ainvoke()d with the expected settings from request
     mocked_chain.ainvoke.assert_called_once_with(
         input=inputs,
         config={
-            'callbacks': [mocked_callback, mocked_langfuse_callback],
-            'metadata': ANY,
+            "callbacks": [mocked_callback, mocked_langfuse_callback],
+            "metadata": ANY,
         },
     )
     # Assert the response is build using the expected settings
@@ -226,77 +211,69 @@ async def test_rag_chain(
     mocked_get_document_compressor_factory(setting=request.compressor_setting)
     # Assert the rag guardrail is called
     mocked_guardrail_parse.assert_called_once_with(
-        os.path.join(request.guardrail_setting.api_base, 'guardrail'),
-        json={
-            'text': [mocked_rag_answer['answer'].answer]
-        },  # TODO MASS : answer.answer pas top, à renomer
+        os.path.join(request.guardrail_setting.api_base, "guardrail"),
+        json={"text": [mocked_rag_answer["answer"]["answer"]]},
     )
+    # Assert the rag guard is called
+    mocked_rag_guard.assert_called_once_with(inputs, llm_answer, response, request.documents_required)
 
 
-@patch(
-    'gen_ai_orchestrator.services.langchain.impls.guardrail.bloomz_guardrail.requests.post'
-)
+@patch("gen_ai_orchestrator.services.langchain.impls.guardrail.bloomz_guardrail.requests.post")
 def test_guardrail_parse_succeed_with_toxicities_encountered(
     mocked_guardrail_response,
 ):
     guardrail = get_guardrail_factory(
-        BloomzGuardrailSetting(
-            provider='BloomzGuardrail', max_score=0.5, api_base='http://test-guard.com'
-        )
+        BloomzGuardrailSetting(provider="BloomzGuardrail", max_score=0.5, api_base="http://test-guard.com")
     ).get_parser()
     rag_response = {
-        'answer': {
-            'status': '',
-            'answer': 'This is a sample text.',
-            'topic': None,
-            'suggested_topics': None,
-            'context': [],
+        "answer": {
+            "status": "",
+            "answer": "This is a sample text.",
+            "topic": None,
+            "suggested_topics": None,
+            "context": [],
         }
     }
 
     mocked_response = MagicMock()
     mocked_response.status_code = 200
     mocked_response.json.return_value = {
-        'response': [
+        "response": [
             [
-                {'label': 'racism', 'score': 0.1},
-                {'label': 'insult', 'score': 0.2},
-                {'label': 'threat', 'score': 0.7},
-                {'label': 'hate speech', 'score': 0.95},
+                {"label": "racism", "score": 0.1},
+                {"label": "insult", "score": 0.2},
+                {"label": "threat", "score": 0.7},
+                {"label": "hate speech", "score": 0.95},
             ]
         ]
     }
 
     mocked_guardrail_response.return_value = mocked_response
-    guardrail_output = guardrail.parse(rag_response['answer']['answer'])
+    guardrail_output = guardrail.parse(rag_response["answer"]["answer"])
 
     mocked_guardrail_response.assert_called_once_with(
-        os.path.join(guardrail.endpoint, 'guardrail'),
-        json={'text': [rag_response['answer']['answer']]},
+        os.path.join(guardrail.endpoint, "guardrail"),
+        json={"text": [rag_response["answer"]["answer"]]},
     )
     assert guardrail_output == {
-        'content': 'This is a sample text.',
-        'output_toxicity': True,
-        'output_toxicity_reason': ['threat', 'hate speech'],
+        "content": "This is a sample text.",
+        "output_toxicity": True,
+        "output_toxicity_reason": ["threat", "hate speech"],
     }
 
 
-@patch(
-    'gen_ai_orchestrator.services.langchain.impls.guardrail.bloomz_guardrail.requests.post'
-)
+@patch("gen_ai_orchestrator.services.langchain.impls.guardrail.bloomz_guardrail.requests.post")
 def test_guardrail_parse_fail(mocked_guardrail_response):
     guardrail = get_guardrail_factory(
-        BloomzGuardrailSetting(
-            provider='BloomzGuardrail', max_score=0.5, api_base='http://test-guard.com'
-        )
+        BloomzGuardrailSetting(provider="BloomzGuardrail", max_score=0.5, api_base="http://test-guard.com")
     ).get_parser()
     rag_response = {
-        'answer': {
-            'status': '',
-            'answer': 'This is a sample text.',
-            'topic': None,
-            'suggested_topics': None,
-            'context': [],
+        "answer": {
+            "status": "",
+            "answer": "This is a sample text.",
+            "topic": None,
+            "suggested_topics": None,
+            "context": [],
         }
     }
 
@@ -308,49 +285,47 @@ def test_guardrail_parse_fail(mocked_guardrail_response):
         HTTPError,
         match=f"Error {mocked_response.status_code}. Bloomz guardrail didn't respond as expected.",
     ):
-        guardrail.parse(rag_response['answer']['answer'])
+        guardrail.parse(rag_response["answer"]["answer"])
 
     mocked_guardrail_response.assert_called_once_with(
-        os.path.join(guardrail.endpoint, 'guardrail'),
-        json={'text': [rag_response['answer']['answer']]},
+        os.path.join(guardrail.endpoint, "guardrail"),
+        json={"text": [rag_response["answer"]["answer"]]},
     )
 
 
-@patch(
-    'gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post'
-)
+@patch("gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post")
 def test_compress_documents_should_succeed(mocked_rerank):
-    bloomz_reranker = BloomzRerank(label='entailement', endpoint='http://example.com')
+    bloomz_reranker = BloomzRerank(label="entailement", endpoint="http://example.com")
     documents = [
         Document(
-            page_content='Page content 1',
+            page_content="Page content 1",
             metadata={
-                'source': 'doc1.pdf',
-                'file_path': 'doc1.pdf',
-                'page': 57,
-                'total_pages': 104,
-                'Producer': 'GPL Ghostscript 9.05',
-                'CreationDate': "D:20230828165103+02'00'",
-                'ModDate': "D:20230828165103+02'00'",
-                'Title': 'DGRC 2023 - ABEI',
-                'Creator': 'PDFCreator Version 1.6.2',
-                'Author': 'F9261',
-                'Keywords': '',
-                'Subject': '',
+                "source": "doc1.pdf",
+                "file_path": "doc1.pdf",
+                "page": 57,
+                "total_pages": 104,
+                "Producer": "GPL Ghostscript 9.05",
+                "CreationDate": "D:20230828165103+02'00'",
+                "ModDate": "D:20230828165103+02'00'",
+                "Title": "DGRC 2023 - ABEI",
+                "Creator": "PDFCreator Version 1.6.2",
+                "Author": "F9261",
+                "Keywords": "",
+                "Subject": "",
             },
         ),
         Document(
-            page_content='Contenu du document 8',
+            page_content="Contenu du document 8",
             metadata={
-                'source': 'incident - v5.pdf',
-                'file_path': 'incident - v5.pdf',
-                'page': 19,
-                'total_pages': 23,
-                'Author': 'F0421',
-                'CreationDate': "D:20231212161411+01'00'",
-                'ModDate': "D:20231212161411+01'00'",
-                'Producer': 'Microsoft: Print To PDF',
-                'Title': "Microsoft Word - P1 - Département des Risques - Déclaration d'un incident - v5.doc",
+                "source": "incident - v5.pdf",
+                "file_path": "incident - v5.pdf",
+                "page": 19,
+                "total_pages": 23,
+                "Author": "F0421",
+                "CreationDate": "D:20231212161411+01'00'",
+                "ModDate": "D:20231212161411+01'00'",
+                "Producer": "Microsoft: Print To PDF",
+                "Title": "Microsoft Word - P1 - Département des Risques - Déclaration d'un incident - v5.doc",
             },
         ),
     ]
@@ -358,59 +333,57 @@ def test_compress_documents_should_succeed(mocked_rerank):
     mocked_response = MagicMock()
     mocked_response.status_code = 200
     mocked_response.json.return_value = {
-        'response': [
+        "response": [
             [
-                {'label': 'entailement', 'score': 0.1},
-                {'label': 'neutral', 'score': 0.2},
-                {'label': 'contradiction', 'score': 0.7},
+                {"label": "entailement", "score": 0.1},
+                {"label": "neutral", "score": 0.2},
+                {"label": "contradiction", "score": 0.7},
             ],
             [
-                {'label': 'entailement', 'score': 0.8},
-                {'label': 'neutral', 'score': 0.2},
-                {'label': 'contradiction', 'score': 0.7},
+                {"label": "entailement", "score": 0.8},
+                {"label": "neutral", "score": 0.2},
+                {"label": "contradiction", "score": 0.7},
             ],
         ]
     }
     mocked_rerank.return_value = mocked_response
 
-    result = bloomz_reranker.compress_documents(documents=documents, query='Some query')
+    result = bloomz_reranker.compress_documents(documents=documents, query="Some query")
 
     mocked_rerank.assert_called_once()
     assert result == [
         Document(
-            page_content='Contenu du document 8',
+            page_content="Contenu du document 8",
             metadata={
-                'source': 'incident - v5.pdf',
-                'file_path': 'incident - v5.pdf',
-                'page': 19,
-                'total_pages': 23,
-                'Author': 'F0421',
-                'CreationDate': "D:20231212161411+01'00'",
-                'ModDate': "D:20231212161411+01'00'",
-                'Producer': 'Microsoft: Print To PDF',
-                'Title': "Microsoft Word - P1 - Département des Risques - Déclaration d'un incident - v5.doc",
-                'retriever_score': 0.8,
+                "source": "incident - v5.pdf",
+                "file_path": "incident - v5.pdf",
+                "page": 19,
+                "total_pages": 23,
+                "Author": "F0421",
+                "CreationDate": "D:20231212161411+01'00'",
+                "ModDate": "D:20231212161411+01'00'",
+                "Producer": "Microsoft: Print To PDF",
+                "Title": "Microsoft Word - P1 - Département des Risques - Déclaration d'un incident - v5.doc",
+                "retriever_score": 0.8,
             },
         )
     ]
 
 
-@patch(
-    'gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post'
-)
+@patch("gen_ai_orchestrator.services.langchain.impls.document_compressor.bloomz_rerank.requests.post")
 def test_compress_documents_with_unknown_label(mocked_rerank):
-    bloomz_reranker = BloomzRerank(label='unknown_label', endpoint='http://example.com')
+    bloomz_reranker = BloomzRerank(label="unknown_label", endpoint="http://example.com")
     documents = [
         Document(
-            page_content='Page content 1',
+            page_content="Page content 1",
             metadata={
-                'source': 'doc1.pdf',
+                "source": "doc1.pdf",
             },
         ),
         Document(
-            page_content='Contenu du document 8',
+            page_content="Contenu du document 8",
             metadata={
-                'source': 'incident - v5.pdf',
+                "source": "incident - v5.pdf",
             },
         ),
     ]
@@ -418,41 +391,119 @@ def test_compress_documents_with_unknown_label(mocked_rerank):
     mocked_response = MagicMock()
     mocked_response.status_code = 200
     mocked_response.json.return_value = {
-        'response': [
+        "response": [
             [
-                {'label': 'entailement', 'score': 0.1},
+                {"label": "entailement", "score": 0.1},
             ]
         ]
     }
     mocked_rerank.return_value = mocked_response
 
     with pytest.raises(GenAIDocumentCompressorUnknownLabelException) as exc:
-        bloomz_reranker.compress_documents(documents=documents, query='Some query')
+        bloomz_reranker.compress_documents(documents=documents, query="Some query")
 
     assert exc.value.error_code.value == 6002
-    assert exc.value.message == 'Unknown Document Compressor label.'
-    assert exc.value.detail == 'Check the Document Compressor label you sent.'
+    assert exc.value.message == "Unknown Document Compressor label."
+    assert exc.value.detail == "Check the Document Compressor label you sent."
 
 
 def test_check_guardrail_output_find_toxicities():
     guardrail_output = {
-        'content': 'This is a sample text.',
-        'output_toxicity': True,
-        'output_toxicity_reason': ['threat', 'hate speech'],
+        "content": "This is a sample text.",
+        "output_toxicity": True,
+        "output_toxicity_reason": ["threat", "hate speech"],
     }
 
     with pytest.raises(GenAIGuardCheckException) as exc_found:
         check_guardrail_output(guardrail_output)
 
     assert exc_found.value.error_code.value == 1004
-    assert 'Guard check failed.' in exc_found.value.message
+    assert "Guard check failed." in exc_found.value.message
 
 
 def test_check_guardrail_output_is_ok():
     guardrail_output = {
-        'content': 'This is a sample text.',
-        'output_toxicity': False,
-        'output_toxicity_reason': [],
+        "content": "This is a sample text.",
+        "output_toxicity": False,
+        "output_toxicity_reason": [],
     }
 
     assert check_guardrail_output(guardrail_output) is True
+
+
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.rag_log")
+def test_rag_guard_fails_if_no_docs_in_valid_answer(mocked_log):
+    question = "Hi!"
+    response = {
+        "answer": {"status": "found_in_context", "answer": "a valid answer"},
+        "documents": [],
+    }
+    try:
+        rag_chain.rag_guard(question, LLMAnswer(**response["answer"]), response, documents_required=True)
+    except Exception as e:
+        assert isinstance(e, GenAIGuardCheckException)
+
+
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.rag_log")
+def test_rag_guard_accepts_no_answer_even_with_docs(mocked_log):
+    question = "Hi!"
+    response = {
+        "answer": {
+            "status": "not_found_in_context",
+            "answer": "Sorry, I don t know.",
+            "context": [
+                {
+                    "chunk": 1,
+                    "sentences": ["str1"],
+                }
+            ],
+        },
+        "documents": ["a doc as a string"],
+    }
+    rag_chain.rag_guard(question, LLMAnswer(**response["answer"]), response, documents_required=True)
+    # No answer found in the retrieved context. The documents are therefore removed from the RAG response.
+    assert response["documents"] == []
+
+
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.rag_log")
+def test_rag_guard_valid_answer_with_docs(mocked_log):
+    question = "Hi!"
+    response = {
+        "answer": {
+            "status": "found_in_context",
+            "answer": "a valid answer",
+        },
+        "documents": ["doc1", "doc2"],
+    }
+    rag_chain.rag_guard(question, LLMAnswer(**response["answer"]), response, documents_required=True)
+    assert response["documents"] == ["doc1", "doc2"]
+
+
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.rag_log")
+def test_rag_guard_no_answer_with_no_docs(mocked_log):
+    question = "Hi!"
+    response = {
+        "answer": {"status": "not_found_in_context", "answer": "Sorry, I don t know."},
+        "documents": [],
+    }
+    rag_chain.rag_guard(question, LLMAnswer(**response["answer"]), response, documents_required=True)
+    assert response["documents"] == []
+
+
+@patch("gen_ai_orchestrator.services.langchain.rag_chain.rag_log")
+def test_rag_guard_without_no_answer_input(mocked_log):
+    """Test that __rag_guard handles missing no_answer input correctly."""
+    question = "Hi!"
+    response = {
+        "answer": {
+            "status": "found_in_context",
+            "answer": "a valid answer",
+        },
+        "documents": [],
+    }
+    with pytest.raises(GenAIGuardCheckException) as exc:
+        rag_chain.rag_guard(question, LLMAnswer(**response["answer"]), response, documents_required=True)
+
+    mocked_log.assert_called_once()
+
+    assert isinstance(exc.value, GenAIGuardCheckException)

From 6ba6d90bc22fcf08da5c9fffab1e3e6776a8fd22 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rodolphe=20K=C3=BCffer?= <rodolphe.kuffer@partnre.com>
Date: Mon, 2 Feb 2026 17:30:20 +0100
Subject: [PATCH 4/7] Front modifications

---
 .../chat-ui-message-debug.component.html              | 11 ++---------
 1 file changed, 2 insertions(+), 9 deletions(-)

diff --git a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
index 361645b06d..a07ba89667 100644
--- a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
+++ b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
@@ -21,20 +21,13 @@
 >
   <div><strong>Status:</strong> {{ message.data.answer?.status }}</div>
 
+  <div *ngIf="message.data.answer?.confidence_score"><strong>Confidence score:</strong>: {{ message.data.answer?.confidence_score }}</div>
+
   <div *ngIf="message.data.answer?.topic && message.data.answer?.topic !== 'unknown'">
     <strong>Topic:</strong> {{ message.data.answer?.topic }}
   </div>
 
-  <div *ngIf="message.data.answer?.suggested_topics?.length">
-    <strong>Suggested topics:</strong>
-    <span *ngFor="let suggestion of message.data.answer.suggested_topics">
-      {{ suggestion }}
-    </span>
-  </div>
-
   <div *ngIf="message.data.answer?.redirection_intent">
     <strong>Redirection intent:</strong> {{ message.data.answer?.redirection_intent }}
   </div>
-
-  <div *ngIf="message.data.answer?.confidence_score"><strong>Confidence score:</strong>: {{ message.data.answer?.confidence_score }}</div>
 </div>

From 2db19a1ecff99f87d49884bc3d77dce582bf6a9f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rodolphe=20K=C3=BCffer?= <rodolphe.kuffer@partnre.com>
Date: Tue, 3 Feb 2026 14:53:53 +0100
Subject: [PATCH 5/7] Front modifications - Wip

---
 .../chat-ui-message-debug.component.html              | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
index a07ba89667..361645b06d 100644
--- a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
+++ b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
@@ -21,13 +21,20 @@
 >
   <div><strong>Status:</strong> {{ message.data.answer?.status }}</div>
 
-  <div *ngIf="message.data.answer?.confidence_score"><strong>Confidence score:</strong>: {{ message.data.answer?.confidence_score }}</div>
-
   <div *ngIf="message.data.answer?.topic && message.data.answer?.topic !== 'unknown'">
     <strong>Topic:</strong> {{ message.data.answer?.topic }}
   </div>
 
+  <div *ngIf="message.data.answer?.suggested_topics?.length">
+    <strong>Suggested topics:</strong>
+    <span *ngFor="let suggestion of message.data.answer.suggested_topics">
+      {{ suggestion }}
+    </span>
+  </div>
+
   <div *ngIf="message.data.answer?.redirection_intent">
     <strong>Redirection intent:</strong> {{ message.data.answer?.redirection_intent }}
   </div>
+
+  <div *ngIf="message.data.answer?.confidence_score"><strong>Confidence score:</strong>: {{ message.data.answer?.confidence_score }}</div>
 </div>

From dacc9a58876534013aa4b13b13cd279ffbb92854 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rodolphe=20K=C3=BCffer?= <rodolphe.kuffer@partnre.com>
Date: Wed, 4 Feb 2026 17:40:00 +0100
Subject: [PATCH 6/7] Remove documentsRequired Rag option + UX improvements

---
 .../models/engines-configurations.ts          |  3 +-
 .../rag/rag-settings/models/rag-settings.ts   |  2 -
 .../rag-settings/rag-settings.component.html  | 18 ------
 .../rag-settings/rag-settings.component.ts    |  8 ---
 .../chat-ui-message-debug.component.html      | 60 +++++++++++++-----
 .../chat-ui-message-debug.component.scss      | 61 +++++++++++++------
 .../chat-ui-message-debug.component.ts        | 43 ++++++++++++-
 .../web/src/app/shared/model/dialog-data.ts   | 14 +++++
 bot/admin/web/src/app/shared/utils/utils.ts   | 36 +++++++++++
 .../components/header/header.component.html   | 10 +++
 .../components/header/header.component.ts     |  4 ++
 .../web/src/app/theme/styles/utilities.scss   |  4 ++
 12 files changed, 198 insertions(+), 65 deletions(-)

diff --git a/bot/admin/web/src/app/rag/rag-settings/models/engines-configurations.ts b/bot/admin/web/src/app/rag/rag-settings/models/engines-configurations.ts
index 758ababaa3..9beab4b51a 100644
--- a/bot/admin/web/src/app/rag/rag-settings/models/engines-configurations.ts
+++ b/bot/admin/web/src/app/rag/rag-settings/models/engines-configurations.ts
@@ -70,7 +70,7 @@ Your job is to surface the right information from provided context.
 
 - **Tone**: neutral, kind, “you” address, light humor when appropriate.
 - **Language**: Introduce technical jargon only when strictly necessary and briefly define it.
-- **Structure**: Use short sentences, bold or bullet points for key ideas, headings to separate the main sections, and fenced \`code\` blocks for examples.
+- **Structure**: Use short sentences, bold or bullet points for key ideas, headings to separate the main sections, and fenced \`code\` blocks for examples. Only include absolute links in your answers.
 - **Style**: Direct and technical tone, with **bold** for important concepts.
 - **Formatting**: Mandatory Markdown, with line breaks for readability.
 - **Examples**: Include a concrete example (code block or CLI command) for each feature.
@@ -89,6 +89,7 @@ Before responding, ensure:
 
 - The documentation actually addresses the question.
 - Your answer is consistent with the docs.
+- If you include a link in your response, make sure it is an absolute link; otherwise, do not include it.
 
 ## Technical Instructions:
 
diff --git a/bot/admin/web/src/app/rag/rag-settings/models/rag-settings.ts b/bot/admin/web/src/app/rag/rag-settings/models/rag-settings.ts
index e605ec6561..c29537b2bb 100644
--- a/bot/admin/web/src/app/rag/rag-settings/models/rag-settings.ts
+++ b/bot/admin/web/src/app/rag/rag-settings/models/rag-settings.ts
@@ -37,6 +37,4 @@ export interface RagSettings {
   indexName: string;
 
   maxDocumentsRetrieved: number;
-
-  documentsRequired: boolean;
 }
diff --git a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.html b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.html
index cf3527fa9f..7faf77dbdd 100644
--- a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.html
+++ b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.html
@@ -408,24 +408,6 @@ <h1 class="flex-grow-1">Rag settings</h1>
                 />
               </tock-form-control>
             </div>
-
-            <div class="col-6 px-3">
-              <tock-form-control
-                label="Documents required"
-                name="documentsRequired"
-                [controls]="documentsRequired"
-                [required]="false"
-                [boldLabel]="false"
-                information="Prohibits LLM from responding if no document could be retrieved. Prevents the bot from engaging in informal conversation and responding to more general questions."
-              >
-                <nb-toggle
-                  formControlName="documentsRequired"
-                  class="nb-toggle-small mt-1"
-                >
-                  Don't allow undocumented answers
-                </nb-toggle>
-              </tock-form-control>
-            </div>
           </div>
         </nb-card-body>
       </nb-card>
diff --git a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.ts b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.ts
index c906ffc48e..12d8cdee99 100644
--- a/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.ts
+++ b/bot/admin/web/src/app/rag/rag-settings/rag-settings.component.ts
@@ -50,8 +50,6 @@ interface RagSettingsForm {
 
   maxDocumentsRetrieved: FormControl<number>;
 
-  documentsRequired: FormControl<boolean>;
-
   questionCondensingLlmProvider: FormControl<AiEngineProvider>;
   questionCondensingLlmSetting: FormGroup<any>;
   questionCondensingPrompt: FormGroup<any>;
@@ -197,7 +195,6 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
     indexSessionId: new FormControl(undefined),
     indexName: new FormControl(undefined),
 
-    documentsRequired: new FormControl(undefined),
     maxDocumentsRetrieved: new FormControl(undefined),
 
     questionCondensingLlmProvider: new FormControl(undefined, [Validators.required]),
@@ -242,10 +239,6 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
     return this.form.get('indexSessionId') as FormControl;
   }
 
-  get documentsRequired(): FormControl {
-    return this.form.get('documentsRequired') as FormControl;
-  }
-
   get maxDocumentsRetrieved(): FormControl {
     return this.form.get('maxDocumentsRetrieved') as FormControl;
   }
@@ -363,7 +356,6 @@ export class RagSettingsComponent implements OnInit, OnDestroy {
 
   setFormDefaultValues(): void {
     this.form.patchValue({
-      documentsRequired: false,
       debugEnabled: false,
       maxMessagesFromHistory: 5,
       maxDocumentsRetrieved: 4
diff --git a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
index 361645b06d..09354d8922 100644
--- a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
+++ b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
@@ -14,27 +14,55 @@
   ~ limitations under the License.
   -->
 
-<div
-  (click)="showDebug()"
-  class="debug d-flex gap-1 p-2"
-  nbTooltip="Show debug infos"
->
-  <div><strong>Status:</strong> {{ message.data.answer?.status }}</div>
+<div class="debug d-flex justify-content-end pr-5 gap-05">
+  <div
+    class="tag"
+    [ngClass]="getStatusClassName()"
+    (click)="showDebug()"
+    nbTooltip="Show details"
+  >
+    Status | <strong class="ml-1">{{ getStatusLabel() }}</strong>
+  </div>
 
-  <div *ngIf="message.data.answer?.topic && message.data.answer?.topic !== 'unknown'">
-    <strong>Topic:</strong> {{ message.data.answer?.topic }}
+  <div
+    class="tag topic"
+    *ngIf="message.data.answer?.topic && message.data.answer?.topic !== 'unknown'"
+    (click)="showDebug()"
+    nbTooltip="Show details"
+  >
+    Topic | <strong class="ml-1">{{ message.data.answer?.topic }}</strong>
   </div>
 
-  <div *ngIf="message.data.answer?.suggested_topics?.length">
-    <strong>Suggested topics:</strong>
-    <span *ngFor="let suggestion of message.data.answer.suggested_topics">
-      {{ suggestion }}
-    </span>
+  <div
+    class="tag suggested-topics"
+    *ngIf="message.data.answer?.suggested_topics?.length"
+    (click)="showDebug()"
+    nbTooltip="Show details"
+  >
+    Suggested topics |
+    <strong class="ml-1">
+      <span *ngFor="let suggestion of message.data.answer.suggested_topics">
+        {{ suggestion }}
+      </span>
+    </strong>
   </div>
 
-  <div *ngIf="message.data.answer?.redirection_intent">
-    <strong>Redirection intent:</strong> {{ message.data.answer?.redirection_intent }}
+  <div
+    class="tag"
+    *ngIf="message.data.answer?.redirection_intent"
+    (click)="showDebug()"
+    nbTooltip="Show details"
+  >
+    Redirection intent | <strong class="ml-1">{{ message.data.answer?.redirection_intent }}</strong>
   </div>
 
-  <div *ngIf="message.data.answer?.confidence_score"><strong>Confidence score:</strong>: {{ message.data.answer?.confidence_score }}</div>
+  <div
+    class="tag"
+    *ngIf="message.data.answer?.confidence_score"
+    [ngStyle]="{ 'background-color': getConfidenceBgColor().bg, color: getConfidenceBgColor().fg }"
+    (click)="showDebug()"
+    nbTooltip="Show details"
+  >
+    Confidence | <strong class="ml-1">{{ message.data.answer?.confidence_score }}</strong>
+  </div>
 </div>
diff --git a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.scss b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.scss
index 5f59982a64..d698b164ed 100644
--- a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.scss
+++ b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.scss
@@ -14,31 +14,56 @@
  * limitations under the License.
  */
 
-@import '@nebular/theme/styles/theming';
-@import '@nebular/theme/styles/themes';
-
 :host {
   display: flex;
-  // justify-content: center;
 
   .debug {
     width: 100%;
-    cursor: pointer;
 
-    font-size: 0.75rem;
-    color: var(--chat-message-sender-text-color);
-    border-top: 1px dashed var(--chat-message-sender-text-color);
-    border-bottom: 1px dashed var(--chat-message-sender-text-color);
-    background-color: var(--background-basic-color-3);
+    margin-top: 0.5em;
+    margin-bottom: 0.5em;
+
+    .tag {
+      display: inline-flex;
+      width: fit-content;
+      padding: 0.2rem 0.5rem;
+      border-radius: 10rem;
+      background-color: var(--border-basic-color-3);
+      font-size: 0.7rem;
+      line-height: 1rem;
+      white-space: nowrap;
+      cursor: pointer;
+
+      &.status-success {
+        background-color: var(--color-success-500);
+        color: white;
+      }
+      &.status-warning {
+        background-color: var(--color-warning-500);
+        color: white;
+      }
+      &.status-info {
+        background-color: var(--color-info-500);
+        color: white;
+      }
+      &.status-danger {
+        background-color: var(--color-danger-500);
+        color: white;
+      }
 
-    margin-left: 1em;
-    // height: 11px;
-    // margin-bottom: 10px;
+      &.topic {
+        background-color: var(--color-success-300);
+        color: var(--color-success-600);
+      }
+      &.suggested-topics {
+        background-color: var(--color-warning-300);
+        color: var(--color-warning-600);
+      }
 
-    // span {
-    //   background: var(--card-background-color);
-    //   padding: 0 5px;
-    //   border-radius: 0.5rem;
-    // }
+      &.confidence {
+        background-color: var(--color-default-300);
+        color: var(--text-basic-color);
+      }
+    }
   }
 }
diff --git a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.ts b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.ts
index 12a45039e6..b6adb66ada 100644
--- a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.ts
+++ b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.ts
@@ -14,10 +14,11 @@
  * limitations under the License.
  */
 
-import { Component, Input, OnInit } from '@angular/core';
+import { Component, Input } from '@angular/core';
 import { NbDialogService } from '@nebular/theme';
-import { Debug } from '../../../../model/dialog-data';
+import { Debug, RagAnswerStatus, RagAnswerStatusLabels } from '../../../../model/dialog-data';
 import { DebugViewerDialogComponent } from '../../../debug-viewer-dialog/debug-viewer-dialog.component';
+import { getContrastYIQ, getInterpolatedColor } from '../../../../utils';
 
 @Component({
   selector: 'tock-chat-ui-message-debug',
@@ -35,4 +36,42 @@ export class ChatUiMessageDebugComponent {
       }
     });
   }
+
+  getStatusClassName(): string {
+    const status = this.message.data.answer?.status;
+    if (!status) {
+      return '';
+    }
+
+    switch (status.toLowerCase()) {
+      case RagAnswerStatus.FOUND_IN_CONTEXT:
+        return 'status-success';
+      case RagAnswerStatus.NOT_FOUND_IN_CONTEXT:
+        return 'status-warning';
+      case RagAnswerStatus.SMALL_TALK:
+        return 'status-info';
+      case RagAnswerStatus.OUT_OF_SCOPE:
+        return 'status-danger';
+      default:
+        return '';
+    }
+  }
+
+  getStatusLabel(): string {
+    const status = this.message.data.answer?.status;
+    if (status) {
+      return RagAnswerStatusLabels[status.toLowerCase()] || status.replace(/_/g, ' ').replace(/^(.)|\s+(.)/g, (c) => c.toUpperCase());
+    }
+    return '';
+  }
+
+  getConfidenceBgColor(): { bg: string; fg: string } {
+    const score = this.message.data.answer?.confidence_score;
+    if (score != null) {
+      const bg = getInterpolatedColor(score);
+      const fg = getContrastYIQ(bg);
+      return { bg, fg };
+    }
+    return { bg: '', fg: '' };
+  }
 }
diff --git a/bot/admin/web/src/app/shared/model/dialog-data.ts b/bot/admin/web/src/app/shared/model/dialog-data.ts
index 428ebb1962..b8253f54a6 100644
--- a/bot/admin/web/src/app/shared/model/dialog-data.ts
+++ b/bot/admin/web/src/app/shared/model/dialog-data.ts
@@ -315,6 +315,20 @@ export class Debug extends BotMessage {
   }
 }
 
+export enum RagAnswerStatus {
+  FOUND_IN_CONTEXT = 'found_in_context',
+  NOT_FOUND_IN_CONTEXT = 'not_found_in_context',
+  SMALL_TALK = 'small_talk',
+  OUT_OF_SCOPE = 'out_of_scope'
+}
+
+export const RagAnswerStatusLabels: Record<RagAnswerStatus, string> = {
+  [RagAnswerStatus.FOUND_IN_CONTEXT]: 'Found in context',
+  [RagAnswerStatus.NOT_FOUND_IN_CONTEXT]: 'Not found in context',
+  [RagAnswerStatus.SMALL_TALK]: 'Small talk',
+  [RagAnswerStatus.OUT_OF_SCOPE]: 'Out of scope'
+};
+
 export class SentenceElement {
   constructor(
     public connectorType: ConnectorType,
diff --git a/bot/admin/web/src/app/shared/utils/utils.ts b/bot/admin/web/src/app/shared/utils/utils.ts
index 7a4164040b..d886dd6c5d 100644
--- a/bot/admin/web/src/app/shared/utils/utils.ts
+++ b/bot/admin/web/src/app/shared/utils/utils.ts
@@ -174,6 +174,42 @@ export function shadeColor(hexcolor: string, amount: number) {
   return '#' + RR + GG + BB;
 }
 
+/**
+ * Interpolates between two hex colors based on a value between 0 and 1.
+ *
+ * @param {number} value - A value between 0 and 1. Clamped to this range if outside.
+ * @param {string} [colorStart="#acbef4"] - The starting hex color (e.g., "#acbef4").
+ * @param {string} [colorEnd="#3366ff"] - The ending hex color (e.g., "#3366ff").
+ * @returns {string} The interpolated hex color as a string.
+ *
+ * @example
+ * // Returns a color halfway between #acbef4 and #3366ff
+ * const interpolatedColor = getInterpolatedColor(0.5, "#acbef4", "#3366ff");
+ */
+export function getInterpolatedColor(value: number, colorStart: string = '#acbef4', colorEnd: string = '#3366ff'): string {
+  // Clamp value between 0 and 1
+  const clampedValue = Math.min(1, Math.max(0, value));
+
+  // Parse hex colors to RGB
+  const parseHex = (hex: string) => {
+    const r = parseInt(hex.slice(1, 3), 16);
+    const g = parseInt(hex.slice(3, 5), 16);
+    const b = parseInt(hex.slice(5, 7), 16);
+    return { r, g, b };
+  };
+
+  const start = parseHex(colorStart);
+  const end = parseHex(colorEnd);
+
+  // Interpolate each channel
+  const r = Math.round(start.r + (end.r - start.r) * clampedValue);
+  const g = Math.round(start.g + (end.g - start.g) * clampedValue);
+  const b = Math.round(start.b + (end.b - start.b) * clampedValue);
+
+  // Convert back to hex
+  return `#${r.toString(16).padStart(2, '0')}${g.toString(16).padStart(2, '0')}${b.toString(16).padStart(2, '0')}`;
+}
+
 export async function copyToClipboard(text: string): Promise<void> {
   if (navigator.clipboard) {
     await navigator.clipboard.writeText(text);
diff --git a/bot/admin/web/src/app/theme/components/header/header.component.html b/bot/admin/web/src/app/theme/components/header/header.component.html
index b76bd7871c..4256d73059 100644
--- a/bot/admin/web/src/app/theme/components/header/header.component.html
+++ b/bot/admin/web/src/app/theme/components/header/header.component.html
@@ -103,6 +103,16 @@
       </nb-select>
     </nb-form-field>
 
+    <button
+      nbButton
+      ghost
+      size="medium"
+      (click)="goToDialogs()"
+      nbTooltip="Go to dialogs monitioring"
+    >
+      <nb-icon icon="wechat"></nb-icon>
+    </button>
+
     <button
       nbButton
       ghost
diff --git a/bot/admin/web/src/app/theme/components/header/header.component.ts b/bot/admin/web/src/app/theme/components/header/header.component.ts
index 840508fd2a..a3e8e5a085 100644
--- a/bot/admin/web/src/app/theme/components/header/header.component.ts
+++ b/bot/admin/web/src/app/theme/components/header/header.component.ts
@@ -127,6 +127,10 @@ export class HeaderComponent implements OnInit, OnDestroy {
     this.menuService.navigateHome();
   }
 
+  goToDialogs() {
+    this.router.navigateByUrl('/analytics/dialogs');
+  }
+
   changeApplication(app) {
     setTimeout((_) => {
       this.state.changeApplicationWithName(app);
diff --git a/bot/admin/web/src/app/theme/styles/utilities.scss b/bot/admin/web/src/app/theme/styles/utilities.scss
index 70840d1b66..da42d7c939 100644
--- a/bot/admin/web/src/app/theme/styles/utilities.scss
+++ b/bot/admin/web/src/app/theme/styles/utilities.scss
@@ -32,6 +32,10 @@
   }
 }
 
+.gap-05 {
+  gap: 0.5rem;
+}
+
 // Should disappear with bootstrap update to v5.
 .column-gap-1 {
   column-gap: 0.25rem;

From bf8ccb0ce64f27e9dd349dd029d5ce64f8ac7b8e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rodolphe=20K=C3=BCffer?= <rodolphe.kuffer@partnre.com>
Date: Thu, 5 Feb 2026 17:11:46 +0100
Subject: [PATCH 7/7] Front - Fix globalMessage footer display + Improve Rag
 debug display

---
 .../chat-ui-message-debug.component.html      |  2 +-
 .../debug-viewer-dialog.component.html        |  5 +-
 .../json-iterator.component.html              | 58 +++++++++++--------
 .../json-iterator.component.scss              | 33 ++++++-----
 .../json-iterator/json-iterator.component.ts  | 53 +++++++++++++++--
 .../components/footer/footer.component.html   | 47 +++++++++------
 .../components/footer/footer.component.ts     | 18 +++++-
 7 files changed, 153 insertions(+), 63 deletions(-)

diff --git a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
index 09354d8922..865f88d462 100644
--- a/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
+++ b/bot/admin/web/src/app/shared/components/chat-ui/chat-ui-message/chat-ui-message-debug/chat-ui-message-debug.component.html
@@ -14,7 +14,7 @@
   ~ limitations under the License.
   -->
 
-<div class="debug d-flex justify-content-end pr-5 gap-05">
+<div class="debug d-flex flex-wrap justify-content-end pr-5 gap-05">
   <div
     class="tag"
     [ngClass]="getStatusClassName()"
diff --git a/bot/admin/web/src/app/shared/components/debug-viewer-dialog/debug-viewer-dialog.component.html b/bot/admin/web/src/app/shared/components/debug-viewer-dialog/debug-viewer-dialog.component.html
index 0f933db352..72d0d3f817 100644
--- a/bot/admin/web/src/app/shared/components/debug-viewer-dialog/debug-viewer-dialog.component.html
+++ b/bot/admin/web/src/app/shared/components/debug-viewer-dialog/debug-viewer-dialog.component.html
@@ -29,6 +29,9 @@
   </nb-card-header>
 
   <nb-card-body>
-    <tock-json-iterator [data]="debug"></tock-json-iterator>
+    <tock-json-iterator
+      [data]="debug"
+      [customOrder]="['user_question', 'condensed_question', 'answer', 'question_answering_prompt', 'documents']"
+    ></tock-json-iterator>
   </nb-card-body>
 </nb-card>
diff --git a/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.html b/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.html
index f06b3718a4..ad6f8bb052 100644
--- a/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.html
+++ b/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.html
@@ -33,31 +33,43 @@
   *ngIf="parentKey"
   (click)="switchDeployed()"
   class="parent-title"
-  [ngClass]="{ isPrimitive: isPrimitive(data), ellipsis: !isPrimitive(data), pointer: !isPrimitive(data) }"
+  [ngClass]="{ pointer: isExpandable() }"
 >
-  <nb-icon
-    icon="caret-right-fill"
-    class="align-middle"
-    *ngIf="!isDeployed && !isPrimitive(data)"
-  ></nb-icon>
-  <nb-icon
-    icon="caret-down-fill"
-    class="align-middle"
-    *ngIf="isDeployed && !isPrimitive(data)"
-  ></nb-icon>
-  <strong> {{ parentKey }} : </strong>
+  <div class="d-flex">
+    <div class="caret-holder">
+      <nb-icon
+        icon="caret-right-fill"
+        class="align-middle"
+        *ngIf="!isDeployed && isExpandable()"
+      ></nb-icon>
+      <nb-icon
+        icon="caret-down-fill"
+        class="align-middle"
+        *ngIf="isDeployed && isExpandable()"
+      ></nb-icon>
+    </div>
+    <div class="grow-1 ellipsis">
+      <strong> {{ parentKey }} : </strong>
 
-  <pre
-    class="pre-format text-primary pl-2"
-    *ngIf="isPrimitive(data)"
-  >
+      <span
+        class="text-primary pl-2"
+        *ngIf="isPrimitive(data) && !isDeployed"
+        >"{{ data }}"</span
+      >
+      <pre
+        class="pre-format text-primary pl-2"
+        *ngIf="isPrimitive(data) && isDeployed"
+      >
 "{{ data }}"</pre
-  >
-  <span
-    class="text-muted pl-2"
-    *ngIf="!isPrimitive(data)"
-    >{{ data | json }}</span
-  >
+      >
+
+      <span
+        class="text-muted pl-2"
+        *ngIf="!isPrimitive(data)"
+        >{{ data | json }}</span
+      >
+    </div>
+  </div>
 </div>
 
 <div
@@ -65,7 +77,7 @@
   class="entry-block"
   [class.isRoot]="isRoot"
 >
-  <ng-container *ngFor="let list of data | keyvalue">
+  <ng-container *ngFor="let list of data | keyvalue: customSort">
     <tock-json-iterator
       [data]="list.value"
       [parentKey]="list.key"
diff --git a/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.scss b/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.scss
index 3e0bec0785..ee002ea4aa 100644
--- a/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.scss
+++ b/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.scss
@@ -14,23 +14,28 @@
  * limitations under the License.
  */
 
-.parent-title {
-  padding: 0.3rem;
+:host {
+  display: block;
 
-  &.isPrimitive {
-    padding-left: 1.8rem;
+  .parent-title {
+    padding: 0.3rem;
   }
-}
 
-.entry-block {
-  padding-left: 1.1rem;
-  &.isRoot {
-    padding-left: 0;
-    margin-left: -0.5rem;
+  .entry-block {
+    padding-left: 1.1rem;
+    &.isRoot {
+      padding-left: 0;
+      margin-left: -0.5rem;
+    }
   }
-}
 
-.pre-format {
-  white-space: pre-wrap;
-  display: inline;
+  .pre-format {
+    white-space: pre-wrap;
+    display: inline;
+  }
+
+  .caret-holder {
+    min-width: 20px;
+    margin-right: 3px;
+  }
 }
diff --git a/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.ts b/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.ts
index d3024f6fc6..0929959ec2 100644
--- a/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.ts
+++ b/bot/admin/web/src/app/shared/components/json-iterator/json-iterator.component.ts
@@ -14,7 +14,7 @@
  * limitations under the License.
  */
 
-import { Component, Input, OnDestroy } from '@angular/core';
+import { AfterViewInit, ChangeDetectionStrategy, ChangeDetectorRef, Component, ElementRef, Input, OnDestroy } from '@angular/core';
 import { isPrimitive } from '../../utils';
 import { JsonIteratorService } from './json-iterator.service';
 import { Subject, takeUntil } from 'rxjs';
@@ -22,20 +22,25 @@ import { Subject, takeUntil } from 'rxjs';
 @Component({
   selector: 'tock-json-iterator',
   templateUrl: './json-iterator.component.html',
-  styleUrls: ['./json-iterator.component.scss']
+  styleUrls: ['./json-iterator.component.scss'],
+  changeDetection: ChangeDetectionStrategy.OnPush
 })
-export class JsonIteratorComponent implements OnDestroy {
+export class JsonIteratorComponent implements AfterViewInit, OnDestroy {
   destroy = new Subject();
 
   @Input() data: { [key: string]: any };
   @Input() isRoot: boolean = true;
   @Input() parentKey: string;
+  @Input() customOrder: string[] = [];
+
+  width = 0;
+  private resizeObserver!: ResizeObserver;
 
   isDeployed: boolean = false;
 
   isPrimitive = isPrimitive;
 
-  constructor(private jsonIteratorService: JsonIteratorService) {
+  constructor(private jsonIteratorService: JsonIteratorService, private elementRef: ElementRef, private cdr: ChangeDetectorRef) {
     this.jsonIteratorService.communication.pipe(takeUntil(this.destroy)).subscribe((evt) => {
       if (evt.type === 'expand' && !this.isDeployed) {
         this.isDeployed = true;
@@ -46,15 +51,55 @@ export class JsonIteratorComponent implements OnDestroy {
     });
   }
 
+  ngAfterViewInit() {
+    this.width = this.elementRef.nativeElement.offsetWidth;
+    this.resizeObserver = new ResizeObserver((entries) => {
+      for (const entry of entries) {
+        this.width = Math.floor(entry.contentRect.width);
+        this.cdr.detectChanges();
+      }
+    });
+    this.resizeObserver.observe(this.elementRef.nativeElement);
+    this.cdr.detectChanges();
+  }
+
+  customSort = (a: { key: string }, b: { key: string }) => {
+    const order = this.customOrder;
+    const aIndex = order.indexOf(a.key);
+    const bIndex = order.indexOf(b.key);
+
+    if (aIndex !== -1 && bIndex !== -1) {
+      return aIndex - bIndex;
+    }
+
+    if (aIndex !== -1) return -1;
+    if (bIndex !== -1) return 1;
+
+    return a.key.localeCompare(b.key);
+  };
+
+  isExpandable(): boolean {
+    if (!this.isPrimitive(this.data)) return true;
+
+    const fontSize = 12;
+    const maxChars = Math.floor(this.width / (0.6 * fontSize));
+    const fullTextLength = String(this.parentKey).length + String(this.data).length + 6; // +6 for parent name, space, colon, space and quotes around the value
+
+    return fullTextLength >= maxChars;
+  }
+
   switchDeployed() {
     this.isDeployed = !this.isDeployed;
+    this.cdr.detectChanges();
   }
 
   expandAll() {
     this.jsonIteratorService.expandAll();
+    this.cdr.detectChanges();
   }
 
   ngOnDestroy(): void {
+    this.resizeObserver.disconnect();
     this.destroy.next(true);
     this.destroy.complete();
   }
diff --git a/bot/admin/web/src/app/theme/components/footer/footer.component.html b/bot/admin/web/src/app/theme/components/footer/footer.component.html
index 3ab9c982cc..5454d1ed41 100644
--- a/bot/admin/web/src/app/theme/components/footer/footer.component.html
+++ b/bot/admin/web/src/app/theme/components/footer/footer.component.html
@@ -15,29 +15,42 @@
   -->
 
 <div>
-  <div class="message" *ngIf="configuration && (configuration | async).globalMessage">
-    {{ (configuration | async).globalMessage}}
+  <div
+    class="message"
+    *ngIf="globalMessage"
+  >
+    {{ globalMessage }}
   </div>
+
   <div class="info">
     Tock v{{ tock_info.build_version }}
-      <a
-        href="https://doc.tock.ai"
-        target="_blank"
-        title="Tock Homepage"
+    <a
+      href="https://doc.tock.ai"
+      target="_blank"
+      title="Tock Homepage"
       ><nb-icon
         icon="info-circle"
         nbTooltip="Tock documentation '{{ tock_info.platform_type }}' version {{ tock_info.build_version }}
              (built from {{ tock_info.source_commit_id }} the {{ tock_info.build_date }})"
-      ></nb-icon></a>
-      &nbsp;
-      <a
-        href="https://github.com/theopenconversationkit/tock"
-        target="_blank"
-      ><nb-icon icon="github" nbTooltip="Tock Sources on GitHub"></nb-icon></a>
-      &nbsp;
-      <a
-        href="https://gitter.im/tockchat/Lobby"
-        target="_blank"
-      ><nb-icon icon="chat-left-text"  nbTooltip="Tock Chat on Gitter"></nb-icon></a>
+      ></nb-icon
+    ></a>
+    &nbsp;
+    <a
+      href="https://github.com/theopenconversationkit/tock"
+      target="_blank"
+      ><nb-icon
+        icon="github"
+        nbTooltip="Tock Sources on GitHub"
+      ></nb-icon
+    ></a>
+    &nbsp;
+    <a
+      href="https://gitter.im/tockchat/Lobby"
+      target="_blank"
+      ><nb-icon
+        icon="chat-left-text"
+        nbTooltip="Tock Chat on Gitter"
+      ></nb-icon
+    ></a>
   </div>
 </div>
diff --git a/bot/admin/web/src/app/theme/components/footer/footer.component.ts b/bot/admin/web/src/app/theme/components/footer/footer.component.ts
index 18fb558f55..9862c5605c 100644
--- a/bot/admin/web/src/app/theme/components/footer/footer.component.ts
+++ b/bot/admin/web/src/app/theme/components/footer/footer.component.ts
@@ -14,20 +14,32 @@
  * limitations under the License.
  */
 
-import { Component } from '@angular/core';
+import { Component, OnInit } from '@angular/core';
 import { tock_info } from '../../../../environments/manifest';
 import { BotSharedService } from '../../../shared/bot-shared.service';
 import { AdminConfiguration } from '../../../shared/model/conf';
+import { take } from 'rxjs';
 
 @Component({
   selector: 'tock-footer',
   templateUrl: './footer.component.html',
   styleUrls: ['./footer.component.scss']
 })
-export class FooterComponent {
+export class FooterComponent implements OnInit {
   constructor(private botSharedService: BotSharedService) {}
 
   tock_info = tock_info;
 
-  configuration = this.botSharedService.getConfiguration()
+  globalMessage = '';
+
+  ngOnInit() {
+    this.botSharedService
+      .getConfiguration()
+      .pipe(take(1))
+      .subscribe((conf: AdminConfiguration) => {
+        if (conf.globalMessage) {
+          this.globalMessage = conf.globalMessage;
+        }
+      });
+  }
 }