Modify prompts of entry and validator

[madina1203]

PR Type

enhancement

---

Description

• Enhanced entry agent prompt to clarify role and task.

• Improved validator prompt for better question validation.

• Added guidelines for handling unclear or complex questions.

• Corrected minor typographical errors in prompts.

---

Changes walkthrough 📝

Relevant files

Enhancement

prompt.py Refine entry agent prompt for task clarity                              app/core/agents/entry/prompt.py Clarified entry agent's role in handling queries. Added instructions for classifying and forwarding queries. Corrected typographical error in the prompt. Emphasized non-answering role of the entry agent. +3/-2      prompt.py Improve validator prompt for robust question validation    app/core/agents/validator/prompt.py Enhanced criteria for question validity. Added flexibility for phrasing in validation. Provided guidelines for handling unclear questions. Updated feedback mechanism for invalid questions. +5/-3

---

Need help?

Type /help how to ... in the comments thread for any questions about PR-Agent usage.

Check out the documentation for more information.

[qodo-code-review]

PR Reviewer Guide 🔍

(Review updated until commit 9e493a0)

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🔒 No security concerns identified

⚡ Recommended focus areas for review Instruction Clarity Confirm that the newly added instruction about not answering questions directly integrates seamlessly with the existing responsibilities and does not lead to any ambiguity in agent behavior. Do not answer the question yourself. Your role is to classify and pass along the request, not to generate final answers. Other team members will handle other queries. Validation Consistency Review the added validation criteria and flexible phrasing to ensure they are consistent with current schema checks and do not conflict with existing logic. -The question is valid if it references any nodes, classes, or properties in the ENPKG — even if indirectly or across multiple relationships. -Allow flexibility for phrasing. As long as the topic aligns with schema elements, the question should be considered valid. Guidelines for Validity: Plant-Specific and Feature-Related Queries: Recognize questions as valid if they involve plant names, metabolites, or features, as long as properties (e.g., has_LCMS, submitted_taxon, has_lab_process, or has_bioassay_results, etc) are mentioned or implied. For instance, questions referencing specific plants, compounds, or taxonomic details should be valid if these attributes can be queried in ENPKG. Grouping, Counting, and Comparing Annotations: Allow questions that require counting, grouping, or comparing annotations if the question mentions properties available in the ENPKG schema (e.g., has_wd_id, has_InChIkey2D, or has_canopus_annotation). These types of questions are valid if they involve any schema properties or identifiers. Valid Question Criteria: -The question is related to one or more of the ENPKG nodes/entities listed below. As long as the entities in the question exist in the schema, even if the question is complex or spans multiple relationships, it is still valid. Not Valid Question Criteria: -It is completely unrelated to any nodes/entities in ENPKG or the topic cannot be mapped to any known classes or properties in the schema. If a question is unclear or possibly valid but lacks detail, do not mark it invalid — ask the user for clarification instead. Google DocString Format While no new functions or classes were added, ensure that any key functions or modules in future updates include Google-style docstrings as per project guidelines.

[github-actions]

Persistent review updated to latest commit 9e493a0

[qodo-code-review]

PR Code Suggestions ✨

Latest suggestions up to 9e493a0
Explore these optional code suggestions:

Category	Suggestion	Impact
General	Clarify deferral instructions Clarify the agent’s role by specifying conditions for deferring response generation, ensuring the intended flow is unambiguous. app/core/agents/entry/prompt.py [5] -You do not directly answer knowledge-based user queries. Your task is to classify them, and forward it with proper context. Other agents will handle the actual response. +You do not directly answer knowledge-based user queries; instead, classify them and forward with necessary context to the appropriate agent for response generation. Suggestion importance[1-10]: 5 __ Why: The suggestion slightly rewords the deferral instruction for clarity without altering functionality, offering only a minor improvement.	Low
	Refine validation criteria Refine the validation criteria to emphasize clear association with known schema elements while still allowing flexible phrasing. app/core/agents/validator/prompt.py [10-11] -The question is valid if it references any nodes, classes, or properties in the ENPKG — even if indirectly or across multiple relationships. -Allow flexibility for phrasing. As long as the topic aligns with schema elements, the question should be considered valid. +The question is valid if it clearly references nodes, classes, or properties in the ENPKG—even indirectly—provided the reference unambiguously aligns with known schema elements. Allow flexible phrasing as long as the core topic relates to these elements. Suggestion importance[1-10]: 5 __ Why: The change refines the wording for validation criteria to enhance clarity; however, it remains a stylistic update with limited impact.	Low
	Standardize unclear query handling Consolidate the instructions for unclear queries to ensure consistency in prompting for clarification where needed. app/core/agents/validator/prompt.py [19-21] -It is completely unrelated to any nodes/entities in ENPKG or the topic cannot be mapped to any known classes or properties in the schema. -If a question is unclear or possibly valid but lacks detail, do not mark it invalid — ask the user for clarification instead. +It is completely unrelated to any nodes/entities in ENPKG or cannot be mapped to any known classes or properties in the schema. For queries that are unclear or borderline in detail, always ask the user for clarification rather than marking them as invalid. Suggestion importance[1-10]: 5 __ Why: This suggestion consolidates instructions for handling unclear queries, improving consistency. The improvement is useful but only a minor refinement.	Low

---

Previous suggestions

Suggestions up to commit 9e493a0

Category	Suggestion	Impact
General	Refine unclear query handling Refine the instruction for borderline cases to ensure that ambiguous queries trigger a request for more detail without being incorrectly invalidated. app/core/agents/validator/prompt.py [21] -If a question is unclear or possibly valid but lacks detail, do not mark it invalid — ask the user for clarification instead. +If a question is unclear or insufficiently detailed, do not mark it as invalid; instead, request further clarification. Apply this suggestion Suggestion importance[1-10]: 5 __ Why: The suggestion rephrases the instruction for borderline cases to improve clarity without changing the intended behavior, representing a moderate improvement.	Low
More