feat: add Kimi Code connector and backend startup disablement

- Add Kimi Code connector implementation with OpenAI-compatible API
- Add backend startup disablement service for selective backend loading
- Update Gemini base connector and model validation
- Update backend configuration provider with environment-based config
- Update application builder to support backend disablement
- Update models controller for modality support
- Add comprehensive unit tests for new components
- Add Kimi Code backend documentation
- Update .gitignore to exclude var/cache/ directory

Author: Mateusz

.gitignore

@@ -125,4 +125,7 @@ var/gemini_oauth_accounts/
 var/kiro_oauth_accounts/
 
 # LevelDB test data (may contain sensitive data)
-leveldb_test/
+leveldb_test/
+
+# Cache directory
+var/cache/

docs/user_guide/backends/kimi-code.md

@@ -0,0 +1,80 @@
+# Kimi Code Backend
+
+The `kimi-code` backend provides an OpenAI-compatible connector for Kimi's coding gateway.
+
+It is implemented as a subclass of the OpenAI-compatible backend connector and targets:
+
+- Base URL: `https://api.kimi.com/coding/v1`
+
+This backend is intended to be used via the [OpenAI Chat Completions frontend](../frontends/openai-chat-completions.md).
+
+## Configuration
+
+### Environment Variables
+
+Set the API key:
+
+```bash
+export KIMI_API_KEY="..."
+```
+
+## Model Naming
+
+The proxy exposes a single model through this backend:
+
+- `kimi-for-coding`
+
+When calling the OpenAI Chat Completions frontend, use the fully-qualified model string:
+
+- `kimi-code:kimi/kimi-for-coding`
+
+Example:
+
+```bash
+curl -X POST http://localhost:8000/v1/chat/completions \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "kimi-code:kimi/kimi-for-coding",
+    "messages": [{"role": "user", "content": "Write a Python function that prints Hello World."}],
+    "stream": true
+  }'
+```
+
+## Multimodal (Text + Image)
+
+This backend advertises the model as accepting:
+
+- Input modalities: `text`, `image`
+- Output modalities: `text`
+
+Example (OpenAI-compatible `image_url` message parts):
+
+```bash
+curl -X POST http://localhost:8000/v1/chat/completions \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "kimi-code:kimi/kimi-for-coding",
+    "messages": [{
+      "role": "user",
+      "content": [
+        {"type": "text", "text": "Describe what you see and suggest refactors."},
+        {"type": "image_url", "image_url": {"url": "https://example.com/screenshot.png"}}
+      ]
+    }],
+    "stream": true
+  }'
+```
+
+## Reasoning Output Compatibility
+
+Some OpenAI-compatible providers stream text using `reasoning_content` while leaving `content` empty.
+Many clients only render `content`.
+
+The `kimi-code` connector mirrors reasoning text into `content` when needed, while keeping the original
+reasoning fields intact. This makes the backend usable with clients that do not understand
+`reasoning_content`.
+
+## Related Documentation
+
+- [Backend Overview](overview.md)
+- [OpenAI Chat Completions Frontend](../frontends/openai-chat-completions.md)

docs/user_guide/backends/overview.md

@@ -17,15 +17,16 @@ The proxy supports the following backend providers out of the box:
 | `gemini-oauth-plan` | Google Gemini (CLI) | OAuth | Users with Google One subscription |
 | `gemini-oauth-free` | Google Gemini (CLI) | OAuth | Free tier users |
 | `gemini-cli-cloud-project` | Google Gemini (GCP) | OAuth + GCP Project | Enterprise, team workflows, central billing |
-| `openrouter` | OpenRouter | API Key | Access to many hosted models |
-| `zenmux` | ZenMux | API Key | OpenAI-compatible ZenMux router |
-| `zai` | ZAI | API Key | Zhipu/Z.ai access |
-| `zai-coding-plan` | ZAI Coding Plan | API Key | Coding-specific workflows |
-| `minimax` | Minimax | API Key | Minimax AI models |
-| `qwen-oauth` | Alibaba Qwen | Local OAuth token | Qwen CLI OAuth |
-| `qwen-oauth` | Alibaba Qwen | Local OAuth token | Qwen CLI OAuth |
-| `hybrid` | Virtual (orchestrates two models) | Inherits from sub-backends | Two-phase reasoning + execution |
-| `antigravity-oauth` | Google Gemini (Antigravity) | Antigravity Token | Internal debugging (Gemini models) |
+| `openrouter` | OpenRouter | API Key | Access to many hosted models |
+| `zenmux` | ZenMux | API Key | OpenAI-compatible ZenMux router |
+| `zai` | ZAI | API Key | Zhipu/Z.ai access |
+| `zai-coding-plan` | ZAI Coding Plan | API Key | Coding-specific workflows |
+| `kimi-code` | Kimi | API Key | Kimi For Coding (OpenAI-compatible) |
+| `minimax` | Minimax | API Key | Minimax AI models |
+| `qwen-oauth` | Alibaba Qwen | Local OAuth token | Qwen CLI OAuth |
+| `qwen-oauth` | Alibaba Qwen | Local OAuth token | Qwen CLI OAuth |
+| `hybrid` | Virtual (orchestrates two models) | Inherits from sub-backends | Two-phase reasoning + execution |
+| `antigravity-oauth` | Google Gemini (Antigravity) | Antigravity Token | Internal debugging (Gemini models) |
 
 ## Frontend APIs
 
@@ -59,10 +60,11 @@ Backends are configured through environment variables and the proxy configuratio
 export OPENAI_API_KEY="sk-..."
 export ANTHROPIC_API_KEY="sk-ant-..."
 export GEMINI_API_KEY="AIza..."
-export OPENROUTER_API_KEY="sk-or-..."
-export ZENMUX_API_KEY="..."
-export ZAI_API_KEY="..."
-export MINIMAX_API_KEY="..."
+export OPENROUTER_API_KEY="sk-or-..."
+export ZENMUX_API_KEY="..."
+export ZAI_API_KEY="..."
+export KIMI_API_KEY="..."
+export MINIMAX_API_KEY="..."
 
 # For GCP-based Gemini
 export GOOGLE_CLOUD_PROJECT="your-project-id"
@@ -120,8 +122,9 @@ For detailed configuration and usage information for each backend, see:
 - [ZAI Backend](zai.md)
 - [Qwen Backend](qwen.md)
 - [MiniMax Backend](minimax.md)
-- [ZenMux Backend](zenmux.md)
-- [Custom Backends](custom-backends.md)
+- [ZenMux Backend](zenmux.md)
+- [Kimi Code Backend](kimi-code.md)
+- [Custom Backends](custom-backends.md)
 
 ## Related Features
 

docs/user_guide/index.md

@@ -110,13 +110,15 @@ Backend provider configuration and usage:
 - **[Antigravity OAuth Backend](backends/antigravity-oauth.md)** - Internal Antigravity OAuth configuration
 - **[Kiro OAuth Auto Backend](backends/kiro-oauth-auto.md)** - Amazon Kiro / Q Developer streaming via self-managed OAuth
 
+- **[Kimi Code Backend](backends/kimi-code.md)** - Kimi For Coding via OpenAI-compatible API
+
 - **[OpenRouter Backend](backends/openrouter.md)** - OpenRouter multi-model access
 - **[ZAI Backend](backends/zai.md)** - Zhipu/Z.ai configuration
 - **[Qwen Backend](backends/qwen.md)** - Alibaba Qwen OAuth configuration
 - **[Minimax Backend](backends/minimax.md)** - Minimax API configuration
 - **[Zenmux Backend](backends/zenmux.md)** - Zenmux API configuration
 - **[OpenCode Zen Backend](backends/opencode-zen.md)** - OpenCode Zen API configuration
-- **[Custom Backends](backends/custom-backends.md)** - Creating and configuring custom backend connectors
+- **[Custom Backends](backends/custom-backends.md)** - Creating and configuring custom backend connectors
 
 ## Debugging
 

src/connectors/gemini_base/config.py

@@ -29,22 +29,13 @@
 
 # Default available models for fallback
 DEFAULT_AVAILABLE_MODELS = [
-    # Current generation (2.5 series) - DEFAULT models
-    "gemini-2.5-pro",
+    # Current generation (3.x series)
+    "gemini-3-pro-preview",
     "gemini-3-flash-preview",
+    # 2.5 series
+    "gemini-2.5-pro",
     "gemini-2.5-flash",
     "gemini-2.5-flash-lite",
-    # Preview models
-    "gemini-2.5-pro-preview-05-06",
-    "gemini-2.5-pro-preview-06-05",
-    "gemini-2.5-flash-preview-05-20",
-    # 2.0 series
-    "gemini-2.0-flash",
-    "gemini-2.0-flash-thinking-exp-1219",
-    "gemini-2.0-flash-preview-image-generation",
-    # 1.5 series
-    "gemini-1.5-pro",
-    "gemini-1.5-flash",
     # Embedding model
     "gemini-embedding-001",
 ]

src/connectors/gemini_base/connector.py

@@ -1166,21 +1166,13 @@ async def _ensure_models_loaded(self) -> None:
             if not self.available_models:
                 # Use a hardcoded list based on gemini-cli's tokenLimits.ts and models.ts
                 self.available_models = [
-                    # Current generation (2.5 series) - DEFAULT models
+                    # Current generation (3.x series)
+                    "gemini-3-pro-preview",
+                    "gemini-3-flash-preview",
+                    # 2.5 series
                     "gemini-2.5-pro",
                     "gemini-2.5-flash",
                     "gemini-2.5-flash-lite",
-                    # Preview models
-                    "gemini-2.5-pro-preview-05-06",
-                    "gemini-2.5-pro-preview-06-05",
-                    "gemini-2.5-flash-preview-05-20",
-                    # 2.0 series
-                    "gemini-2.0-flash",
-                    "gemini-2.0-flash-thinking-exp-1219",
-                    "gemini-2.0-flash-preview-image-generation",
-                    # 1.5 series
-                    "gemini-1.5-pro",
-                    "gemini-1.5-flash",
                     # Embedding model
                     "gemini-embedding-001",
                 ]

src/connectors/gemini_base/model_validation.py

@@ -195,21 +195,13 @@ class ModelListManager:
 
     # Default fallback model list
     DEFAULT_MODELS: list[str] = [
-        # Current generation (2.5 series) - DEFAULT models
+        # Current generation (3.x series)
+        "gemini-3-pro-preview",
+        "gemini-3-flash-preview",
+        # 2.5 series
         "gemini-2.5-pro",
         "gemini-2.5-flash",
         "gemini-2.5-flash-lite",
-        # Preview models
-        "gemini-2.5-pro-preview-05-06",
-        "gemini-2.5-pro-preview-06-05",
-        "gemini-2.5-flash-preview-05-20",
-        # 2.0 series
-        "gemini-2.0-flash",
-        "gemini-2.0-flash-thinking-exp-1219",
-        "gemini-2.0-flash-preview-image-generation",
-        # 1.5 series
-        "gemini-1.5-pro",
-        "gemini-1.5-flash",
         # Embedding model
         "gemini-embedding-001",
     ]