Cuga compatiable

.secrets.baseline

@@ -3,7 +3,7 @@
     "files": "^.secrets.baseline$",
     "lines": null
   },
-  "generated_at": "2026-02-24T16:20:06Z",
+  "generated_at": "2026-02-26T16:00:29Z",
   "plugins_used": [
     {
       "name": "AWSKeyDetector"
@@ -87,7 +87,7 @@
         "verified_result": null
       },
       {
-        "hashed_secret": "1ed5b7962b3d8356ccb9f4ccbbb0f17e5fc39724",
+        "hashed_secret": "11fa7c37d697f30e6aee828b4426a10f83ab2380",
         "is_secret": false,
         "is_verified": false,
         "line_number": 18,

CONFIGURATION.md

@@ -11,19 +11,28 @@ export OPENAI_API_KEY=sk-...
 
 ### Custom LLM Configuration
 
-Kaizen uses [LiteLLM](https://docs.litellm.ai/) and supports using a LiteLLM proxy server for centralized LLM access:
+Kaizen uses [LiteLLM](https://docs.litellm.ai/) and supports OpenAI-compatible proxy endpoints (including LiteLLM) via standard OpenAI environment variables:
 
 ```bash
-# LiteLLM Proxy Configuration
-LITELLM_PROXY_API_KEY="your-proxy-token"
-LITELLM_PROXY_API_BASE="https://your-litellm-proxy.com"
+# OpenAI-compatible endpoint configuration (works with LiteLLM)
+export OPENAI_API_KEY="your-api-key"
+export OPENAI_BASE_URL="https://your-litellm-proxy.com/v1"
 
 # Kaizen Model Configuration
-KAIZEN_TIPS_MODEL="your-model-name"
-KAIZEN_CONFLICT_RESOLUTION_MODEL="your-model-name"
-KAIZEN_CUSTOM_LLM_PROVIDER="your-custom-llm-provider"
+export KAIZEN_TIPS_MODEL="openai/gpt-4o-mini"
+export KAIZEN_CONFLICT_RESOLUTION_MODEL="openai/gpt-4o-mini"
+export KAIZEN_FACT_EXTRACTION_MODEL="openai/gpt-4o-mini"
+export KAIZEN_MODEL_NAME="openai/gpt-4o-mini"
+export KAIZEN_CUSTOM_LLM_PROVIDER="openai"
 ```
 
+Model selection precedence:
+1. Task-specific models: `KAIZEN_TIPS_MODEL`, `KAIZEN_CONFLICT_RESOLUTION_MODEL`, `KAIZEN_FACT_EXTRACTION_MODEL`
+2. Global Kaizen fallback: `KAIZEN_MODEL_NAME`
+3. Built-in default: `gpt-4o`
+
+If `KAIZEN_*_MODEL` are unset, set `KAIZEN_MODEL_NAME` to control all Kaizen LLM calls.
+
 ## Environment Variables
 
 All configuration variables are prefixed with `KAIZEN_`.
@@ -34,9 +43,11 @@ All configuration variables are prefixed with `KAIZEN_`.
 |----------|-------------------------------------------------------------------------------|------------------------------------------|
 | `KAIZEN_BACKEND` | Backend provider (`milvus` or `filesystem`)                                   | `milvus`                                 |
 | `KAIZEN_NAMESPACE_ID` | Namespace ID for isolation                                                    | `kaizen`                                 |
-| `KAIZEN_TIPS_MODEL` | Model for generating tips (e.g. `openai/gpt-4o` for proxy with custom models) | `gpt-4o`                                 |
-| `KAIZEN_CONFLICT_RESOLUTION_MODEL` | Model for resolving conflicts (e.g. `openai/gpt-4o` for proxy with custom models)  | `gpt-4o`                                 |
-| `KAIZEN_CUSTOM_LLM_PROVIDER` | LiteLLM provider (use `openai` for proxy with custom models) | `None`                                   |
+| `KAIZEN_TIPS_MODEL` | Model for tip generation only | `KAIZEN_MODEL_NAME` -> `gpt-4o` |
+| `KAIZEN_CONFLICT_RESOLUTION_MODEL` | Model for conflict resolution only | `KAIZEN_MODEL_NAME` -> `gpt-4o` |
+| `KAIZEN_FACT_EXTRACTION_MODEL` | Model for fact extraction only | `KAIZEN_MODEL_NAME` -> `gpt-4o` |
+| `KAIZEN_MODEL_NAME` | Global fallback model for all Kaizen LLM calls | `gpt-4o` |
+| `KAIZEN_CUSTOM_LLM_PROVIDER` | LiteLLM provider (use `openai` for OpenAI-compatible endpoints) | `None`                                   |
 | `KAIZEN_EMBEDDING_MODEL` | Embedding model                                                               | `sentence-transformers/all-MiniLM-L6-v2` |
 
 ### Milvus Backend Settings
@@ -146,4 +157,3 @@ except ImportError:
 | `KAIZEN_TRACING_ENDPOINT` | Phoenix collector endpoint | `http://localhost:6006/v1/traces` |
 
 > **Note**: Auto-patching skips if existing tracing is detected. Use `enable_tracing(force=True)` to override.
-

README.md

@@ -31,12 +31,12 @@ uv sync && source .venv/bin/activate
 
 ### Configuration
 
-Set your OpenAI API key:
+For direct OpenAI usage:
 ```bash
 export OPENAI_API_KEY=sk-...
 ```
 
-For detailed configuration options (custom LLM providers, backends, etc.), see [CONFIGURATION.md](CONFIGURATION.md).
+For LiteLLM proxy usage and model selection (including global fallback via `KAIZEN_MODEL_NAME`), see [CONFIGURATION.md](CONFIGURATION.md).
 
 ### Running the MCP Server
 

kaizen/backend/filesystem.py

@@ -241,10 +241,14 @@ def _search_entities_internal(
             for ent in entities:
                 match = True
                 for key, value in filters.items():
-                    # Check top-level field first, then metadata
-                    ent_value = ent.get(key)
-                    if ent_value is None and ent.get("metadata"):
-                        ent_value = ent["metadata"].get(key)
+                    if key.startswith("metadata."):
+                        metadata_key = key.split(".", 1)[1]
+                        ent_value = (ent.get("metadata") or {}).get(metadata_key)
+                    else:
+                        # Check top-level field first, then metadata
+                        ent_value = ent.get(key)
+                        if ent_value is None and ent.get("metadata"):
+                            ent_value = ent["metadata"].get(key)
                     if ent_value != value:
                         match = False
                         break