Merge pull request #5180 from BerriAI/litellm_allow_controlling_logged_tags_langfuse

[Feat-Proxy+langfuse] LiteLLM-specific Tags on Langfuse - `cache_hit`, `cache_key`

docs/my-website/docs/proxy/logging.md

@@ -278,6 +278,42 @@ curl --location 'http://0.0.0.0:4000/chat/completions' \
 }'
 ```
 
+
+### LiteLLM-specific Tags on Langfuse - `cache_hit`, `cache_key`
+
+Use this if you want to control which LiteLLM-specific fields are logged as tags by the LiteLLM proxy. By default LiteLLM Proxy logs no LiteLLM-specific fields
+
+| LiteLLM specific field               | Description                                           | Example Value                                       |
+|------------------------|-------------------------------------------------------|------------------------------------------------|
+| `cache_hit`            | Indicates whether a cache hit occured (True) or not (False)   | `true`, `false`                                |
+| `cache_key`            | The Cache key used for this request                | `d2b758c****`|
+| `proxy_base_url`       | The base URL for the proxy server, the value of env var `PROXY_BASE_URL` on your server                | `https://proxy.example.com`|
+| `user_api_key_alias`   | An alias for the LiteLLM Virtual Key.| `prod-app1`        |
+| `user_api_key_user_id` | The unique ID associated with a user's API key.       | `user_123`, `user_456`                         |
+| `user_api_key_user_email` | The email associated with a user's API key.        | `user@example.com`, `admin@example.com`        |
+| `user_api_key_team_alias` | An alias for a team associated with an API key.    | `team_alpha`, `dev_team`                       |
+
+
+**Usage**
+
+Specify `langfuse_default_tags` to control what litellm fields get logged on Langfuse
+
+Example config.yaml 
+```yaml
+model_list:
+  - model_name: gpt-4
+    litellm_params:
+      model: openai/fake
+      api_key: fake-key
+      api_base: https://exampleopenaiendpoint-production.up.railway.app/
+
+litellm_settings:
+  success_callback: ["langfuse"]
+
+  # 👇 Key Change
+  langfuse_default_tags: ["cache_hit", "cache_key", "proxy_base_url", "user_api_key_alias", "user_api_key_user_id", "user_api_key_user_email", "user_api_key_team_alias", "semantic-similarity", "proxy_base_url"]
+```
+
 ### 🔧 Debugging - Viewing RAW CURL sent from LiteLLM to provider
 
 Use this when you want to view the RAW curl request sent from LiteLLM to the LLM API 

litellm/__init__.py

@@ -53,18 +53,7 @@ _known_custom_logger_compatible_callbacks: List = list(
     get_args(_custom_logger_compatible_callbacks_literal)
 )
 callbacks: List[Union[Callable, _custom_logger_compatible_callbacks_literal]] = []
-_langfuse_default_tags: Optional[
+langfuse_default_tags: Optional[List[str]] = None
-    List[
-        Literal[
-            "user_api_key_alias",
-            "user_api_key_user_id",
-            "user_api_key_user_email",
-            "user_api_key_team_alias",
-            "semantic-similarity",
-            "proxy_base_url",
-        ]
-    ]
-] = None
 _async_input_callback: List[Callable] = (
     []
 )  # internal variable - async custom callbacks are routed here.

litellm/integrations/braintrust_logging.py

@@ -181,9 +181,9 @@ class BraintrustLogger(CustomLogger):
 
                     # generate langfuse tags - Default Tags sent to Langfuse from LiteLLM Proxy
                     if (
-                        litellm._langfuse_default_tags is not None
+                        litellm.langfuse_default_tags is not None
-                        and isinstance(litellm._langfuse_default_tags, list)
+                        and isinstance(litellm.langfuse_default_tags, list)
-                        and key in litellm._langfuse_default_tags
+                        and key in litellm.langfuse_default_tags
                     ):
                         tags.append(f"{key}:{value}")
 
@@ -307,9 +307,9 @@ class BraintrustLogger(CustomLogger):
 
                     # generate langfuse tags - Default Tags sent to Langfuse from LiteLLM Proxy
                     if (
-                        litellm._langfuse_default_tags is not None
+                        litellm.langfuse_default_tags is not None
-                        and isinstance(litellm._langfuse_default_tags, list)
+                        and isinstance(litellm.langfuse_default_tags, list)
-                        and key in litellm._langfuse_default_tags
+                        and key in litellm.langfuse_default_tags
                     ):
                         tags.append(f"{key}:{value}")
 

litellm/integrations/langfuse.py

@@ -366,12 +366,11 @@ class LangFuseLogger:
             clean_metadata = {}
             if isinstance(metadata, dict):
                 for key, value in metadata.items():
-
                     # generate langfuse tags - Default Tags sent to Langfuse from LiteLLM Proxy
                     if (
-                        litellm._langfuse_default_tags is not None
+                        litellm.langfuse_default_tags is not None
-                        and isinstance(litellm._langfuse_default_tags, list)
+                        and isinstance(litellm.langfuse_default_tags, list)
-                        and key in litellm._langfuse_default_tags
+                        and key in litellm.langfuse_default_tags
                     ):
                         tags.append(f"{key}:{value}")
 
@@ -386,6 +385,11 @@ class LangFuseLogger:
                     else:
                         clean_metadata[key] = value
 
+            # Add default langfuse tags
+            tags = self.add_default_langfuse_tags(
+                tags=tags, kwargs=kwargs, metadata=metadata
+            )
+
             session_id = clean_metadata.pop("session_id", None)
             trace_name = clean_metadata.pop("trace_name", None)
             trace_id = clean_metadata.pop("trace_id", litellm_call_id)
@@ -468,9 +472,9 @@ class LangFuseLogger:
             clean_metadata["litellm_response_cost"] = cost
 
             if (
-                litellm._langfuse_default_tags is not None
+                litellm.langfuse_default_tags is not None
-                and isinstance(litellm._langfuse_default_tags, list)
+                and isinstance(litellm.langfuse_default_tags, list)
-                and "proxy_base_url" in litellm._langfuse_default_tags
+                and "proxy_base_url" in litellm.langfuse_default_tags
             ):
                 proxy_base_url = os.environ.get("PROXY_BASE_URL", None)
                 if proxy_base_url is not None:
@@ -583,6 +587,27 @@ class LangFuseLogger:
             verbose_logger.error(f"Langfuse Layer Error - {traceback.format_exc()}")
             return None, None
 
+    def add_default_langfuse_tags(self, tags, kwargs, metadata):
+        """
+        Helper function to add litellm default langfuse tags
+
+        - Special LiteLLM tags:
+            - cache_hit
+            - cache_key
+
+        """
+        if litellm.langfuse_default_tags is not None and isinstance(
+            litellm.langfuse_default_tags, list
+        ):
+            if "cache_hit" in litellm.langfuse_default_tags:
+                _cache_hit_value = kwargs.get("cache_hit", False)
+                tags.append(f"cache_hit:{_cache_hit_value}")
+            if "cache_key" in litellm.langfuse_default_tags:
+                _hidden_params = metadata.get("hidden_params", {}) or {}
+                _cache_key = _hidden_params.get("cache_key", None)
+                tags.append(f"cache_key:{_cache_key}")
+        return tags
+
 
 def _add_prompt_to_generation_params(
     generation_params: dict, clean_metadata: dict

litellm/proxy/proxy_config.yaml

@@ -40,4 +40,6 @@ general_settings:
 litellm_settings:
   fallbacks: [{"gemini-1.5-pro-001": ["gpt-4o"]}]
   success_callback: ["langfuse", "prometheus"]
+  langfuse_default_tags: ["cache_hit", "cache_key", "proxy_base_url", "user_api_key_alias", "user_api_key_user_id", "user_api_key_user_email", "user_api_key_team_alias", "semantic-similarity", "proxy_base_url"]
   failure_callback: ["prometheus"]
+  cache: True