Merge pull request #96 from trubrics/remove_default_latency

remove default latency

Author: jeffkayne

CHANGELOG.md

@@ -10,6 +10,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [1.8.2] - 2025-02-10
 ### Fixed
 - Added missing flushing limit values, and setup automated flush if limit is reached
+### Changed
+- Change LLM tracking latency to seconds and remove default latency from event properties
 
 
 ## [1.8.1] - 2025-02-07

README.md

@@ -40,7 +40,7 @@ client.track_llm(
     assistant_id="gpt-4",
     generation="The capital of France is Paris.",
     properties={"model": "gpt-4"},
-    latency=150  # milliseconds
+    latency=1.5  # seconds
 )
 ```
 

examples/sdk_test.py

@@ -0,0 +1,16 @@
+import os
+
+from trubrics import Trubrics
+
+trubrics = Trubrics(api_key=os.environ["TRUBRICS_API_KEY"])
+trubrics.track(event="Sign up", user_id="user_id")
+
+trubrics.track_llm(
+    user_id="user_id",
+    prompt="What is Trubrics?",
+    assistant_id="gpt4o",
+    generation="Trubrics is a product analytics platform for AI applications.",
+    latency=2,
+)
+
+trubrics.close()

trubrics/main.py

@@ -116,7 +116,7 @@ def track_llm(
         generation: str,
         properties: dict | None = None,
         timestamp: datetime | None = None,
-        latency: int = 1,
+        latency: float | None = None,
     ):
         """
         Track an LLM prompt and generation.
@@ -127,10 +127,10 @@ def track_llm(
             generation (str): The generated response from the LLM.
             properties (dict | None): Additional properties to track.
             timestamp (datetime | None): The timestamp of the generation event. If None, the current time in UTC is used.
-            latency (int): The latency in milliseconds between the prompt and the generation. Defaults to 1.
+            latency (float | None): The latency in seconds between the prompt and the generation. Defaults to 1.
         """
         generation_timestamp = timestamp or datetime.now(timezone.utc)
-        prompt_timestamp = generation_timestamp - timedelta(milliseconds=latency)
+        prompt_timestamp = generation_timestamp - timedelta(seconds=latency or 1)
 
         self.track(
             event="Prompt",
@@ -146,7 +146,7 @@ def track_llm(
                 "$text": generation,
                 "$assistant_id": assistant_id,
                 "$prompt": prompt,
-                "latency(ms)": latency,
+                "latency(s)": latency,
                 **(properties or {}),
             },
             timestamp=generation_timestamp,