LiteLLM AI Observability installation

Let AI instrument your LLM calls for you

Skip the manual setup — run this in your project and the wizard installs the SDK and wires up AI Observability for you.

Learn more

1
LiteLLM Requirements
Required
Note: Use LiteLLM as a Python SDK or as a proxy server. PostHog observability requires LiteLLM version 1.77.3 or higher.

Install LiteLLM

Required

Choose your installation method based on how you want to use LiteLLM:

pip install litellm
# Install via pip
pip install 'litellm[proxy]'

# Or run via Docker
docker run --rm -p 4000:4000 ghcr.io/berriai/litellm:latest

Configure PostHog observability

Required

Configure PostHog by setting your project token and host as well as adding posthog to your LiteLLM callback handlers. You can find your project token in your project settings.

import os
import litellm

# Set environment variables
os.environ["POSTHOG_API_KEY"] = "<ph_project_token>"
os.environ["POSTHOG_API_URL"] = "https://us.i.posthog.com"  # Optional, defaults to https://app.posthog.com

# Enable PostHog callbacks
litellm.success_callback = ["posthog"]
litellm.failure_callback = ["posthog"]  # Optional: also log failures
# config.yaml
model_list:
- model_name: gpt-5-mini
  litellm_params:
    model: gpt-5-mini

litellm_settings:
  success_callback: ["posthog"]
  failure_callback: ["posthog"]  # Optional: also log failures

environment_variables:
  POSTHOG_API_KEY: "<ph_project_token>"
  POSTHOG_API_URL: "https://us.i.posthog.com"  # Optional

Call LLMs through LiteLLM

Required

When you use LiteLLM to call an LLM provider, PostHog automatically captures an $ai_generation event. Identity and trace data travel through metadata, since LiteLLM has no dedicated posthog_trace_id parameter.

from posthog import Posthog
import time, uuid, json

posthog = Posthog("<ph_project_token>", host="https://us.i.posthog.com")

trace_id = str(uuid.uuid4())

response = litellm.completion(
    model="gpt-5-mini",
    messages=[{"role": "user", "content": "What's the weather in Paris?"}],
    tools=tools,
    metadata={
        "user_id": "user_123",                # Maps to PostHog distinct_id
        "company": "company_id_in_your_db",   # Custom property
        "$ai_session_id": "conversation-abc",
        "$ai_trace_id": trace_id,
    },
)
# Start the proxy (if not already running)
litellm --config config.yaml

# Make a request to the proxy
curl -X POST http://localhost:4000/chat/completions                                       -H "Content-Type: application/json"                                       -d '{
    "model": "gpt-5-mini",
    "messages": [
      {"role": "user", "content": "Tell me a fun fact about hedgehogs"}
    ],
    "metadata": {
      "user_id": "user_123",
      "company": "company_id_in_your_db", # Custom property
      "$ai_session_id": "conversation-abc" # Groups calls into one session
    }
  }'

Notes:

This works with streaming responses by setting stream=True.

To disable logging for specific requests, add {"no-log": true} to metadata.

Pass $ai_session_id in metadata to group calls from the same conversation into one PostHog session.

If you want to capture LLM events anonymously, do not pass a user_id in metadata.

See our docs on anonymous vs identified events to learn more.

You can expect captured $ai_generation events to have the following properties:

Property	Description
`$ai_model`	The specific model, like `gpt-5-mini` or `claude-4-sonnet`
`$ai_latency`	The latency of the LLM call in seconds
`$ai_time_to_first_token`	Time to first token in seconds (streaming only)
`$ai_tools`	Tools and functions available to the LLM
`$ai_input`	List of messages sent to the LLM
`$ai_input_tokens`	The number of tokens in the input (often found in response.usage)
`$ai_output_choices`	List of response choices from the LLM
`$ai_output_tokens`	The number of tokens in the output (often found in `response.usage`)
`$ai_total_cost_usd`	The total cost in USD (input + output)
[...]	See full list of properties

Capture tool calls as spans

Optional

Capture each tool call as a span yourself, as the example below does right after the generation that triggered it.

for call in response.choices[0].message.tool_calls or []:
    start = time.time()
    result = run_tool(call.function.name, json.loads(call.function.arguments))

    posthog.capture(
        distinct_id="user_123",
        event="$ai_span",
        properties={
            "$ai_trace_id": trace_id,
            "$ai_session_id": "conversation-abc",
            "$ai_span_id": str(uuid.uuid4()),
            "$ai_span_name": call.function.name,
            "$ai_input_state": call.function.arguments,
            "$ai_output_state": result,
            "$ai_latency": time.time() - start,
        },
    )

See spans for the full list of span properties.

Capture embeddings

Optional

PostHog can also capture embedding generations as $ai_embedding events through LiteLLM:

response = litellm.embedding(
    input="The quick brown fox",
    model="text-embedding-3-small",
    metadata={
        "user_id": "user_123",  # Maps to PostHog distinct_id
        "company": "company_id_in_your_db"  # Custom property
    }
)
# Make an embeddings request to the proxy
curl -X POST http://localhost:4000/embeddings                                       -H "Content-Type: application/json"                                       -d '{
    "input": "The quick brown fox",
    "model": "text-embedding-3-small",
    "metadata": {
      "user_id": "user_123",
      "company": "company_id_in_your_db" # Custom property
    }
  }'

Verify traces and generations
Recommended
Confirm LLM events are being sent to PostHog
Let's make sure LLM events are being captured and sent to PostHog. Under AI Observability, you should see rows of data appear in the Traces and Generations tabs.

Check for LLM events in PostHog

Next steps

Recommended

Now that you're capturing AI conversations, continue with the resources below to learn what else AI Observability enables within the PostHog platform.

Resource	Description
Basics	Learn the basics of how LLM calls become events in PostHog.
Generations	Read about the `$ai_generation` event and its properties.
Traces	Explore the trace hierarchy and how to use it to debug LLM calls.
Spans	Review spans and their role in representing individual operations.
Anaylze LLM performance	Learn how to create dashboards to analyze LLM performance.

LiteLLM AI Observability installation

Contents

LiteLLM Requirements

Install LiteLLM

Configure PostHog observability

Call LLMs through LiteLLM

Capture tool calls as spans

Capture embeddings

Verify traces and generations

Next steps

Community questions

Was this page useful?