Implementation:BerriAI Litellm PostHog Logger

Attribute	Value
Sources	`litellm/integrations/posthog.py`
Domains	Logging, Analytics, LLM Monitoring, Integrations
Last Updated	2026-02-15 16:00 GMT

Overview

The PostHogLogger sends LLM analytics events to PostHog following the PostHog LLM Analytics specification.

Description

PostHogLogger extends CustomBatchLogger to capture LLM call data as PostHog events ($ai_generation or $ai_embedding). It constructs event payloads with properties following PostHog's LLM Analytics format, including model information, input/output data, token counts, costs, latency, error information, trace IDs, and custom metadata. The logger supports both sync (immediate send) and async (batched) modes, per-request credential overrides via standard_callback_dynamic_params, distinct user ID resolution (from metadata, end_user, or trace_id), and graceful shutdown with atexit flush. Events are sent to PostHog's /batch/ endpoint grouped by API credentials. A mock mode is available via POSTHOG_MOCK environment variable.

Usage

Import and register PostHogLogger when you want to send LLM analytics to PostHog. Requires POSTHOG_API_KEY environment variable.

Code Reference

Source Location

litellm/integrations/posthog.py

Signature

class PostHogLogger(CustomBatchLogger):
    def __init__(self, **kwargs)

Import

from litellm.integrations.posthog import PostHogLogger

I/O Contract

Inputs

Parameter	Type	Required	Description
`kwargs`	`dict`	No	Additional arguments passed to `CustomBatchLogger`.

Environment Variables:

Variable	Required	Description
`POSTHOG_API_KEY`	Yes	PostHog project API key.
`POSTHOG_API_URL`	No	PostHog API URL. Defaults to `"https://us.i.posthog.com"`.
`POSTHOG_MOCK`	No	Enable mock mode for testing.

Key Methods

Method	Returns	Description
`log_success_event(kwargs, response_obj, start_time, end_time)`	`None`	Sync logging - sends event immediately.
`async_log_success_event(kwargs, response_obj, start_time, end_time)`	`None`	Async logging - queues event for batch send.
`async_log_failure_event(kwargs, response_obj, start_time, end_time)`	`None`	Async failure logging - queues event for batch send.
`create_posthog_event_payload(kwargs)`	`PostHogEventPayload`	Creates a PostHog event payload from request kwargs.
`async_send_batch()`	`None`	Sends batched events grouped by credentials to PostHog.

Outputs

Output	Type	Description
Side effect	HTTP POST	Sends events to `{posthog_host}/batch/` endpoint with `$ai_generation` or `$ai_embedding` event types.

Usage Examples

import litellm

# Register the callback
litellm.success_callback = ["posthog"]
litellm.failure_callback = ["posthog"]

response = litellm.completion(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}],
    metadata={
        "user_id": "user-123",  # Used as PostHog distinct_id
    },
)

# Direct instantiation
from litellm.integrations.posthog import PostHogLogger

logger = PostHogLogger()

Related Pages

BerriAI_Litellm_Braintrust_Logger - another analytics/observability integration
BerriAI_Litellm_OpenMeter_Logger - another metering/analytics integration
BerriAI_Litellm_MLflow_Logger - another tracing integration

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment