feat(llma): Add Pydantic AI instrumentation via OpenTelemetry

[andrewm4894]

Summary

Add OpenTelemetry-based instrumentation for Pydantic AI agents. This provides a foundation for OTel-instrumented AI frameworks.

New modules:

• posthog.ai.otel - Generic OTel SpanExporter that translates spans to PostHog events
• posthog.ai.pydantic_ai - Pydantic AI-specific wrapper and setup function

Usage

from posthog import Posthog
from posthog.ai.pydantic_ai import instrument_pydantic_ai
from pydantic_ai import Agent

posthog = Posthog(project_api_key="phc_xxx", host="https://us.i.posthog.com")

# One-liner setup
instrument_pydantic_ai(posthog, distinct_id="user_123")

# Use Pydantic AI normally - automatically traced
agent = Agent("openai:gpt-4")
result = await agent.run("Hello")

Architecture

Pydantic AI Agent -> OTel Spans -> PydanticAISpanExporter -> PostHogSpanExporter -> PostHog Events

• PostHogSpanExporter: Generic exporter following GenAI semantic conventions
• PydanticAISpanExporter: Normalizes Pydantic AI message formats and tool attribute names

Changes

• Add posthog/ai/otel/ with generic OTel exporter
• Add posthog/ai/pydantic_ai/ with Pydantic AI wrapper and instrument_pydantic_ai()
• Add optional dependencies: pip install posthog[pydantic-ai] or pip install posthog[otel]
• Add 60 tests covering both modules

Test plan

• Unit tests for OTel exporter (29 tests)
• Unit tests for Pydantic AI exporter (24 tests)
• Unit tests for instrument function (7 tests)
• End-to-end testing with llm-analytics-apps: feat: Add Pydantic AI provider with PostHog OTel instrumentation llm-analytics-apps#27

[greptile-apps]

_{13 files reviewed, 4 comments}

_{Edit Code Review Agent Settings | Greptile}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".