Storage and Observability¶

Storage Providers¶

Track all agent interactions: - SQL-based with SQLModel/SQLAlchemy - Per-agent or shared database - Analytics via CLI: agentpool history stats

Observability¶

Logfire + OpenTelemetry integration: - Structured logging with context via logfire.configure() - Auto-instrumentation: logfire.instrument_pydantic_ai(), logfire.instrument_mcp(), logfire.instrument_fastapi(app) - Manual instrumentation: @logfire.instrument decorator and with logfire.span(...) context manager - Trace agent execution end-to-end across async task boundaries and subagent delegation - Export to any OTLP-compatible backend (SigNoz, Jaeger, Honeycomb, etc.) via OTEL env vars - Disabled in tests via env vars (see conftest.py) - See Telemetry & Span Instrumentation for mandatory practices