ai-parrot-openlit-bridge

Minimal OTLP endpoint validation helper for AI-Parrot's OpenLIT integration (FEAT-462 — Unified Telemetry Bus).

AI-Parrot no longer depends on the openlit Python SDK (a monkey-patching LLM SDK instrumentor whose pinned openai version range conflicted with ai-parrot's own openai dependency). OpenLIT is now a pure deployment-time OTLP endpoint — configure ObservabilityConfig.otlp_targets (multi-endpoint OTLP export) or OpenLitUsageRecorder (usage-only spans) in ai-parrot itself. This package provides two small, optional helpers on top of that:

1. validate_endpoint(url) — an async probe that checks whether an OTLP collector is reachable, with zero heavy dependencies (only aiohttp, already a workspace dependency).
2. parrot-openlit-check — a CLI wrapper around the same probe.
3. A bundled docker-compose.openlit.yml to run an OpenLIT collector + dashboard locally.

Installation

pip install ai-parrot-openlit-bridge
# or, via the ai-parrot extra:
pip install "ai-parrot[observability-openlit]"

Usage

Python

from ai_parrot_openlit_bridge import validate_endpoint

status = await validate_endpoint("http://localhost:4318")
if status.reachable:
    print(f"OTLP collector reachable (status={status.status_code})")
else:
    print(f"Unreachable: {status.error}")

CLI

parrot-openlit-check http://localhost:4318
# ✅ Endpoint reachable: http://localhost:4318
#    Status: 200

Exit code is 0 when reachable, 1 otherwise — suitable for health-check scripts / CI smoke tests.

Local OpenLIT collector

docker compose -f docker-compose.openlit.yml up -d
parrot-openlit-check http://localhost:4318

Then point ai-parrot at it, e.g.:

export OBSERVABILITY_ENABLED=true
export OBSERVABILITY_BACKEND=otel
export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318