pelennor 0.1.1

Send your AI agent's OpenTelemetry traces to Pelennor.

pelennor

Send your AI agent's OpenTelemetry traces to Pelennor, where they become conversations you can read, search, and turn into evals.

pip install pelennor

Quickstart

import pelennor

pelennor.init(api_key="plk_...")   # the key from your agent's Lens page

with pelennor.conversation(id=thread_id, user="cust_123"):
    agent.invoke(...)

That's it. Every LLM call, tool call, and span your agent produces inside the conversation block is exported to Pelennor and stitched into a single conversation keyed by id.

init() reads PELENNOR_API_KEY from the environment when you don't pass it, so nothing but the key need live in code. Traces go to Pelennor's hosted endpoint by default; set endpoint= (or PELENNOR_ENDPOINT) only if you self-host.

Why the conversation block?

OpenTelemetry has no concept of a conversation. Each request is its own trace, and the next turn arrives later as a completely separate one. pelennor.conversation(id=...) stamps a shared id onto every span started inside it, so Pelennor can join a whole multi-turn exchange back together. The id is whatever your app already calls a thread, session, or chat. Reuse it across turns and the turns line up.

It's built on contextvars, so the binding follows your code across await points and worker threads, which is where agent frameworks do their work.

Auto-instrumentation

The base install exports spans and stamps conversation ids. To capture LLM and tool calls automatically, install the extra for your stack:

pip install "pelennor[openai]"        # OpenAI
pip install "pelennor[langchain]"     # LangChain / LangGraph
pip install "pelennor[anthropic]"     # Anthropic
pip install "pelennor[llama-index]"   # LlamaIndex

init() activates whichever of these are installed. Already using OpenInference or OpenLLMetry? Pelennor reads both, so you can point your existing exporter at it and skip this SDK entirely.

Options

pelennor.init(
    api_key="plk_...",
    endpoint="https://self-hosted",  # or PELENNOR_ENDPOINT; defaults to the cloud
    service_name="support-bot",      # shows up on your traces
    capture_content=False,           # send metadata only, never message text
    instrument=True,                 # auto-activate installed instrumentors
)

capture_content=False tells the instrumentors not to record prompt or completion text, for when conversation content can't leave your environment.

It will never break your agent

init() never raises. A missing key, an unreachable endpoint, or a bad span all degrade to "no telemetry" and a log line, never an exception in your request path. Telemetry is not allowed to take your agent down.

License

MIT