parlot-instrumentation-pipecat

Parlot OpenTelemetry instrumentation for Pipecat production voice AI agents.

Emits Parlot’s Conversation Contract (parlot.session, parlot.turn, GenAI/voice spans) so Pipecat sessions show up in the same debugger as LiveKit and LangGraph.

Install

pip install parlot-instrumentation-pipecat
# or: pip install "parlot[pipecat]"

Requirements

For parlot.turn (user/agent transcript on the timeline), the pipeline must include LLMUserAggregator and LLMAssistantAggregator — typically via LLMContextAggregatorPair. Parlot does not create them. Put the assistant aggregator after transport.output().

Call parlotize() before building aggregators so handlers auto-wire. If they already exist, call attach_aggregators(user, assistant). Without aggregators you may still get session + STT/LLM/TTS spans, but not contract turns.

Your own aggregator event handlers and observers=[...] keep working — Parlot appends alongside them.

Instrument

from pipecat.processors.aggregators.llm_response_universal import LLMContextAggregatorPair
from parlot.instrumentation.pipecat import attach_aggregators, parlotize

parlotize("my-pipecat-agent", version="0.1.0")

context_aggregator = LLMContextAggregatorPair(context)
# attach_aggregators(...) only if the pair was built before parlotize()

pipeline = Pipeline([
    transport.input(),
    stt,
    context_aggregator.user(),
    llm,
    tts,
    transport.output(),
    context_aggregator.assistant(),
])

worker = PipelineWorker(
    pipeline,
    params=PipelineParams(enable_metrics=True, enable_usage_metrics=True),
    enable_tracing=True,
    enable_turn_tracking=True,
)

export PARLOT_ENDPOINT=https://ingest.parlot.ai
export PARLOT_API_KEY=<org-scoped-key>

Example

See examples/pipecat/minimal-voice/ (adapted from Pipecat's open-telemetry/langfuse demo).

Docs

• Pipecat guide
• Concepts