OpenTelemetry export

pip install "agentlens-toolkit[otel]" adds an exporter that sends traces to any OpenTelemetry backend (Jaeger, Grafana Tempo, Phoenix, Langfuse, ...) using the GenAI semantic conventions. Each run is exported as invoke_agent <name>, LLM calls as chat <model> (gen_ai.provider.name, gen_ai.request.model, gen_ai.usage.input_tokens, ...), tools as execute_tool <tool> (gen_ai.tool.name), with the recorded timings and errors:

import agentlens
from agentlens.integrations.otel import OTelExporter, export_trace, fan_out
from agentlens.store import SQLiteStore

exporter = OTelExporter()          # the global TracerProvider; configure OTLP as usual
with SQLiteStore("runs.db") as store:
    with agentlens.run("support", sink=fan_out(store.save_trace, exporter)):
        ...                        # your agent, with wrap_llm / wrap_tool
    for trace in store.list_traces(name="support"):    # or export what is already stored
        export_trace(trace)

Prompts, outputs and tool arguments are not exported unless you pass capture_content=True (they may hold personal data). Export errors are logged, never raised into the agent.

API: agentlens.integrations.otel.