# Eldros

## Docs

- [Capturing conversations](https://platform.eldros.ai/docs/capture-options.md): Wrap each exchange in a turn — transcript recorded, traces nested, sessions grouped.
- [Sessions, turns & traces](https://platform.eldros.ai/docs/concepts/sessions.md): The three-layer model behind every judged conversation.
- [Simulation & episodes](https://platform.eldros.ai/docs/concepts/simulation.md): How simulated conversations are identified — and how to attach your internal traces to a test run.
- [Configuration](https://platform.eldros.ai/docs/configuration.md): The API key, every setting, and exactly what init() does at startup.
- [Eldros](https://platform.eldros.ai/docs/index.md): Judge every production conversation your AI agent has — scores, flagged issues, and the trace behind every reply.
- [LiveKit Agents](https://platform.eldros.ai/docs/integrations/livekit.md): One call before session.start() — every conversation turn recorded, session-grouped with your LLM traces.
- [Pipecat](https://platform.eldros.ai/docs/integrations/pipecat.md): Hook the turn-stopped events, finalize at pipeline end — transcripts and Pipecat's native traces in one pipeline.
- [Quickstart](https://platform.eldros.ai/docs/quickstart.md): From install to judged conversations in three steps.
- [LLM tracing](https://platform.eldros.ai/docs/tracing.md): Every model call becomes a trace span automatically — no change to your model code.