Ingestion overview

Ingestion is how trace data gets from your application into Moda: your app sends LLM requests, responses, and tool calls to the Ingestion API, and Moda groups them into traces — one per agent run, keyed by the conversation_id field (the trace ID) — for analysis. Use this page to pick the integration path that matches your stack. All paths authenticate with a Moda API key (called an ingestion key in the dashboard), created at Settings → Ingestion keys. See Authentication for key management.

​

Choose an integration

The SDKs and the OpenTelemetry path all deliver data to the same Ingestion API, so you can mix paths — for example, a Python backend on the SDK and a data pipeline on the HTTP API — and the data lands in the same traces.

​

What gets captured

​

Environments

Every event is tagged with one of three environments: development, staging, or production. Common aliases (dev, stg, prod) are normalized; unrecognized values are recorded as production. Each integration page shows how to set the environment for that path, and the dashboard lets you filter by it.

​

What happens after ingest

1. Events are grouped into traces by conversation_id (the trace ID). New traces appear on the Traces page within minutes.
2. Signal analyses — user emotion, tool failures, hallucinations, laziness — typically complete within minutes of ingest.
3. Problems and use-case clusters build automatically as evidence accumulates.

See Analysis for what each stage produces. Retrying a failed request is safe. Delivery semantics and limits are covered in Reliability.

​

Next steps

• Quickstart — send your first trace end to end.
• Python SDK or Node.js SDK — auto-instrumentation for OpenAI and Anthropic apps.
• Data model — how traces, messages, and signals relate.
• Reliability — limits, retries, and error handling.