Transcript Service
Transcript Service
Internal service that turns audio URLs into validated CustomerInteraction@v1.0 documents.
Transcript Service is the Ontopix transcription engine. It accepts an audio URL,
transcribes the conversation with ElevenLabs, optionally corrects domain-specific
terms with Bedrock, resolves participant roles, and writes a validated
CustomerInteraction@v1.0 document for downstream Ontopix services.
Documentation
| Page | Description |
|---|---|
| API Reference | Job creation, status polling, callbacks, request fields, and status lifecycle. |
| Processing Pipeline | Lambda responsibilities, Step Functions flow, data boundaries, and failure path. |
| Sandbox | LocalStack-based development environment and local request workflow. |
| Operations | Required tasks, infrastructure commands, constraints, and escalation rules. |
How It Fits
The service owns the conversion from source audio into the canonical Ontopix interaction format. It does not perform semantic enrichment or audit scoring.
flowchart LR
Audio[Audio URL] --> Transcript[transcript-service]
Transcript --> Interaction[CustomerInteraction@v1.0]
Interaction --> Enrich[enrich-service]
Interaction --> Audit[audit-service]
Output Contract
Successful jobs produce CustomerInteraction@v1.0 documents validated through
@ontopix/schemas. The service must not persist or return non-canonical
transcription shapes as final output.
Related Documentation
- Schemas — canonical
CustomerInteraction@v1.0contract. - Engineering Handbook — repository, task, sandbox, and infrastructure patterns.
- Enrich Service — semantic enrichment of transcript output.