Retrieval and Documents
Orlo Platform supports task-scoped document ingestion and retrieval-backed inference.
Documents
Documents attach to tasks. Current supported ingestion modes include:
- inline text
- inline markdown
- pre-chunked content
- PDF ingestion with async extraction
Retrieval
Retrieval is task-scoped. Multiple documents under the same task can contribute context to one inference.
PDF flow
PDFs are handled asynchronously:
- upload or start direct upload
- finalize the document
- extract text
- chunk content
- embed chunks
- mark the document ready
Attribution
Retrieval-backed inference returns attribution in debug/audit contexts so you can inspect what source material contributed to the response.