Ingest a file's chunks as documents into the knowledge graph

Chunk an uploaded file via the doc-chunker and queue its sections as graph documents.

Without force, acts on the file's latest ingestion job: none yet starts one; completed returns it as is, after retrying the graph removal of any document whose DELETE failed (503 if that fails again); failed/pending resumes it (re-chunking if it has no documents yet); in_progress returns 409, unless it has had no update for graph.ingestion_stale_after_seconds, in which case it is treated as abandoned and resumed. A running job's worker bumps updated_at periodically (see GraphIngestionService._heartbeat), so only a job whose worker is gone (e.g. a server restart) goes stale.

curl -X POST "https://deepfellow-server-host/v1/graph/files/5eb7cf5a86d9755df3a6c593/ingest?force=false" \  -H "OpenAI-Organization: string" \  -H "OpenAI-Project: string"

{
  "status": "queued",
  "document_count": 0,
  "job_id": "string"
}

{
  "detail": [
    {
      "loc": [
        "string"
      ],
      "msg": "string",
      "type": "string",
      "input": null,
      "ctx": {}
    }
  ]
}