Document uploads

Audience

Integrators uploading evidence documents to a project.

What you'll accomplish

Upload a file via the presigned flow, finalize the source, and know where to poll for extraction status.

Project documents are sources (kind: document). Uploading triggers background extraction. The complete response includes sourcePid for polling.

Requires projects:write. See Authentication.

Presigned upload

File bytes go directly to object storage. This is the supported path for integrators and large files.

POST /projects/:projectPid/documents/upload-init
  → uploadUrl + storage key
PUT uploadUrl                         (no Authorization header)
POST /projects/:projectPid/documents/upload-complete
  → sourcePid

Step-by-step

1. Init — presigned PUT URL (valid five minutes):

curl -sS -X POST 'https://api.air.thalus.ai/projects/<projectPid>/documents/upload-init' \
  -H 'Authorization: Bearer <rawKey>' \
  -H 'Content-Type: application/json' \
  -d '{"filename":"evidence.pdf","contentType":"application/pdf"}'

2. Upload — PUT raw file to uploadUrl with the same Content-Type. No AIR auth header.

3. Complete — start extraction:

curl -sS -X POST 'https://api.air.thalus.ai/projects/<projectPid>/documents/upload-complete' \
  -H 'Authorization: Bearer <rawKey>' \
  -H 'Content-Type: application/json' \
  -d '{
    "s3Key": "<storage key from init>",
    "filename": "evidence.pdf",
    "contentType": "application/pdf"
  }'

Poll GET /projects/:projectPid/documents until status is no longer processing — Async jobs.

Limits and supported types

Limit	Value
Max file size	50 MiB per file
Max artifacts per assessment	5
Presigned URL TTL	5 minutes

Supported extensions: .pdf, .docx, .pptx, .xlsx, .txt, .md, .csv.

Very large files may time out during extraction even when upload succeeds.

Common failures

Symptom	Likely cause	Fix
403 on init	Missing projects:write	Add scope or use integrations:full
404 on init	Wrong projectPid or wrong domain	GET /domains/<domainPid>/projects
PUT fails or 403 from storage	Wrong Content-Type, expired URL, or corrupt body	Re-run init within five minutes; match Content-Type
409 on complete	Storage key already used or upload never finished	Fresh init → PUT → complete
Source error after complete	Unsupported or corrupt file	Check lastError on document row
Stuck in processing	Large file or parser timeout	Wait with backoff; check lastError; retry regenerate if needed

After upload

Action	Route
List documents	GET /projects/:projectPid/documents
List artifacts	GET /projects/:projectPid/artifacts
Download original	GET /projects/:projectPid/documents/:sourcePid/download
Re-run extraction	POST /projects/:projectPid/documents/:sourcePid/regenerate
Archive	POST /projects/:projectPid/documents/:sourcePid/archive

Deleting a document archives derived artifacts used as assessment evidence.

Related

• Integration flow — upload in a full pipeline
• Async jobs — poll until connected
• Errors — 409 and upload troubleshooting