Integrate Modal-hosted Marker and add env-based switch to offload layout extraction from local RQ

[priyankeshh]

Summary

This PR adds a clean integration path to run Marker on Modal instead of locally via RQ. It introduces an environment-based switch (MARKER_RUN_MODE) and an HTTP client that forwards PDFs to a Modal-hosted Marker service. This lets us use proper GPU/compute and avoid HF Space resource limits. Layout correctness improvements can be iterated later; this PR focuses on wiring and compute offload.

Key changes

• ocr_jobs.py
  • Loads .env early to pick up configuration.
  • Adds a modal execution branch that calls the Modal endpoint for output_format=json and parses the response.
  • Logs the selected run mode and endpoint for visibility.
• extralit_server/integrations/marker_modal_client.py
  • New lightweight client to POST PDFs to Modal /convert, with timeout and basic error handling.
  • Reads MARKER_MODAL_BASE_URL and optional MARKER_MODAL_TIMEOUT_SECS from env.
• .env.example
  • Documents MARKER_RUN_MODE=modal and MARKER_MODAL_BASE_URL required to enable Modal integration.

Configuration

• Set the following variables (local .env for dev; real env in prod/worker runtime):
  • MARKER_RUN_MODE=modal
  • MARKER_MODAL_BASE_URL=https://YOUR-ENDPOINT.modal.run
  • Optional: MARKER_MODAL_TIMEOUT_SECS=600

How to test

1. Ensure a Modal Marker service is deployed and healthy

• Deploy the example from datalab-to/marker (examples) with JSON path fixed
• Verify it responds:
  • curl https://YOUR-ENDPOINT.modal.run/health
  • curl -F "file=@/path/to/sample.pdf" -F "output_format=json" https://YOUR-ENDPOINT.modal.run/convert

2. Local CLI test (bypassing RQ)

• In repo root (so .env is picked up):
  • python extralit-server/src/extralit_server/jobs/ocr_jobs.py "/path/to/sample.pdf" --extract-text
• Expected logs:
  • Starting Marker layout extraction ... (mode=modal)
  • Using Modal endpoint: https://YOUR-ENDPOINT.modal.run
• Expected output: JSON with tables, figures, text_blocks, metadata.

Checklist

• Env-based switch to Modal
• New Modal client and wiring in ocr_jobs
• .env.example updated
• Local CLI test path documented
• Optional: add Modal auth header support
• Optional: retries/backoff and timeouts tuning

[JonnyTran]

Hey @priyankeshh, these code are handling nested dict for both function arguments and returns types, and it's better to define our own data types with pydantic.BaseModel or import a defined type from marker. We want to avoid using these nested dict parsing method since it's hard to maintain this code when there's no typehinting. Can you define the models at extralit-server/src/extralit_server/api/schemas/v1/document/layout.py?

You can see here for an example: https://github.com/Extralit/extralit/blob/e919e0453c808c89a4e7bfa331f1542fde5c2674/extralit-server/src/extralit_server/api/schemas/v1/document/metadata.py

[priyankeshh]

refactored the code to use Pydantic models for layout parsing and defined them in layout.py as suggested.

[JonnyTran]

Save later for determining modal external call due to complexity