feat: add OCR language selection to /extract API and improve LLM prompt

📄 Merge Request Description

🧩 Overview

This MR introduces parameterized OCR language selection and a refined LLM prompt to improve multilingual extraction accuracy and usability.

---

🔧 Changes

Change	File(s)	Summary
API update	bookextractor/main.py	Added lang enum (en, te, devanagari) as a Form field to /extract. The selected value is passed to process_pdf.
Pipeline signature	bookextractor/pipeline.py	Updated method signature to process_pdf(pdf_path, benchmark=False, lang="en"), forwarding lang to parse_pdf_via_vparse.
VParse client	bookextractor/vparse_client.py	Updated function parse_pdf_via_vparse(file_path, lang="en") to include lang_list: <lang> in the request payload.
Prompt overhaul	bookextractor/pipeline.py (method: extract_semantic_fields)	Rewritten prompt with explicit instructions, multilingual handling, and strict JSON output rules.
Enum definition	bookextractor/main.py	Added OCRLanguage enum for Swagger UI dropdown.

---

🧾 Commits

• f771dea (HEAD → feat/dockerize) feat: add OCR language selection to /extract API and improve LLM prompt

• 523ec9c feat(docker): containerize bookextractor + vparse with model persistence

• 77ba5ba fix(docker): add build deps and CPU torch to BookExtractor Dockerfile

• d82fdc6 feat(docker): add multi-backend VParse OCR with per-mode Dockerfiles

• 132a006 (origin/feat/multi-format-pipeline-phase1, feat/multi-format-pipeline-phase1) feat: implement multi-format pipeline (Phase 1)

• 629b082 feat: implement multi-format pipeline (Phase 1)

---

🧪 Testing

• API

  • Swagger UI displays lang dropdown with supported values

• Manual Test

curl -F file=@sample.pdf -F lang=en http://localhost:8000/extract

• Returns JSON with correctly extracted English metadata

• Unit Tests

  • Existing tests pass without modification:

    • tests/test_pipeline.py
    • tests/test_vparse_client.py

---

✅ Checklist

• Code compiles

python -m py_compile $(git ls-files '*.py')

• Documentation updated (README.md, API usage section)
• No new lint warnings (ruff)
• CI pipeline passes (gitlab-ci.yml)

---

🔗 Related Issues

• Closes #5 (closed) – “Add multilingual OCR language selection”

---

🚀 Status

Ready for review & merge into feat/multi-format-pipeline-phase1.