refactor: move inline OCR imports to module level with a global availability flag (#591)

[suhaniiz]

📋 PR Checklist

Thank you for contributing to PDF-Assistant-RAG! 🎉
Please fill out this template before submitting. PRs without it filled in will be closed.

---

🔗 Related Issue

Closes #591

---

📝 What does this PR do?

Optimizes performance and readability within the image captioning module by refactoring how OCR dependencies are managed.

• Moved the expensive inline PIL and pytesseract imports out of the _ocr_caption function loop and up to the module initialization level.
• Implemented a top-level global boolean flag HAS_OCR to handle conditional dependency loading safely at runtime.
• Updated _ocr_caption to immediately evaluate the state of HAS_OCR rather than spinning up repetitive try/except execution paths on every function invocation. This removes unnecessary interpreter overhead during long document batch processing routines.

---

🗂️ Type of Change

• 🐛 Bug fix
• ✨ New feature
• 🔧 Refactor / code cleanup
• 📝 Documentation update
• 🎨 UI / styling change
• ⚙️ CI / tooling / config change
• 🧪 Tests

---

🧪 How was this tested?

• Verified error safety handling locally when target libraries are absent vs. present.
• Confirmed function behavior maintains standard string fallbacks cleanly without altering functional outputs.

---

📸 Screenshots (if UI change)

N/A

---

⚠️ Anything to flag for reviewers?

None. This change does not alter the external API surface or functionality of caption_image.

---

✅ Self-Review Checklist

• My branch is based on dev, not main
• I have not added any secrets / API keys
• I have not modified main branch or any HuggingFace deployment config
• My code follows the existing style (no unnecessary formatting changes)
• I have updated relevant docs / comments if needed

[suhaniiz]

Hi @param20h,

Since I am the author who originally opened this issue (#591), I have a deep understanding of the performance overhead we wanted to solve here.

I've successfully isolated the heavy inline imports up to the module level and tied them to a clean HAS_OCR global flag initialization. This completely eliminates the repetitive try/except overhead during batch chunking while keeping the fallback safe.

Please take this PR into consideration—it perfectly bridges the issue I highlighted. Ready for review! 🚀

[suhaniiz]

@param20h

[param20h]

merge conflicts'

[suhaniiz]

i cannot see the conflicts to be resolved!