fix: resolve type hint mismatch in deduplicator and broaden embedder exceptions (#259)

[prasiddhi-105]

Linked issue

Closes #259

What this PR does

Resolves a type system mismatch between the main scan workflow engine and the machine learning deduplication / embedding logic. Previously, type hints on the deduplicator signature led to a silent failure where Pydantic Finding objects were being processed with dictionary-style getattr() defaults, resulting in zero-information empty text embeddings. This PR introduces a robust runtime instance-based type extractor, fixes the functional type hints, adds a type assertion guard inside the scanning controller loop, and broadens initialization exception wrappers for SentenceTransformer to protect the server boot process from third-party runtime failures (e.g., PyTorch out-of-memory or HuggingFace timeouts).

Type of change

• Bug fix
• New feature
• ML model / training pipeline
• Refactor (no behaviour change)
• Documentation
• Tests only

ML tier (if applicable)

• Tier 1 — Triage
• Tier 2 — Predictive
• Tier 3 — Autonomous
• Not ML-related

Stack affected

• Backend
• Frontend
• Both

---

Changes

Backend

• backend/app/ml/deduplicator.py: Upgraded data type signatures and annotations to explicitly accept structural collections of Union[Finding, Dict[str, Any]].
• backend/app/ml/embedder.py: Introduced a localized polymorph conditional parser _extract_text to handle extraction paths natively using string parsing via .get() for raw dictionaries or getattr() for model properties.
• backend/app/ml/embedder.py: Widen the initialization block check statement from except ImportError to except Exception to capture connection drops or memory faults gracefully.
• backend/app/main.py: Deployed a clear static type check assertion boundary condition wrapper on the engine array loop directly before calculating cosine similarity.

Frontend

• No frontend changes.

New dependencies

• None.

Database / schema changes

• None.

---

Testing

How did you test this?

• Validated code layout alignment against python syntax trees locally.
• Verified parsing function capabilities by running tests using standard dictionaries and live Pydantic models.
• Mocked initialization dependency faults to check that the network loading layer correctly outputs a system warning instead of completely crashing the engine startup loop.

Checklist

• Tested locally end-to-end (upload ZIP or GitHub URL → scan → findings returned correctly)
• New ML model falls back gracefully when model file is absent
• No new console.error or unhandled Python exceptions introduced
• Added or updated tests where applicable
• requirements.txt / package.json updated if new dependencies added
• New model files (.pkl, .pt, etc.) are gitignored, not committed

---

Anything reviewers should focus on

Please check the structural implementation of the new private helper _extract_text within embedder.py to ensure it seamlessly toggles between dictionary-like and object-like item evaluations without degrading execution processing loops during full scan passes.

[github-actions]

🎉 Thank you @prasiddhi-105 for submitting a Pull Request!

We're excited to review your contribution.

Before Review

✅ Ensure all CI checks pass
✅ Complete the PR template
✅ Link the related issue

⚡ Want faster reviews and contributor support?

Join our Discord community:

🔗 https://discord.gg/FcXuyw2Rs

Maintainers and mentors are active there and can help resolve blockers quickly.

Happy Contributing! 🚀

[github-actions]

✅ PR template check passed!

@arpit2006 this PR is ready for your review. 🚀

[github-actions]

✅ PR template check passed!

@arpit2006 this PR is ready for your review. 🚀

[arpit2006]

@prasiddhi-105 , Resolve CI fails!

[arpit2006]

@prasiddhi-105 , Any Update.

[prasiddhi-105]

@prasiddhi-105 , Any Update.

Yeah on it, I'll give u an update tomorrow morning, please do consider !