Skip to content

feat: replace Docling with structured local knowledge search - #66

Merged
salim4n merged 1 commit into
devfrom
feat/structured-office-extraction
Aug 24, 2026
Merged

feat: replace Docling with structured local knowledge search#66
salim4n merged 1 commit into
devfrom
feat/structured-office-extraction

Conversation

@salim4n

@salim4n salim4n commented Aug 24, 2026

Copy link
Copy Markdown
Member

Summary

  • remove Docling from the runtime and replace it with isolated Bun extractors for PDF, DOCX, PPTX and XLSX
  • migrate knowledge embeddings to pinned Qwen3 via private TEI gRPC and add versioned ParadeDB hybrid search with BGE reranking
  • add durable extraction metadata, provenance, model-revision migrations and deployment sizing documentation

Validation

  • bun run check
    • 587 Bun unit/HTTP tests
    • 43 crawler tests
    • TypeScript, architecture, backend and Next.js production build
  • bun run test:integration
    • 153 PostgreSQL integration tests
  • changed-file secret scan and git diff --check

Safety

  • no provider canary or outbound send executed
  • no OpenAI embedding fallback retained
  • no Docling service or secret remains in the standard runtime

@salim4n
salim4n merged commit 4933ea6 into dev Aug 24, 2026
1 check passed
@salim4n
salim4n deleted the feat/structured-office-extraction branch August 24, 2026 13:01
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant