feat: v3.0 — CIS PDF chunker, DOCX professional, RAG complete, progress bar, code cleanup

Browse Source

- CIS PDF chunker: segments by recommendation (54/54 complete with desc+rationale+rem), 7000 chars target, 500 overlap
- RAG: 60s timeout + retry, direct SQL fetch by recommendationNumber, concatenate all chunks
- Auto-detect embedding dimension from DDL (empty tables), auto-detect model per dimension
- DOCX: camo strip (Pillow), A4 centered cover, code blocks (Consolas), lists with hanging indent
- Compliance report: progress bar (RAG X/35 percentage), Audit/Verification bold
- Code cleanup: removed redundant imports (io, base64, time, Response, HTMLResponse), duplicate DB query
- EmbeddingsPage: auto-select ADB config fix
- ReportsPage: removed DOCX preview (docx-preview lib)
- README updated to v3.0 with all new features

...

This commit is contained in:

nogueiraguh

2026-03-31 10:44:13 -03:00

parent c81909d0a1

commit 1b3c02d10b

4 changed files with 852 additions and 284 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

852

backend/app.py

File diff suppressed because it is too large Load Diff