docs: enrich Tech stack table and align README with code state - #6
Merged
Conversation
The README's Tech stack table treated GLiNER and DeBERTa zero-shot the same way as FastAPI and ruff: a one-line role tag. Reviewers from non-ML backgrounds (and ML folks who haven't tracked span-based NER) will skim past names they do not recognise. While in the area, a few other stale facts needed catching up. README: - Replace the Tech stack `Role` column with `What it does`. Niche items (GLiNER, DeBERTa, BM25, hierarchical chunking) get one sentence that explains what they are and the differentiator. Universal tools stay terse. Presidio is explicitly identified as Microsoft's. - Coverage badge 92% -> 93% to match the project number. - Faithfulness placeholder: replace the literal `<measured>` token with `N/A (informational, requires Ollama)` in both the README expected output and the matching `scripts/eval.py` print line. Same change in the response-headers example for `X-Guardian-Faithfulness`, which is annotated as roadmap. - Companion-projects framing: surface the three axes (cost, quality, privacy) as a bullet list. AGENTS.md: - Mirror the README's companion-projects framing. - Add `retrieval/_real_backends.py` to the pyright-excluded list. It was added during the retrieval PR but the agent doc still listed only `pii/_real_backends.py` and `dashboard/app.py`. context/architecture.md: - Add `guardian_rag/retrieval/_real_backends.py` to the module map next to the pii one. - Test count and coverage: 125 -> 130 tests, ~92% -> ~93%.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Rolecolumn becomesWhat it does. Niche libraries (GLiNER, DeBERTa zero-shot, BM25, hierarchical chunking) get one sentence that explains what they are and the differentiator. Universal tools stay terse.<measured>placeholder thatscripts/eval.pywas printing and the README was quoting. Now both sayN/A (informational, requires Ollama).X-Guardian-Faithfulness:header annotated as reserved / on the roadmap.AGENTS.md.AGENTS.mdpyright-excluded list now includesretrieval/_real_backends.pyfor completeness.context/architecture.mdmodule map gets the matching row, and the test count / coverage line is brought up to date.Test plan
uv run ruff check .cleanuv run pyright0 errors / 0 warningsuv run pytest130 passed, 93.31% coverageuv run python scripts/eval.py --with-glineroutput matches the README block verbatim