Essential session storage is required to sign in. Optional analytics remain off unless you allow them.
Build a retrieval-augmented chat: ingest a folder of markdown, chunk + embed it, retrieve top-k for each question, and answer with citations. Include an eval script with at least 10 Q/A pairs.
Sensible chunking; top-k tuned; citations point right.
Retrieved text treated as data; injection-resistant prompting.
Repeatable eval script with pass/fail criteria.
Token usage measured; caching where it pays.
Use this README structure to make the work reviewable.
# RAG chat over your own docs ## Goal Build a retrieval-augmented chat: ingest a folder of markdown, chunk + embed it, retrieve top-k for each question, and answer with citations. Include an eval script with at least 10 Q/A pairs. ## Acceptance criteria - [ ] Retrieval quality: Sensible chunking; top-k tuned; citations point right. - [ ] Prompt hygiene: Retrieved text treated as data; injection-resistant prompting. - [ ] Evals: Repeatable eval script with pass/fail criteria. - [ ] Cost awareness: Token usage measured; caching where it pays. ## Verification - [ ] Document installation and run commands. - [ ] Record automated checks and their exact commands. - [ ] Add representative output, screenshots, or a short demo where useful. - [ ] Confirm failure paths and known constraints. ## Decision log ### Decision title - Context: - Choice: - Alternatives considered: - Tradeoffs: ## Evidence - Link each rubric criterion to the file, test, or artifact that demonstrates it. ## Known limitations - List what is intentionally out of scope and what should be improved next.
DevPath fetches a read-only, size-capped snapshot. Submitted code is never executed.