Repository navigation
Records Office redesign, backend hardening, and README screenshots - #3
Merged
Merged
Conversation
- Formatter no longer makes an LLM call: it passes the analyst's answer through and builds truncated citation previews (only analyst and critic hit the LLM now) - Cache ChatOllama instances in get_llm() per (model, base_url, temperature, max_tokens) - Add ?limit= to GET /api/query/history (default 50, max 200) and truncate AgentRun output summaries before persisting - Handle empty documents: files that parse to no text get status="empty" and skip FAISS indexing; UTF-8 fast path before chardet fallback Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Replace the dark glassmorphism theme with a light, subject-grounded identity: Avanam (ஆவணம், Tamil for document/record) styled as a registry. - New token system in index.css: record paper, ledger green chrome, iron ink, vermilion reserved for verification, brass for revisions - Type: Besley (display), Catamaran (UI, Latin+Tamil), IBM Plex Mono (data) - New ConfidenceSeal component: SVG stamp on answers — VERIFIED at the critic's 0.7 threshold, PROVISIONAL below it - Pipeline visualizer restyled as a routing slip with a brass revision loop - Lazy-loaded views, capped per-message agent events, tightened TS types, sidebar starts collapsed on narrow viewports, scrollable hero empty state Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Add docs/screenshots (chat hero, live pipeline with revision loop, stamped answer, document registry, query ledger) and a README screenshots section - Correct README facts: Avanam = ஆவணம் (document/record), React 19, qwen2.5:7b default model, Records Office design system - Update CLAUDE.md for the current architecture; add AGENTS.md contributor guide Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2d7eb89aee
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
- Router: only short-circuit to a canned chat reply when the query consists entirely of greeting/chat phrases — mixed queries like "Hi, summarize the uploaded contract" now route through retrieval - DocumentUpload: stop the Browse button's click from bubbling to the dropzone root, which opened a second file picker - Fix ruff findings (unused confidence_score in formatter, unused imports) and apply black to the files touched by this PR Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Takes main's hardened backend (thread-safe vector store, cosine scoring, JSON-mode LLM, retriever short-circuit, Alembic entrypoint, pytest suite) and keeps this branch's Records Office frontend redesign. Notable resolutions: - router: main's classify_intent() structure with the PR-review fix ported onto it — chat patterns are now full-match, so mixed queries like "hello summarize this doc" route to search (passes main's router tests) - query route: main's version plus this branch's ?limit= param and truncated JSON agent-run summaries; persist real pipeline duration - pipeline: restore revision_count in the terminal SSE event (the UI's "Revised ×N" badge reads it) - frontend types: merge main's AgentEventData/QueryHistoryItem with this branch's agent-name aliases; drop useApi.ts (deleted on main, unused) - docs: keep main's CLAUDE.md and add a Records Office design-system section; README keeps screenshots with main's llama3.2:3b model facts Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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
ConfidenceSealcomponent that stamps answers VERIFIED (vermilion) at the critic's 0.7 threshold or PROVISIONAL (brass) below it. The pipeline visualizer is restyled as a routing slip with a visible revision loop. Also: lazy-loaded views, capped per-message agent events, tightened TS types, and mobile fixes (collapsed sidebar default, scrollable hero).get_llm()cachesChatOllamainstances;GET /api/query/historygains?limit=; agent-run summaries are truncated before persisting; empty documents getstatus="empty"and skip FAISS indexing.docs/screenshots/, plus factual fixes (React 19,qwen2.5:7bdefault model, corrected name etymology). CLAUDE.md refreshed for the current architecture; AGENTS.md contributor guide added.Screenshots
See the README's new 📸 Screenshots section on this branch — chat hero, live pipeline mid-revision, a stamped answer with citations, the document registry, and the query ledger.
Test plan
npm run build(tsc + vite) passesnpm run lintpassesdocs/screenshots/)docker-compose up) — needs Ollama running🤖 Generated with Claude Code