Skip to content

UI tour

  • Get Started

    The fastest way to bring the stack up and learn the “happy path”.

  • Dashboard

    System status, monitoring links, storage, help, and glossary.

  • Get Started

    The fastest way to bring the stack up and learn the “happy path”.

  • Dashboard

    System status, monitoring links, storage, help, and glossary.

  • RAG

    Retrieval, graph views, indexing, reranker config, and learning studios.

  • Admin

    Secrets + integrations that power provider-backed features.

Quickstart Indexing Searching Configuration

Default dev URL

http://127.0.0.1:5173/web

How the UI talks to the backend

In dev, the UI uses relative requests like /api/search and your dev server proxies them to the backend (default http://127.0.0.1:8012).

flowchart LR
  UI["Vite UI (5173)"] --> API["/api/* (proxy)"]
  API --> BE["Backend (8012)"]
  BE --> PG["Postgres"]
  BE --> N4J["Neo4j (optional)"]

UI snapshots (current surfaces)

Chat + dataset iteration loop

Chat with dataset context and source grounding

This is the core operator loop: ask, inspect sources, adjust retrieval settings, and re-run.

RAG graph inspection

Graph explorer with entity table and relationships

The graph view is where you validate entity extraction quality and relationship coverage.

Recall gating and memory policy controls

Recall gating controls in chat settings

This panel controls when memory is indexed and when recall is injected per message.

Main tabs (what they’re for)

The UI is organized into top-level tabs. Here’s the practical meaning of each:

Tab What you do there Typical “first click”
Get Started Bring-up flow + sanity checks /web/start
Dashboard System status, monitoring, storage, help /web/dashboard?subtab=system
Chat Chat UI + chat settings /web/chat?subtab=ui
Grafana Embed Grafana dashboards/config (when enabled) /web/grafana?subtab=dashboard
Benchmark Run/inspect benchmarks /web/benchmark
RAG Core tri-brid features (retrieval/indexing/graph/reranker) /web/rag?subtab=retrieval
Eval Analysis Analyze eval runs and datasets /web/eval?subtab=analysis
Infrastructure Docker status, MCP servers, paths/stores /web/infrastructure?subtab=services
Admin Secrets and integrations /web/admin?subtab=secrets

Startup load: Dashboard → Storage is lazy-loaded

To keep first paint cheap and avoid unnecessary database IO, ragweld does not fetch storage/indexing metrics until you open the Storage subtab.

  • On initial load of /web/dashboard?subtab=system, the UI makes zero calls to /api/index/stats.
  • Open Dashboard → Storage to trigger the fetch and render per-corpus storage and indexing stats.
  • If you expected to see numbers but the panel looks empty, you are probably still on the System subtab. Click Storage (or use the panel’s Refresh control) to populate it.
Verify in your browser’s Network panel
  • Go to /web/dashboard (the System subtab).
  • Open DevTools → Network and filter for /api/index/stats.
  • You should see no requests until you click the Storage subtab.
  • This behavior is by design to reduce background load on Postgres at startup.

If something looks empty

Most RAG panels depend on a selected corpus and a completed index. If you haven’t indexed yet, start at Indexing.

The single most important UI control: corpus selection

Many screens include a corpus selector (sometimes labeled “repo”). This determines the corpus_id used for:

  • indexing
  • retrieval
  • graph views
  • per-corpus configuration and stats

If you see “wrong results”, the most common cause is simply that you’re looking at the wrong corpus.

Quick troubleshooting inside the UI

Where to check readiness
  • Dashboard → System Status
  • /api/ready (raw endpoint)
Where to find logs
  • Infrastructure → Docker (if you’re running via Docker)
  • Your terminal where ./start.sh is running
Where to verify secrets
  • Admin → Secrets
  • /api/secrets/check?...