Overview dashboard

What it does

The Overview dashboard shows the Bigwave cycle as an interactive flow diagram with nodes for Catalog, Chat, Feedback, Datasets, Training, Adapters, and Evals. Each node displays live data (e.g. message count in Chat, pass rate in Evals). Click a node to navigate to that page. The model dropdown at the top lets you filter by model, and live data refreshes every 15 seconds.

Prerequisites

• A running stack with at least one model loaded.
• The app is open at the default landing page (Overview).

Steps

1. View the Overview page

Open the app home (/admin) or click Overview in the left rail. A full-page diagram renders with nodes arranged in a cycle:

           Catalog
             ↓
       ← Chat ←
       ↓     ↑
   Feedback  |
       ↓     |
   Datasets  |
       ↓     |
   Training  |
       ↓     |
   Adapters  |
       ↓     |
    Evals ───→

2. Read live metrics on each node

Each node displays a count or status:

• Catalog — total models available.
• Chat — # of completions in the current period.
• Feedback — # of thumbs up/down in the current period.
• Datasets — # of datasets in the workspace.
• Training — # of jobs (completed + running).
• Adapters — # of adapters (active + inactive).
• Evals — most recent run's pass rate (e.g. "87%").

Metrics update every 15 seconds from the backend.

3. Filter by model (optional)

The model dropdown at the top-right of the page lets you filter the flow to one model. Select a model; the diagram updates to show metrics only for that model's pipeline. Select "All models" to see the full workspace.

4. Click a node to navigate

Click any node to jump to that feature's page:

• Catalog → Model Catalog page.
• Chat → Chat panel opens in the right sidebar.
• Feedback → Feedback Dashboard (detailed feedback list).
• Datasets → Datasets page.
• Training → Training page.
• Adapters → Adapters page.
• Evals → Evals page.

5. Monitor progress

Watch the metrics update in real-time. A high chat count + low feedback count suggests you need more user feedback; high training count + low eval count suggests you need to validate trained adapters.

Verify

• Open the Overview page (default on app load).
• Verify the diagram renders with all 7 nodes visible.
• Click on Chat node; the chat panel opens.
• Click on Datasets node; you navigate to the Datasets page.
• Metrics update every 15 seconds; send a chat and watch the Chat node's count increase.
• Use the model dropdown to filter; the metrics are recomputed for the selected model.

Common issues

Symptom	Cause	Fix
Diagram doesn't render or shows "Error"	Backend metrics endpoint is down or returning 5xx	Check api container health; restart if needed
Node counts are all zero	No data exists yet in the workspace	Send a chat (guide 01), capture feedback (guide 03), upload a dataset (guide 04)
Metrics don't update	Auto-refresh is stalled due to network issues	Refresh the page manually
Model dropdown has no options	No models are loaded	Go to Catalog and load a model first
Clicking a node does nothing	Browser JavaScript error or route not configured	Check browser console (F12) for errors

Related

• Model catalog — browse and load models to populate the Catalog node.
• Send a chat message — generates chats visible in the Chat node.
• Capture user feedback (👍/👎) — increases the Feedback node metric.
• Upload or build a dataset — populates the Datasets node.
• Train a LoRA adapter (DPO) — increases Training and Adapters nodes.
• Run an eval against a golden set — shows pass rate in the Evals node.