Dashboard Guide
A tour of the SourcePrep desktop dashboard — a configurable workspace of panels for monitoring your index, searching code, assembling context, and tuning the engine.
Overview
The dashboard is a modular grid of panels. Pick the panels you want, drag them where you want, and the layout persists across sessions. There is no fixed left/right split — every panel is independently closeable, resizable, and rearrangeable.
See the live SourcePrep dashboard for the full integrated workspace. The individual panels below are each rendered with the same components the dashboard uses.
Panel categories
Every panel belongs to one of four categories. Use this when choosing what to add to your layout.
- Status: what the engine is doing — Knowledge Status, Pipeline progress, Code-Graph coverage, Atlas, Audit, Knowledge Activity, Goalposts, Roadmap.
- Search: querying the index — Search, Search Results, File Tree, Trace.
- Context: what gets sent to your AI — Context Options, Context Output, Architecture, Concepts.
- Config: tuning the engine — Deep Analysis, Token Budget, Agent Ops.
Adding, moving, and resetting panels
The Panel Picker dropdown (top-right of the dashboard) is how you add and remove panels from your layout. It also exposes layout management:
- Toggle: show or hide any panel — non-essential panels are closeable and persist their hidden state.
- Refit: recompact the grid after closing panels.
- Reset: restore the default layout (with a confirmation step).
- Copy / Paste: serialize the current layout to clipboard so you can share it with a teammate or move it between machines.
Panel Picker — toggle visibility, reset, refit, and copy/paste your layout.
Key panels
Most users start with these. The full list lives in the Panel Picker.
Knowledge Status
High-level health of the index — file counts, coverage, last build, and stale-file count. The fastest way to see whether the engine has kept up with your last edits.
~/code/sourceprep
Knowledge Status — chunks indexed, embedding model, last build, and freshness at a glance.
Code-Graph Coverage
Inventory view of which files made it into the structural code graph, which are queued, which were excluded, and which need re-tracing after a recent change. Use the Untraced and Stale tabs to nudge specific files into the pipeline.
Code-Graph Coverage — manage the inventory of indexed files.
Pipeline
Visualizes the 15-stage enrichment pipeline (Sync → Enrich → Finalize). Each stage shows its status, last run, and provenance (deterministic re-use vs. fresh build). For a full breakdown of what each stage does and why, see How It Works → Graph Enrichment.
mixed — expand to inspect
mixed — expand to inspect
mixed — expand to inspect
Pipeline panel — every stage visible with live progress.
Search & Context
The Search panel runs queries against the index; results stream in, and the Context Output panel renders the assembled context as it would be sent to your AI. Pair them with Context Options to control budget, atlas routing, and trace expansion.
Search panel — find code by meaning, with the assembled context preview alongside.
Agent Operations
Cross-cutting view of MCP-aware agents using your index — which editors are connected, which projects they're scoped to, and how many tool calls they've issued. Useful when more than one agent is in flight at the same time.
Engines
Paperclip Integration
Install the SourcePrep skill so Paperclip agents can use structural codebase intelligence tools (prep, prep_search, prep_impact).
Push Settings
Agent Operations Panel — live connections, scope, and throughput per agent.
Roadmap
LLM-synthesized roadmap derived from your codebase, open tickets, and recent commits. Bi-directional GitHub sync keeps it in step with the work that's actually moving.
Questions for You (1)
Should federated search merge results by relevance score or by project priority?
Federated search across multiple projects requires a merge strategy.
✨ Your answer will shape the next AI Propose cycle
North Star
Implement MCP streaming responses
Strategic Overview
SourcePrep is an epistemic intelligence engine for autonomous agents.
Sprint Intelligence
Based on average velocity of 3.2 nodes/sprint.
Sprint 6
Roadmap Panel — synthesized milestones, sprint suggestions, and burndown.
Activity Heatmap
Year-view of indexing and enrichment activity per day. Surface dormant areas of the codebase and spot anomalies in incremental-rebuild frequency at a glance.
Knowledge Activity
Less
More
Activity Heatmap — daily indexing + enrichment volume across the year.
Settings
The Settings panel (Config category) is the configuration surface for the project. It is organized into sub-pages: Pipeline defaults, Chunking and Embeddings, Source globs, Trace, Integrations, and Destructive Actions. Most settings autosave; look for the inline status indicator next to each field.
What it looks like with an agent attached
With the dashboard running and an MCP-aware editor connected, your agent operates against the same panels you see — using SourcePrep's context to navigate the codebase while you watch progress in real time.
An agent live-editing UI code with SourcePrep's structural context loaded alongside the IDE.
Learn more
- Managing projects — adding repos, scope control, per-project settings.
- Graph Enrichment — what the 15 pipeline stages do.
- Code Graph — the structural backbone the dashboard surfaces.