Dashboard

Centralized management for all ContextManager features — 5 tabs, one WebView.

Opening the Dashboard

Click the status bar item (shows active project name)
Run command: ContextManager: Open Dashboard
Click the 📖 icon in the Activity Bar → project → Dashboard

Tab Overview

Tab	What’s Inside
🧠 Intelligence	Auto-Capture/Auto-Learn controls, observations feed, AI Distill
Knowledge	All cards with search, filter, folders, select/deselect (drives hook injection), inline editing, card health, card queue
Sessions	Tracked chat sessions, pending capture counts, and bind/rebind controls for multi-project routing
Context	Project goals, conventions, tool hints, working notes, key files
⚙ Settings	Core extension settings, hook controls, and data import/export

Intelligence Tab

The command center for the intelligence pipeline:

Controls

Auto-Capture ON/OFF toggle with observation count
Auto-Learn ON/OFF toggle
🤖 Distill Observations button — extract conventions, tool hints, and working notes from observations

Observations Feed

Per-source filter pills: dynamically generated from observation sources
Each observation shows timestamp, participant source, and prompt preview
Type badges (bugfix, feature, change, etc.) shown as summary counts in the header
Actions: Delete (✕), Distill with AI (🧠)

Custom Workflows

+ New Workflow — define name, trigger, prompt template, output mode, target card, and max items
Template variable buttons — grouped by category (Queue, Card, Project, Collections, Event) for easy insertion into prompts
AI and Template modes — choose whether the rendered template goes through the model or is saved directly
Trigger badges — color-coded: blue (auto-queue), gray (manual), purple (event triggers like convention-learned, card-created, card-updated, observation-created)
▶ Run — manually execute any workflow
Enable/Disable toggle, Edit, Delete actions
See Custom Workflows for full documentation

Sessions Tab

The session-routing control plane for multi-project setups:

Tracked sessions — each Copilot chat session appears with origin, last activity, and a first-prompt label
Pending capture counts — see how many hook events are waiting for assignment because no project was explicitly bound yet
Bind & Import Pending — attach an unbound session to a project and immediately materialize queued captures into that project’s memory
Rebind From Now — switch future captures to a different project without rewriting past captures
Dismiss / Delete — hide noisy sessions or permanently remove stale, unbound routing state
Bulk operations — select multiple sessions with checkboxes and Dismiss or Delete them in one action

Single-project setups usually do not need this tab, because tools and hook capture can resolve the only project automatically. See Session Routing for the full walkthrough.

Knowledge Tab

Card Management

Search bar - keyword filtering across all cards
Category filter - architecture, pattern, convention, explanation, note, all
Folder tree - collapsible hierarchy with drag-and-drop
Card actions - Select, Edit, Delete, Refine, 📌 Pin, 👁 Include, 🗃 Archive, 🌐 Global
Generate with AI - enter a topic and the LLM generates a structured card using the Language Model API directly
Card checkboxes drive hook injection - cards checked here are automatically included in the session context injected into every prompt via the agent hook system

AI Synthesis with Custom Prompt

Select items from the card queue and click ✨ AI Synthesize
The editor opens with an optional custom prompt textarea
Enter specific instructions (e.g., “Focus on security”) or leave blank for defaults
Click ✨ AI Draft to generate — errors now show descriptive reasons instead of generic failures

Inline Editing

Click Edit on any card to expand the editor
Auto-sizing textarea (min 300px, max 80vh)
Scroll position preserved during edit
Rendered markdown preview with LaTeX support

Card Health

Collapsible analytics section
Total / selected / stale / never-used counts
Top 5 most-used cards
Duplicate detection (Jaccard ≥40%)

Card Queue

Pending count, Distill into Cards, Clear Queue
Check items to include in distill, or save individual items directly as cards
Tags are auto-generated by the LLM during distill and propagated through approval
See Card Queue for details

Replace, Delete, Refine with AI, Create Card from Selection, Ask Question

Context Tab

Project Metadata

Project name, goals, key files
Auto-saves on edit (800ms debounce)

🏗 Conventions

Codebase conventions learned by the agent
Enable/disable toggle per convention
Edit, delete actions
Injected into all AI prompts when enabled

🔧 Tool Hints

Search patterns that work (and don’t work) in this codebase
Checkbox selection for prompt injection
If none selected, top 5 by recency are injected by default

📝 Working Notes

Insights discovered during codebase exploration
Expandable details with staleness badges (🟢 fresh, ⚠️ possibly-stale, 🔴 stale)
Matched by file path for task-relevant injection
Actions: Mark Fresh, Discard, Delete

Settings Tab

The most commonly used extension settings are editable directly in the dashboard, alongside hook controls and data management.

Dedicated AI model selectors — choose separate model families for background extraction, AI workflows, and AI Draft / Synthesize Card

Data Management

📦 Export All Data / Import Data - full dump/restore
Export Current Project / Import Project - single-project with duplicate handling

Keyboard Shortcuts

Shortcut	Action
`1`	Intelligence
`2`	Knowledge
`3`	Sessions
`4`	Context
`5`	Settings
`Ctrl+K`	Focus search
`Ctrl+Shift+K`	Deselect all cards
`Ctrl+N`	New card

Rendering

Markdown - full rendering with headers, lists, code blocks, blockquotes, tables
LaTeX - $...$ inline, $$...$$ display, KaTeX with MathML output
Mermaid - ` ```mermaid ` blocks displayed as styled code blocks with syntax highlighting
Code syntax - language-tagged code blocks with VS Code theme colors

Next Steps

Language Model Tools →