spc

State Memory (state-memory-mcp)

This project tracks workflow state, tasks, design decisions, and blockers using state-memory-mcp with project slug "spc".

1. Priority Order

Before doing any coding or investigation:

1. manage_sessions(action: "start") — Start a tracking session for full change attribution.
2. get_analytics(action: "summary") — Run to understand current project state, active branches, and overall progress.
3. manage_tasks(action: "next") — Query prioritized runnable tasks.
4. manage_tasks(action: "find_blockers") — Identify any active blockers preventing progress.
5. manage_nodes(action: "list") — Find pending tasks, past decisions, or milestones.
6. query_graph(action: "trace") — Trace what depends on or blocks a task.

2. When to Write to the Graph

You MUST update the graph as you work:

• Starting a session: Always call manage_sessions(action: "start", agent_id: "my-agent") to track all mutations under a unique session.
• Starting a new task: Create a node with manage_nodes(action: "create", type: "task", title: "...", session_id: session_id).
• Making a design or implementation decision: Document it with manage_nodes(action: "create", type: "decision", title: "...", metadata: { "rationale": "..." }, session_id: session_id).
• Encountering a blocker: Record the blocker with manage_nodes(action: "create", type: "blocker", title: "...", session_id: session_id) and connect it using manage_edges(action: "add", type: "blocks", source_id: blocker_id, target_id: task_id, session_id: session_id).
• Adding observation notes: Atomically log notes using manage_nodes(action: "add_note", text: "...", attach_to: node_id).
• Batch updates: Bulk update tasks/nodes using manage_nodes(action: "batch_update", ids: ["..."], status: "done").
• Completing a task: Update status to done using manage_tasks(action: "complete", task_id: task_id) or manage_nodes(action: "update", id: task_id, status: "done").
• Creating/generating a new file: Create an artifact node with manage_nodes(action: "create", type: "artifact", title: "...", session_id: session_id) and connect it using manage_edges(action: "add", type: "produces", source_id: task_id, target_id: artifact_id).

3. Workflow Pattern

1. Start of session: Call manage_sessions(action: "start") to align and track work, then run get_analytics(action: "summary"), manage_tasks(action: "next"), and manage_tasks(action: "find_blockers").
2. Task decomposition: Decompose user requests into tasks and add them to the graph.
3. Execution: Mark tasks as "in_progress", document design decisions as they occur, and log blockers if you hit any obstacles.
4. Validation & Resolution: Run run_diagnostics(action: "validate") to ensure no cycles/orphans/contradictions, mark tasks as "done", document completed artifacts, and resolve blockers. Call manage_sessions(action: "end") to finalize.

4. Codebase Seeding on Initialization

If the project was just initialized or is missing high-level structure (Plans, Milestones, Decisions):

1. Inspect the Codebase: Read the README and core files to understand the roadmap and architecture.
2. Scaffold the Roadmap: Create a plan node (e.g., "Project Roadmap") and add milestone nodes representing key target phases, connecting them using part_of edges.
3. Scaffold Architecture: Create decision nodes representing core technical choices (e.g., choice of databases, frameworks) and link them to the milestones/tasks using decided_in edges.

Visual Memory (vision-memory-mcp)

This project utilizes vision-memory-mcp to cache visual states, record layout transitions, provide element grounding, and avoid repetitive LLM vision calls.

1. Mandatory Workflow & Priority

1. Orient: Call get_session_context to align your visual state context at the start of work.
2. Search: Call recall_memory (text/image search) before recreating duplicate UI state paths.
3. Ingest/Verify: ALWAYS call analyze_screenshot before querying any front-end vision models.
   • Cache Hit (is_known: true): Do NOT use vision models; read the returned description as context and use grounded_elements (selectors, coordinates) for action target selection.
   • Cache Miss (is_known: false): Query your vision model, then run analyze_screenshot with both the image and description to seed the cache.
4. Action Target Execution: Use predict_next_action to retrieve grounded_target handles (target_selector, target_coords) for deterministic UI clicks and typing.
5. Transitions: Call record_outcome after every click/type/scroll action to construct navigation paths.
6. Privacy & Cleanup: Call forget_state to purge sensitive or secret states from storage.

2. Tool Reference Summary (15 Core MCP Tools)

• analyze_screenshot: Ingest screenshot(s) (single or batch via items), lookup cache, return layout description and grounded elements.
• recall_memory: Search visual memory by description query or base64 image query (read-only).
• record_outcome: Save UI action execution outcomes, transitions, or log visual blockers (action_type: 'blocker').
• get_navigation_paths: Find path between states using BFS navigation graph.
• predict_next_action: Predict best next UI action and target coordinates based on transition success rates and AX tree grounding.
• compare_states: Compare visual states structurally (has_layout_change) or compare video recordings (video_a_id/video_b_id).
• get_session_context: Fetch aggregated visual context, recent/frequent states, transitions, cache hit ratios, token savings metrics, and server version info.
• manage_snapshot: Unified snapshot management (save, diff, export, restore) for visual checkpoints and regression detection.
• manage_visual_spec: Visual SDD design contract baseline registration (set), live verification (verify), and listing (list).
• manage_video: Unified video memory operations for ingestion (ingest), semantic search (search), and keyframe timelines (timeline).
• create_evidence_pack: Create cryptographic, multi-modal evidence pack linking video keyframes, state graph tasks, and visual proof.
• export_trajectories: Export multimodal visual transitions and joint workflow trajectories (json, llava, qwen2_vl, joint).
• undo_visual_mutation: Revert accidental state or transition edge ingestions.
• forget_state: Purge a specific state and vector embedding from storage for privacy.
• wait_for_visual_state: Poll for target visual state until present or timeout occurs.

3. Agent Permissions & Auto-Run Configuration

To allow cache query and ingestion commands to run automatically without prompting:

• Google Antigravity (~/.gemini/config/config.json): Add these rules to your "globalPermissionGrants" -> "allow" list:
  • "command(vision-memory-mcp)" (Allow running the CLI without parameters prompts)
  • "read_file(.*\\.gemini/antigravity/brain/.*)" (Allow reading captured screenshots)
  • "write_file(.*\\.gemini/antigravity/brain/.*)" (Allow saving visual states)
• VS Code / Cursor IDE (settings.json): Ensure the agent has execution permissions for command(vision-memory-mcp) and read/write access to the workspace's local .vision-memory-mcp/ cache directory.