@hybridlabor-api/aos 4.3.0 → 4.4.1

package/mcps/tdmcp/docs/superpowers/plans/2026-06-06-body-bubbles.md

@@ -1,138 +0,0 @@
-# Body Bubbles Implementation Plan
-
-> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
-
-**Goal:** Build a TouchDesigner generator that creates a MediaPipe-ready bubble installation: open palm emits light bubbles, bubbles live for 30 seconds, collide with the screen bounds/floor, and can be pushed/lifted by tracked body points.
-
-**Architecture:** Add one Layer 1 tool, `create_body_bubbles`, using the existing `runBuild`/`createSystemContainer`/`finalize` pattern. The tool creates a self-contained COMP with optional hand/body CHOP inputs, a Script CHOP that simulates bubble physics, a Script SOP + Geometry/Render TOP path for visible circles, and a control panel for live tuning.
-
-**Tech Stack:** TypeScript, Zod schemas, tdmcp Layer 1 orchestration, TouchDesigner Script CHOP/SOP Python, MSW/Vitest offline tests, existing CLI registry.
-
----
-
-### Task 1: Offline Unit Test
-
-**Files:**
-- Create: `tests/unit/createBodyBubbles.test.ts`
-
-- [ ] **Step 1: Write a failing test**
-
-Create a Vitest/MSW test that imports `createBodyBubblesImpl` and `createBodyBubblesSchema`, captures `/api/nodes` and `/api/exec`, then asserts:
-
-```ts
-await createBodyBubblesImpl(makeCtx(), {
-  name: "body_bubbles",
-  parent_path: "/project1",
-  hand_chop_path: "/project1/mp_hand_adapter/hand",
-  body_chop_path: "/project1/mp_body_adapter/pose",
-  bubble_count: 96,
-  lifetime_seconds: 30,
-  emit_on_open_palm: true,
-  floor_bounce: 0.22,
-  wall_bounce: 0.82,
-  gravity: 0.06,
-  body_radius: 0.08,
-  hand_emit_rate: 18,
-  output_resolution: [1280, 720],
-  expose_controls: true,
-});
-```
-
-Expected captured behavior:
-- creates a `scriptCHOP` named `bubble_sim`
-- creates a `scriptSOP` named `bubble_sop`
-- creates a `geometryCOMP`, `renderTOP`, and `nullTOP` named `out1`
-- writes Python containing `LIFETIME = 30.0`, `open_palm`, `floor_y`, and `body_radius`
-- exposes controls named `EmitRate`, `Gravity`, `BodyRadius`, and `Lifetime`
-
-- [ ] **Step 2: Run the test and confirm RED**
-
-Run: `rtk npx vitest run tests/unit/createBodyBubbles.test.ts`
-
-Expected: fail because `src/tools/layer1/createBodyBubbles.js` does not exist yet.
-
-### Task 2: Layer 1 Tool
-
-**Files:**
-- Create: `src/tools/layer1/createBodyBubbles.ts`
-
-- [ ] **Step 1: Implement schema and builder**
-
-Export `createBodyBubblesSchema`, `createBodyBubblesImpl`, and `registerCreateBodyBubbles`.
-
-- [ ] **Step 2: Build the TD network**
-
-Use `createSystemContainer(ctx, args.parent_path, args.name)` and create:
-- `scriptCHOP bubble_sim`
-- `scriptSOP bubble_sop`
-- `geometryCOMP bubbles_geo`
-- `constantMAT bubble_mat`
-- `cameraCOMP cam`
-- `lightCOMP light`
-- `renderTOP render`
-- `nullTOP out1`
-
-- [ ] **Step 3: Install Python callbacks**
-
-Use `builder.python` to write:
-- `bubble_sim_cb` callback DAT: owns particle state, detects palm openness from landmarks 0/4/8/12/16/20, emits near the palm center, integrates velocity/position, applies floor/wall collisions, and kills/recycles bubbles at `lifetime_seconds`.
-- `bubble_sop_cb` callback DAT: reads `bubble_sim` channels and emits one circular polyline per live bubble.
-
-- [ ] **Step 4: Expose controls**
-
-Add controls bound to stable custom parameters or node parameters:
-- `EmitRate`
-- `Gravity`
-- `BodyRadius`
-- `Lifetime`
-- `WallBounce`
-- `FloorBounce`
-
-### Task 3: Registry and CLI Wiring
-
-**Files:**
-- Modify: `src/tools/layer1/index.ts`
-- Modify: `src/cli/agent.ts`
-
-- [ ] **Step 1: Register the tool**
-
-Import `registerCreateBodyBubbles` and append it to `layer1Registrars`.
-
-- [ ] **Step 2: Add CLI command**
-
-Import `createBodyBubblesImpl/createBodyBubblesSchema` and add a command entry:
-
-```ts
-{
-  command: "create_body_bubbles",
-  schema: createBodyBubblesSchema,
-  handler: createBodyBubblesImpl,
-  summary: "Create a MediaPipe body-interactive bubble physics installation.",
-}
-```
-
-Use the exact local command registry shape discovered in `src/cli/agent.ts`.
-
-### Task 4: Verification
-
-- [ ] **Step 1: Run focused test**
-
-Run: `rtk npx vitest run tests/unit/createBodyBubbles.test.ts`
-
-Expected: pass.
-
-- [ ] **Step 2: Run typecheck**
-
-Run: `rtk npm run typecheck`
-
-Expected: pass.
-
-- [ ] **Step 3: Run build**
-
-Run: `rtk npm run build`
-
-Expected: pass.
-
-- [ ] **Step 4: Report live TD status**
-
-If the local TouchDesigner bridge is unavailable, report that offline gates passed and live bridge validation remains pending.

package/mcps/tdmcp/docs/superpowers/plans/2026-06-08-deep-doppelganger-implementation.md

@@ -1,201 +0,0 @@
-# Deep Doppelganger Implementation Plan
-
-> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
-
-**Goal:** Build the live `deep_doppelganger` TouchDesigner project in the running TD session.
-
-**Architecture:** Use existing tdmcp-agent builders for MediaPipe tracking and major visual systems, then run one controlled TD Python integration pass to create the final compositor, controls, and `out_video`. Keep the project arranged in clear zones and validate through the live bridge.
-
-**Tech Stack:** TouchDesigner 2025.32820, tdmcp bridge 0.6.1, tdmcp-agent, MediaPipe TouchDesigner package, TD Python via `exec python`.
-
----
-
-## Task 1: Runtime Preflight
-
-**Files:**
-- Read: `docs/superpowers/specs/2026-06-08-deep-doppelganger-design.md`
-- Live TD: `/project1`
-
-- [ ] **Step 1: Confirm bridge health**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js info --output json --no-color
-```
-
-Expected: `connected: true`, endpoint `http://127.0.0.1:9980`, `bridge_stale: false`.
-
-- [ ] **Step 2: Confirm MediaPipe package path**
-
-Run:
-
-```bash
-rtk proxy sh -lc 'find "$HOME/.tdmcp/packages" "$HOME/tdmcp-packages" -maxdepth 5 -iname "MediaPipe.tox" 2>/dev/null | head -20'
-```
-
-Expected: at least `/Users/pantani/.tdmcp/packages/installed/mediapipe-touchdesigner/release/toxes/MediaPipe.tox`.
-
-## Task 2: Build Tracking Sources
-
-**Files:**
-- Live TD: `/project1/MediaPipe`, `/project1/mp_hand_adapter`, body tracking outputs
-
-- [ ] **Step 1: Build body tracking**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js body-tracking --params '{"parent_path":"/project1","build_skeleton":true}' --output json --no-color
-```
-
-Expected: MediaPipe engine is loaded, a body tracking adapter/skeleton is created, and the output includes the created paths.
-
-- [ ] **Step 2: Build hand tracking**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js hand-tracking --params '{"parent_path":"/project1","max_hands":2,"coordinate_space":"world","adapter_name":"mp_hand_adapter"}' --output json --no-color
-```
-
-Expected: hand landmark CHOP output exists and can be read by the integration pass.
-
-## Task 3: Build Visual Entities
-
-**Files:**
-- Live TD: `/project1/deep_doppelganger`
-
-- [ ] **Step 1: Create the root container**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js container --params '{"parent_path":"/project1","name":"deep_doppelganger"}' --output json --no-color
-```
-
-Expected: `/project1/deep_doppelganger` exists.
-
-- [ ] **Step 2: Build optical flow from the camera**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js optical-flow --params '{"name":"motion_flow","parent_path":"/project1/deep_doppelganger","source":"/project1/MediaPipe/video","resolution":[960,540],"sensitivity":5.5,"smoothing":0.72,"blur":2,"direction_from":"edges"}' --output json --no-color
-```
-
-Expected: a motion-flow container exists and returns an output TOP path.
-
-- [ ] **Step 3: Build spectral point cloud**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js point-cloud --params '{"source":"existing","existing":"/project1/MediaPipe/video","resolution":192,"depth_scale":1.65,"point_size":0.018,"rotate":8,"expose_controls":true,"parent_path":"/project1/deep_doppelganger"}' --output json --no-color
-```
-
-Expected: a point-cloud container exists and returns an output TOP path.
-
-- [ ] **Step 4: Build feedback tunnel from the camera**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js feedback-tunnel --params '{"name":"feedback_tunnel","parent_path":"/project1/deep_doppelganger","source":"/project1/MediaPipe/video","zoom":1.035,"rotate":1.6,"hue_shift":0.012,"decay":0.93,"resolution":[1280,720]}' --output json --no-color
-```
-
-Expected: a feedback tunnel container exists and returns an output TOP path.
-
-## Task 4: Integrate Final Composite And Controls
-
-**Files:**
-- Create: `_workspace/deep_doppelganger/integrate_deep_doppelganger.py`
-- Live TD: `/project1/deep_doppelganger/out_video`
-
-- [ ] **Step 1: Create the integration script**
-
-Write `_workspace/deep_doppelganger/integrate_deep_doppelganger.py` with TD Python that:
-
-- Creates or reuses `/project1/deep_doppelganger`.
-- Creates select TOPs for camera, motion flow, point cloud, and feedback tunnel.
-- Builds `ghost_delay`, `flow_glitch`, `layer_mixer`, and `out_video`.
-- Adds custom controls: `Shock`, `Ghostdelay`, `Depthscale`, `Tunnelopen`, `Glitchbite`, `Trackingpanic`, `Cameraopacity`, `Blackout`.
-- Arranges nodes in non-overlapping zones.
-- Prints a JSON report with `out_video`, created nodes, warnings, and missing sources.
-
-- [ ] **Step 2: Generate the exec payload**
-
-Create the JSON parameters file consumed by `exec python`:
-
-```bash
-rtk node -e 'const fs=require("fs"); const script=fs.readFileSync("_workspace/deep_doppelganger/integrate_deep_doppelganger.py","utf8"); fs.writeFileSync("_workspace/deep_doppelganger/integrate_payload.json", JSON.stringify({script, return_output:true}, null, 2));'
-```
-
-Expected: `_workspace/deep_doppelganger/integrate_payload.json` exists and
-contains a non-empty `script` field plus `return_output: true`.
-
-- [ ] **Step 3: Execute the integration script**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js exec python --allow-unsafe --params-file _workspace/deep_doppelganger/integrate_payload.json --output json --no-color
-```
-
-Expected: `/project1/deep_doppelganger/out_video` exists at `1280x720` in this
-Non-Commercial TD session and the report says `ok: true`.
-
-## Task 5: Output And Validation
-
-**Files:**
-- Live TD: `/project1/deep_doppelganger/out_video`
-
-- [ ] **Step 1: Create output window**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js output --params '{"source_path":"/project1/deep_doppelganger/out_video","output_type":"window","resolution":"720p","parent_path":"/project1/deep_doppelganger"}' --output json --no-color
-```
-
-Expected: a Window COMP/output wrapper exists.
-
-- [ ] **Step 2: Arrange the network**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js arrange --params '{"path":"/project1/deep_doppelganger","recursive":true}' --output json --no-color
-```
-
-Expected: nodes are laid out in readable zones.
-
-- [ ] **Step 3: Validate errors**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js nodes errors --params '{"path":"/project1/deep_doppelganger","recursive":true}' --output json --no-color
-```
-
-Expected: no fatal errors on the final output path. Camera or MediaPipe permission warnings are reported honestly if present.
-
-- [ ] **Step 4: Capture preview**
-
-Run:
-
-```bash
-rtk node dist/cli/agent.js preview /project1/deep_doppelganger/out_video --out _workspace/deep_doppelganger/out_video.png --no-color
-```
-
-Expected: `_workspace/deep_doppelganger/out_video.png` is written and is not blank.
-
-Note: on a licensed TouchDesigner system, the same graph can be promoted to
-`1920x1080`; the validated local target is `1280x720` because TouchDesigner
-Non-Commercial caps TOP output resolution.
-
-## Self-Review
-
-- Spec coverage: bridge, tracking, visual entities, final output, fallback behavior, and layout validation are covered.
-- Red-flag scan: no vague implementation steps are used; every command is concrete.
-- Type consistency: command names and parameter names match schemas inspected from the local tdmcp-agent command catalog.

package/mcps/tdmcp/docs/superpowers/plans/2026-06-10-ai-party-ollama-integration.md

@@ -1,55 +0,0 @@
-# AI Party Ollama Integration Implementation Plan
-
-> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
-
-**Goal:** Make the AI Party POC easy to run with a local Ollama ShowIntent model while keeping policy enforcement outside the model.
-
-**Architecture:** Add an isolated ShowIntent Ollama adapter that converts operator text into a Hermes-style candidate, then feed it through the existing `runAiPartyGateway` path. Add a setup helper that checks/starts Ollama, inspects the configured model, and reports the exact commands needed to pull or package it. The existing deterministic fallback remains available and all LLM output is still schema- and policy-checked.
-
-**Tech Stack:** TypeScript, Node fetch/spawn, existing `ShowIntentSchema`, `AiPartyGatewaySchema`, Ollama `/api/chat`, existing Vitest harness.
-
----
-
-### Task 1: ShowIntent Ollama Adapter
-
-**Files:**
-- Create: `src/automation/showIntentOllama.ts`
-- Test: `tests/unit/showIntentOllama.test.ts`
-
-- [x] Write a failing test proving valid Ollama JSON becomes a Hermes candidate.
-- [x] Write a failing test proving malformed/unsafe LLM JSON fails closed.
-- [x] Implement `runShowIntentOllama(input, deps)` with injected fetch for tests.
-- [x] Verify with `npm test -- tests/unit/showIntentOllama.test.ts`.
-
-### Task 2: CLI Integration
-
-**Files:**
-- Modify: `src/cli/agent.ts`
-- Test: `tests/unit/aiPartyGateway.test.ts`
-
-- [x] Write a failing test for `tdmcp-agent ai-party --llm`.
-- [x] Write a failing test for `tdmcp-agent ai-party llm-setup`.
-- [x] Add `--llm` parsing for the `ai-party` command only.
-- [x] Route `--llm` through the Ollama adapter, then through `runAiPartyGateway`.
-- [x] Add `llm-setup` as a dry-run setup/status command.
-- [x] Verify with focused CLI tests.
-
-### Task 3: Docs And Commands
-
-**Files:**
-- Modify: `docs/reference/cli.md`
-- Modify: `docs/reference/environment.md`
-- Optional: `docs/AI_PARTY_LLM_TRAINING_PLAN.md`
-
-- [x] Document `tdmcp-agent ai-party --llm`.
-- [x] Document `tdmcp-agent ai-party llm-setup`.
-- [x] Keep clear that `tdmcp chat` stays general-purpose and this model is ShowIntent-only.
-
-### Task 4: Verification
-
-**Commands:**
-- [x] `npm run typecheck`
-- [x] `npm run lint`
-- [x] `npm test -- tests/unit/showIntentOllama.test.ts tests/unit/aiPartyGateway.test.ts tests/unit/showIntentTraining.test.ts`
-- [x] `npm test`
-- [x] `npm run docs:build`

package/mcps/tdmcp/docs/superpowers/plans/2026-06-24-kinect-wall-harp-calibration-wizard.md

@@ -1,53 +0,0 @@
-# Kinect Wall Harp Calibration Wizard Implementation Plan
-
-> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
-
-**Goal:** Add an on-projection calibration wizard for the Kinect wall harp so the performer can calibrate without chat-timed sampling.
-
-**Architecture:** The existing `hand_tracker` will expose both mapped and raw Kinect coordinates. The existing `strings_visual` Script TOP will own the calibration state machine, draw target overlays, capture stable raw hand points, write calibration parameters on the parent COMP, and pause harp triggering during calibration.
-
-**Tech Stack:** TypeScript tool generator, embedded TouchDesigner Python callbacks, Vitest/MSW unit tests, live TouchDesigner bridge on port `9980`.
-
----
-
-## Task 1: Add Generator Contract
-
-**Files:**
-- Modify: `src/tools/layer1/createKinectWallHarp.ts`
-- Modify: `tests/unit/createKinectWallHarp.test.ts`
-
-- [ ] Add `calibration_hold_ms` to the Zod schema and payload interface with default `900`.
-- [ ] Expose `Calibrationmode`, `Manualcapture`, `Resetcalibration`, and `Calibrationholdms` custom parameters.
-- [ ] Assert the generated Python contains the calibration parameters and storage keys.
-
-## Task 2: Preserve Raw Hand Coordinates
-
-**Files:**
-- Modify: `src/tools/layer1/createKinectWallHarp.ts`
-
-- [ ] Extend `HAND_CHOP_CODE` hand tuples to include `raw_x`, `raw_y`, `cal_x`, and `cal_y`.
-- [ ] Store those channels in `tdmcp_hands_latest`.
-- [ ] Keep mapped `x/y` unchanged for the harp logic.
-
-## Task 3: Add Visual Calibration Wizard
-
-**Files:**
-- Modify: `src/tools/layer1/createKinectWallHarp.ts`
-
-- [ ] In `VISUAL_TOP_CODE`, pause harp/audio drive while `Calibrationmode` is enabled.
-- [ ] Draw four projected targets and a progress bar.
-- [ ] Auto-capture after the configured hold time when the raw hand point is stable.
-- [ ] Capture immediately when `Manualcapture` is toggled.
-- [ ] Reset state when `Resetcalibration` is toggled.
-- [ ] Apply `Inputmirrorx`, `Inputleft`, `Inputright`, `Inputtop`, and `Inputbottom` from the four captured raw points.
-
-## Task 4: Verify and Apply Live
-
-**Files:**
-- Modify: live TouchDesigner project `/project1/kinect_wall_harp`
-
-- [ ] Run focused Vitest.
-- [ ] Run Biome on touched source/test files.
-- [ ] Run build.
-- [ ] Rebuild/apply the harp in TouchDesigner with OSC Kinect mode and UMC202HD audio settings.
-- [ ] Confirm `/api/network/project1/errors` is empty.