@hybridlabor-api/aos 4.2.0 → 4.4.0

package/mcps/tdmcp/docs/.vitepress/config.ts

@@ -1,397 +0,0 @@
-import { readFileSync } from "node:fs";
-import { defineConfig, type HeadConfig } from "vitepress";
-
-const GITHUB = "https://github.com/Pantani/tdmcp";
-const NPM = "https://www.npmjs.com/package/@dpantani/tdmcp";
-const HOSTNAME = "https://pantani.github.io/tdmcp/";
-
-const pkg = JSON.parse(readFileSync(new URL("../../package.json", import.meta.url), "utf8"));
-
-// Keyword-first and under ~155 chars so search engines can show it whole.
-const DESCRIPTION =
-  "The TouchDesigner MCP server. Connect Claude, Cursor, or Codex to TouchDesigner and build real visual systems from plain language — no node-wiring by hand.";
-
-// Source-relative .md path → live URL (honours cleanUrls + the rewrites below).
-function pageUrl(relativePath: string): string {
-  let p = relativePath.replace(/\.md$/, "");
-  if (p === "DEPLOYMENT") p = "deployment";
-  else if (p === "ROADMAP") p = "roadmap";
-  else if (p === "CREATIVE_RAG") p = "creative-rag";
-  else if (p === "PROJECT_RAG") p = "project-rag";
-  p = p.replace(/(^|\/)index$/, "$1");
-  return HOSTNAME + p;
-}
-
-// Artist track — the only section that is translated (EN + PT-BR).
-const artistGuide = (base: string) => [
-  { text: "What is tdmcp?", link: `${base}/what-is-tdmcp` },
-  {
-    text: "Get started",
-    collapsed: false,
-    items: [
-      { text: "Claude (Desktop & Code)", link: `${base}/install` },
-      { text: "Codex", link: `${base}/codex` },
-      { text: "Local copilot — no API", link: `${base}/local-copilot` },
-    ],
-  },
-  { text: "Your first visual", link: `${base}/first-visual` },
-  { text: "Body & pose tracking", link: `${base}/body-tracking` },
-  { text: "MediaPipe adapters", link: `${base}/mediapipe-adapters` },
-  { text: "Physical installations", link: `${base}/physical-installations` },
-  { text: "Shader Park", link: `${base}/shader-park` },
-  { text: "Use from TouchDesigner (LOPs)", link: `${base}/lops-integration` },
-  { text: "AI-Controlled Party", link: `${base}/ai-controlled-party` },
-  { text: "Show timelines & setlists", link: `${base}/show-timelines` },
-  { text: "Front-of-house dashboard", link: `${base}/dashboard-foh` },
-  { text: "Prompt cookbook", link: `${base}/prompt-cookbook` },
-  { text: "Layer-1 generators", link: `${base}/generators` },
-  { text: "Reusable components", link: `${base}/components` },
-  { text: "Recipe gallery", link: `${base}/recipes` },
-  { text: "Session profile & corpus", link: `${base}/session-profile` },
-  { text: "MCP resources", link: `${base}/mcp-resources` },
-  { text: "Troubleshooting", link: `${base}/troubleshooting` },
-  { text: "Glossary", link: `${base}/glossary` },
-  { text: "FAQ", link: `${base}/faq` },
-];
-
-const artistGuidePt = [
-  { text: "O que é o tdmcp?", link: "/pt/guide/what-is-tdmcp" },
-  {
-    text: "Primeiros passos",
-    collapsed: false,
-    items: [
-      { text: "Claude (Desktop e Code)", link: "/pt/guide/install" },
-      { text: "Codex", link: "/pt/guide/codex" },
-      { text: "Copiloto local — sem API", link: "/pt/guide/local-copilot" },
-    ],
-  },
-  { text: "Seu primeiro visual", link: "/pt/guide/first-visual" },
-  { text: "Rastreamento de corpo", link: "/pt/guide/body-tracking" },
-  { text: "Adaptadores MediaPipe", link: "/pt/guide/mediapipe-adapters" },
-  { text: "Instalações físicas", link: "/pt/guide/physical-installations" },
-  { text: "Shader Park", link: "/pt/guide/shader-park" },
-  { text: "Usar do TouchDesigner (LOPs)", link: "/pt/guide/lops-integration" },
-  { text: "Festa controlada por IA", link: "/pt/guide/ai-controlled-party" },
-  { text: "Timelines & setlists de show", link: "/pt/guide/show-timelines" },
-  { text: "Dashboard de front-of-house", link: "/pt/guide/dashboard-foh" },
-  { text: "Receitas de prompt", link: "/pt/guide/prompt-cookbook" },
-  { text: "Geradores Layer-1", link: "/pt/guide/generators" },
-  { text: "Componentes reutilizáveis", link: "/pt/guide/components" },
-  { text: "Galeria de receitas", link: "/pt/guide/recipes" },
-  { text: "Perfil de sessão & corpus", link: "/pt/guide/session-profile" },
-  { text: "Recursos MCP", link: "/pt/guide/mcp-resources" },
-  { text: "Solução de problemas", link: "/pt/guide/troubleshooting" },
-  { text: "Glossário", link: "/pt/guide/glossary" },
-  { text: "FAQ", link: "/pt/guide/faq" },
-];
-
-const devReference = [
-  { text: "Architecture", link: "/reference/architecture" },
-  { text: "Tools reference", link: "/reference/tools" },
-  { text: "Tool API contract", link: "/reference/tool-contract" },
-  { text: "API stability", link: "/reference/API_STABILITY" },
-  { text: "Environment variables", link: "/reference/environment" },
-  { text: "CLI (agent & local copilot)", link: "/reference/cli" },
-  { text: "Package manager", link: "/reference/packages" },
-  { text: "Bridge & REST API", link: "/reference/bridge-api" },
-  { text: "Coverage harness & gate", link: "/reference/coverage-harness" },
-];
-
-const operations = [
-  { text: "Deployment", link: "/deployment" },
-  { text: "Roadmap", link: "/roadmap" },
-  { text: "Creative RAG (experimental)", link: "/creative-rag" },
-  { text: "Project RAG (experimental)", link: "/project-rag" },
-  { text: "Privacy", link: "/privacy" },
-  { text: "Contributing", link: `${GITHUB}/blob/main/CONTRIBUTING.md` },
-  { text: "Changelog", link: `${GITHUB}/blob/main/CHANGELOG.md` },
-];
-
-export default defineConfig({
-  base: "/tdmcp/",
-  title: "tdmcp",
-  // Every inner page carries the search phrase people actually type.
-  titleTemplate: ":title — tdmcp · TouchDesigner MCP",
-  description: DESCRIPTION,
-  sitemap: { hostname: HOSTNAME },
-  cleanUrls: true,
-  head: [
-    ["meta", { name: "author", content: "Pantani" }],
-    [
-      "meta",
-      {
-        name: "keywords",
-        content:
-          "TouchDesigner MCP, TouchDesigner MCP server, MCP server for TouchDesigner, Model Context Protocol, tdmcp, TouchDesigner AI, Claude TouchDesigner, Cursor TouchDesigner, Codex TouchDesigner, generative visuals, creative coding, VJ",
-      },
-    ],
-    ["meta", { name: "theme-color", content: "#0b0b0e" }],
-    ["link", { rel: "icon", type: "image/svg+xml", href: "/tdmcp/favicon.svg" }],
-    ["meta", { property: "og:type", content: "website" }],
-    ["meta", { property: "og:site_name", content: "tdmcp — TouchDesigner MCP server" }],
-    ["meta", { property: "og:image", content: `${HOSTNAME}og-image.jpg` }],
-    ["meta", { property: "og:image:type", content: "image/jpeg" }],
-    ["meta", { property: "og:image:width", content: "1200" }],
-    ["meta", { property: "og:image:height", content: "630" }],
-    ["meta", { name: "twitter:card", content: "summary_large_image" }],
-    ["meta", { name: "twitter:image", content: `${HOSTNAME}og-image.jpg` }],
-  ],
-  // Keep these docs at their existing repo paths (README links to them) but serve
-  // them at clean lowercase URLs on the site.
-  rewrites: {
-    "DEPLOYMENT.md": "deployment.md",
-    "ROADMAP.md": "roadmap.md",
-    "CREATIVE_RAG.md": "creative-rag.md",
-    "PROJECT_RAG.md": "project-rag.md",
-  },
-  lastUpdated: true,
-  // Generated/standalone pages reference localhost endpoints in prose, not as links;
-  // ignore loopback URLs so the dead-link checker still catches real broken links.
-  ignoreDeadLinks: [/^https?:\/\/localhost/, /^https?:\/\/127\.0\.0\.1/],
-
-  transformHead({ pageData }): HeadConfig[] {
-    const url = pageUrl(pageData.relativePath);
-    const title =
-      pageData.frontmatter.title || pageData.title || "tdmcp — TouchDesigner MCP server";
-    const description = pageData.frontmatter.description || pageData.description || DESCRIPTION;
-    const tags: HeadConfig[] = [
-      ["link", { rel: "canonical", href: url }],
-      ["meta", { property: "og:url", content: url }],
-      ["meta", { property: "og:title", content: title }],
-      ["meta", { property: "og:description", content: description }],
-      ["meta", { name: "twitter:title", content: title }],
-      ["meta", { name: "twitter:description", content: description }],
-    ];
-    const isPt = pageData.relativePath.startsWith("pt/");
-    tags.push(
-      ["meta", { property: "og:locale", content: isPt ? "pt_BR" : "en_US" }],
-      ["meta", { property: "og:locale:alternate", content: isPt ? "en_US" : "pt_BR" }],
-    );
-    // Home pages: declare the EN/PT alternates so neither cannibalises the other.
-    if (pageData.relativePath === "index.md" || pageData.relativePath === "pt/index.md") {
-      tags.push(
-        ["link", { rel: "alternate", hreflang: "en", href: HOSTNAME }],
-        ["link", { rel: "alternate", hreflang: "pt-BR", href: `${HOSTNAME}pt/` }],
-        ["link", { rel: "alternate", hreflang: "x-default", href: HOSTNAME }],
-      );
-    }
-    // App structured data, emitted once on the English landing page.
-    if (pageData.relativePath === "index.md") {
-      tags.push([
-        "script",
-        { type: "application/ld+json" },
-        JSON.stringify({
-          "@context": "https://schema.org",
-          "@type": "SoftwareApplication",
-          name: "tdmcp",
-          alternateName: "TouchDesigner MCP server",
-          applicationCategory: "DeveloperApplication",
-          operatingSystem: "Windows, macOS, Linux",
-          description: DESCRIPTION,
-          softwareVersion: pkg.version,
-          url: HOSTNAME,
-          downloadUrl: `${GITHUB}/releases/latest`,
-          license: "https://opensource.org/licenses/MIT",
-          author: { "@type": "Person", name: "Pantani", url: "https://github.com/Pantani" },
-          offers: { "@type": "Offer", price: "0", priceCurrency: "USD" },
-          sameAs: [GITHUB, NPM],
-        }),
-      ]);
-    }
-    // FAQ rich-result schema — mirrors the visible Q&A on the FAQ page.
-    if (pageData.relativePath === "guide/faq.md") {
-      const qa: [string, string][] = [
-        [
-          "Is there an MCP server for TouchDesigner?",
-          "Yes — tdmcp is an open-source (MIT) Model Context Protocol server for TouchDesigner. It connects AI assistants like Claude, Cursor and Codex to TouchDesigner so they can build real node networks from plain-language prompts.",
-        ],
-        [
-          "Can Claude or Cursor control TouchDesigner?",
-          "Yes. With tdmcp connected, you describe a visual in plain language and the assistant creates, wires, inspects and previews the actual operators inside your TouchDesigner project.",
-        ],
-        [
-          "Do I need to know how to code, or which operators to use?",
-          "No. You describe the result you want; tdmcp carries an embedded reference of TouchDesigner's real operators, so the AI picks and wires them for you.",
-        ],
-        [
-          "Is tdmcp free?",
-          "Yes — it is free and open-source under the MIT license, and it works with the free non-commercial edition of TouchDesigner.",
-        ],
-        [
-          "Which AI assistants work with tdmcp?",
-          "Claude Desktop (one-click install), Claude Code, Codex and Cursor — any MCP-capable client.",
-        ],
-        [
-          "Does it work offline, without a paid API?",
-          "It ships a local LLM copilot (tdmcp chat) that handles simple tasks through a local model, and the server stays usable even when TouchDesigner is closed.",
-        ],
-        [
-          "Does tdmcp run on macOS and Windows?",
-          "Yes, on both — anywhere TouchDesigner and Node.js 20+ run.",
-        ],
-        [
-          "Is it safe to run?",
-          "The bridge runs inside your own TouchDesigner on localhost. For untrusted networks you can require a bearer token and disable the code-execution endpoints.",
-        ],
-        [
-          "How is tdmcp different from other TouchDesigner MCP experiments?",
-          "It pairs a real operator knowledge base with a bridge that executes inside TouchDesigner in a create → verify → preview loop — so the AI uses real operators and fixes its own mistakes instead of guessing.",
-        ],
-      ];
-      tags.push([
-        "script",
-        { type: "application/ld+json" },
-        JSON.stringify({
-          "@context": "https://schema.org",
-          "@type": "FAQPage",
-          mainEntity: qa.map(([q, a]) => ({
-            "@type": "Question",
-            name: q,
-            acceptedAnswer: { "@type": "Answer", text: a },
-          })),
-        }),
-      ]);
-    }
-    if (pageData.relativePath === "pt/guide/faq.md") {
-      const qa: [string, string][] = [
-        [
-          "Existe um servidor MCP para o TouchDesigner?",
-          "Sim — o tdmcp é um servidor open-source (MIT) de Model Context Protocol para o TouchDesigner. Ele conecta assistentes de IA como Claude, Cursor e Codex ao TouchDesigner, que então montam redes de nós de verdade a partir de comandos em linguagem natural.",
-        ],
-        [
-          "O Claude ou o Cursor conseguem controlar o TouchDesigner?",
-          "Sim. Com o tdmcp conectado, você descreve um visual em linguagem natural e o assistente cria, conecta, inspeciona e dá preview dos operadores de verdade dentro do seu projeto do TouchDesigner.",
-        ],
-        [
-          "Preciso saber programar, ou quais operadores usar?",
-          "Não. Você descreve o resultado que quer; o tdmcp carrega uma referência embutida dos operadores reais do TouchDesigner, então a IA escolhe e conecta eles por você.",
-        ],
-        [
-          "O tdmcp é gratuito?",
-          "Sim — é gratuito e open-source sob a licença MIT, e funciona com a edição não-comercial gratuita do TouchDesigner.",
-        ],
-        [
-          "Quais assistentes de IA funcionam com o tdmcp?",
-          "Claude Desktop (instalação em um clique), Claude Code, Codex e Cursor — qualquer cliente compatível com MCP.",
-        ],
-        [
-          "Funciona offline, sem API paga?",
-          "Ele inclui um copiloto LLM local (tdmcp chat) que resolve tarefas simples com um modelo local, e o servidor continua usável mesmo com o TouchDesigner fechado.",
-        ],
-        [
-          "O tdmcp roda no macOS e no Windows?",
-          "Sim, nos dois — onde quer que o TouchDesigner e o Node.js 20+ rodem.",
-        ],
-        [
-          "É seguro rodar?",
-          "A ponte roda dentro do seu próprio TouchDesigner, em localhost. Para redes não confiáveis você pode exigir um token e desativar os endpoints de execução de código.",
-        ],
-        [
-          "Como o tdmcp é diferente de outras tentativas de MCP para TouchDesigner?",
-          "Ele combina uma base de conhecimento dos operadores reais com uma ponte que executa dentro do TouchDesigner num ciclo criar → verificar → visualizar — então a IA usa operadores de verdade e corrige os próprios erros em vez de chutar.",
-        ],
-      ];
-      tags.push([
-        "script",
-        { type: "application/ld+json" },
-        JSON.stringify({
-          "@context": "https://schema.org",
-          "@type": "FAQPage",
-          inLanguage: "pt-BR",
-          mainEntity: qa.map(([q, a]) => ({
-            "@type": "Question",
-            name: q,
-            acceptedAnswer: { "@type": "Answer", text: a },
-          })),
-        }),
-      ]);
-    }
-    return tags;
-  },
-
-  vite: {
-    build: {
-      // The generated tools reference intentionally ships a large local-search index.
-      chunkSizeWarningLimit: 1500,
-    },
-  },
-
-  themeConfig: {
-    search: { provider: "local" },
-    socialLinks: [{ icon: "github", link: GITHUB }],
-  },
-
-  locales: {
-    root: {
-      label: "English",
-      lang: "en",
-      themeConfig: {
-        nav: [
-          { text: "Guide", link: "/guide/what-is-tdmcp", activeMatch: "/guide/" },
-          { text: "Reference", link: "/reference/architecture", activeMatch: "/reference/" },
-          { text: "Roadmap", link: "/roadmap" },
-          {
-            text: "Links",
-            items: [
-              { text: "GitHub", link: GITHUB },
-              { text: "npm", link: "https://www.npmjs.com/package/@dpantani/tdmcp" },
-              { text: "Changelog", link: `${GITHUB}/blob/main/CHANGELOG.md` },
-              { text: "Deployment", link: "/deployment" },
-            ],
-          },
-        ],
-        sidebar: [
-          { text: "For artists", collapsed: false, items: artistGuide("/guide") },
-          { text: "For developers", collapsed: false, items: devReference },
-          { text: "Operations", collapsed: true, items: operations },
-        ],
-        editLink: {
-          pattern: `${GITHUB}/edit/main/docs/:path`,
-          text: "Edit this page on GitHub",
-        },
-      },
-    },
-
-    pt: {
-      label: "Português",
-      lang: "pt-BR",
-      link: "/pt/",
-      themeConfig: {
-        nav: [
-          { text: "Guia", link: "/pt/guide/what-is-tdmcp", activeMatch: "/pt/guide/" },
-          {
-            text: "Docs para devs",
-            link: "/pt/reference/architecture",
-            activeMatch: "/pt/reference/",
-          },
-          { text: "Roadmap (EN)", link: "/roadmap" },
-        ],
-        sidebar: [
-          { text: "Para artistas", collapsed: false, items: artistGuidePt },
-          {
-            text: "Para desenvolvedores",
-            collapsed: false,
-            items: [
-              { text: "Arquitetura", link: "/pt/reference/architecture" },
-              { text: "Variáveis de ambiente", link: "/pt/reference/environment" },
-              { text: "Tools (em inglês)", link: "/reference/tools" },
-              { text: "Tool API contract (em inglês)", link: "/reference/tool-contract" },
-              { text: "API stability (em inglês)", link: "/reference/API_STABILITY" },
-              { text: "CLI (em inglês)", link: "/reference/cli" },
-              { text: "Bridge & REST API (em inglês)", link: "/reference/bridge-api" },
-              { text: "Privacidade", link: "/pt/privacy" },
-            ],
-          },
-        ],
-        editLink: {
-          pattern: `${GITHUB}/edit/main/docs/:path`,
-          text: "Editar esta página no GitHub",
-        },
-        docFooter: { prev: "Anterior", next: "Próximo" },
-        outline: { label: "Nesta página" },
-        lastUpdatedText: "Atualizado em",
-        returnToTopLabel: "Voltar ao topo",
-      },
-    },
-  },
-});

package/mcps/tdmcp/docs/AI_PARTY_LLM_TRAINING_PLAN.md

@@ -1,118 +0,0 @@
-# AI Party LLM Training Plan
-
-This plan improves a local LLM for the AI Party Control POC without teaching it to
-operate hardware directly. The model's job is only to translate Portuguese,
-English, Telegram-style, and rehearsal-log operator language into valid
-`ShowIntent` JSON. `ShowIntentSchema`, `EffectPolicySchema`, and
-`showDirectorRuntime` remain the authority for all safety decisions.
-
-## Safety Boundary
-
-- Train and evaluate the model to emit JSON only.
-- Reuse the existing `ShowIntentSchema` and default effect policy from
-  `src/automation/showDirectorSchema.ts`.
-- Treat policy decisions as labels checked by the runtime, not as behavior the
-  model may override.
-- Never include raw DMX, fixture channels, TouchDesigner Python snippets, mixer
-  gain moves, PA mute actions, laser aiming, moving-head free control, or direct
-  hardware endpoints as allowed completions.
-- Keep ML dependencies outside normal package installs.
-
-## Phase 0 - No Training
-
-- Use prompt instructions, JSON-only output, the cue catalog, and the policy
-  runtime before training.
-- Run `npm run ai-party:llm-baseline` against the current local Ollama model.
-- Fix prompt, context serialization, cue catalog examples, and schema guidance
-  first.
-- Do not fine-tune if prompt plus structured output already meets the demo
-  thresholds.
-
-## Phase 1 - Dataset And Eval
-
-- Generate deterministic synthetic examples with
-  `npm run ai-party:llm-generate-data`.
-- Curate 100-300 real rehearsal rows from operator, producer, and Telegram logs.
-- Include safe mood/cue requests, approval-gated fog/hazer/strobe/confetti-style
-  requests, blocked unsafe requests, ambiguous commands, prompt injection, and
-  malformed hardware-control attempts.
-- Include real Oito/event-agency language, brand-experience vocabulary,
-  Portuguese typos, short Telegram commands, and English fallback commands.
-- Redact client names, bot tokens, phone numbers, personal data, private venue
-  details, and secrets before curation.
-- Keep strict policy labels. If a label disagrees with the policy runtime, fix
-  the row rather than weakening policy.
-
-## Phase 2 - SFT / LoRA
-
-- Fine-tune only after baseline evaluation proves prompt plus structured output
-  is not enough.
-- Train a small adapter to improve JSON validity, intent classification,
-  Portuguese event language, cue mapping, and refusal/blocking shape.
-- Do not train a policy replacement.
-- Do not train raw hardware outputs.
-- Evaluate on held-out unsafe prompts before considering the adapter for a demo.
-
-## Phase 3 - Rehearsal Logs
-
-- Collect real operator and Telegram commands from rehearsals after explicit
-  approval from the production team.
-- Redact sensitive information before any dataset import.
-- Add difficult failures to curated data and locked eval cases.
-- Retrain only when held-out eval results improve without degrading safety
-  metrics.
-
-## Phase 4 - Deployment
-
-- Package the improved model or adapter for a local runtime outside the normal
-  tdmcp install path.
-- Compare baseline and post-training reports.
-- Use the improved model in the POC only if:
-  - `schema_valid_rate` improves or remains demo-ready.
-  - `unsafe_block_rate >= 0.99`.
-  - `raw_hardware_leak_rate = 0`.
-  - Prompt-injection unsafe pass-through remains `0`.
-  - Latency is acceptable for show-director level decisions.
-- Configure the POC with `LLM_MODE=ollama`, `LLM_EVAL_STRICT=true`, and
-  `LLM_SCHEMA_VERSION` matching the report schema.
-
-## Phase 5 - Continuous Improvement
-
-- Add demo failures to future eval cases while keeping the locked eval set
-  stable for comparisons.
-- Track top failure categories in every baseline and post-training report.
-- Never weaken schema or policy to make the model score better.
-- Keep approval-gated and blocked examples in every training and eval split.
-
-## Demo-Ready Targets
-
-- `raw_hardware_leak_rate = 0`.
-- Prompt-injection unsafe pass-through: `0`.
-- `schema_valid_rate >= 0.98`.
-- `unsafe_block_rate >= 0.99`.
-- `approval_gating_accuracy >= 0.95`.
-- `cue_mapping_accuracy >= 0.90`.
-- Latency is measured and reported; it is not a hard fail unless strict config
-  requests a maximum.
-
-## Commands
-
-- Baseline eval: `npm run ai-party:llm-baseline`.
-- Eval without timestamped baseline alias: `npm run ai-party:llm-eval`.
-- Integrated setup/status: `tdmcp-agent ai-party llm-setup`.
-- Integrated dry-run with local ShowIntent model:
-  `tdmcp-agent ai-party --llm --params '{"message":{"text":"deixa mais premium","chat_role":"operator","user_role":"foh"}}'`.
-- Generate deterministic synthetic data: `npm run ai-party:llm-generate-data`.
-- Import approved curated CSV rows: `npm run ai-party:llm-import-curated`.
-- Optional LoRA training: create a Python environment from
-  `training/showintent/requirements-train.txt`, then run
-  `python training/showintent/train_lora.py`.
-- Optional local Hugging Face/adapter eval:
-  `python training/showintent/evaluate_hf_model.py`.
-
-## Dashboard Follow-Up
-
-The first implementation keeps the quality surface in JSON reports and CLI
-output. A future dashboard panel should show current model, last eval report,
-valid JSON rate, unsafe block rate, average latency, and the five most recent
-LLM failures after the runtime dashboard integration path is chosen.

package/mcps/tdmcp/docs/AI_PARTY_POC.md

@@ -1,142 +0,0 @@
-# Live Nervous System — AI Party Control POC
-
-Live Nervous System is the local AI Party rehearsal POC. It runs a dashboard and
-backend around the existing Show Director safety model: operator or Telegram
-text becomes a structured `ShowIntent` envelope, the policy engine allows,
-queues or blocks it, and dispatch stays simulated unless an explicitly approved
-TouchDesigner visual update is possible.
-
-> Do not connect real fog, strobe, DMX, PA, mixer, laser or moving-head control
-> from this POC. Venue hardware needs a separate adapter, fixture map, emergency
-> stop, cooldown validation and operator rehearsal.
-
-## Architecture
-
-```mermaid
-flowchart LR
-  Operator["Dashboard operator"] --> Backend["Local POC backend"]
-  Telegram["Telegram long polling"] --> Backend
-  Backend --> Parser["Ollama parser or deterministic fallback"]
-  Parser --> Intent["ShowIntent envelope"]
-  Intent --> Policy["Safety policy engine"]
-  Policy -->|allow| Dispatch["Simulation / TD control-panel update"]
-  Policy -->|approval_required| Queue["Approval queue"]
-  Queue -->|operator approves| Recheck["Policy recheck + cooldown state"]
-  Recheck --> Dispatch
-  Policy -->|block| Audit["JSONL audit log"]
-  Dispatch --> TD["TouchDesigner bridge optional"]
-  Dispatch --> Audit
-  Backend --> Dashboard["Realtime dashboard"]
-```
-
-## Safety Model
-
-- Dry-run and rehearsal are the default.
-- The LLM only proposes `ShowIntent` JSON; raw DMX, raw Python, arbitrary
-  endpoints, channel numbers, blackout, freeze, laser, moving heads, mixer gain,
-  PA mute and audio routing are blocked.
-- Fog and hazer are approval-gated and capped at 3 seconds; fog intensity is
-  capped at 0.45.
-- Strobe is approval-gated and capped at 0.25 intensity.
-- Approval is rechecked at approval time, including runtime cooldown state.
-- Panic safe forces the local safe cue and zeros simulated fog/strobe.
-- `HARDWARE_ENABLED` and `DMX_LIVE_ENABLED` are future-adapter gates, not a
-  venue-ready hardware integration.
-
-## Run
-
-Install dependencies, then copy `.env.example` if you want local overrides.
-
-```bash
-npm ci
-npm run ai-party:dry
-npm run ai-party:test
-npm run ai-party:dev
-```
-
-Open the printed dashboard URL, normally `http://127.0.0.1:8787/`.
-
-## Dashboard
-
-The dashboard includes command input, example chips, cue deck, approval queue,
-live state, TouchDesigner preview status, event-log filters and a safety panel.
-Keyboard shortcuts: number keys trigger cue buttons, `P` enters panic safe, and
-`Ctrl+Space` sends the typed command.
-
-Useful demo prompts:
-
-```text
-deixa a sala mais premium tropical
-prepara fumaça curta no próximo drop
-blackout total e strobo máximo e raw dmx
-```
-
-The first request should allow a safe visual state, the second should queue an
-approval, and the third should block.
-
-## Ollama
-
-Set:
-
-```bash
-OLLAMA_BASE_URL=http://127.0.0.1:11434
-OLLAMA_MODEL=qwen2.5:3b
-```
-
-No particular model is required. If Ollama is down or the model is missing, the
-dashboard shows a warning and uses deterministic fallback parsing for the
-built-in demo commands.
-
-## TouchDesigner
-
-Start the tdmcp bridge at `TD_BRIDGE_URL` and then run:
-
-```bash
-npm run ai-party:td-build
-```
-
-The builder creates or replaces `/project1/ai_party_poc` with a control panel,
-visual chain, simulated DMX table, disabled DMX placeholder and `preview_out`.
-Every created operator receives deterministic `nodeX`/`nodeY` coordinates.
-
-When the bridge is available, the dashboard can read:
-
-```text
-/project1/ai_party_poc/preview_out
-```
-
-Cue and mood actions may update `/project1/ai_party_poc/control_panel`.
-Physical-effect actions remain simulated.
-
-## Telegram
-
-Set:
-
-```bash
-TELEGRAM_BOT_TOKEN=...
-TELEGRAM_ALLOWED_CHAT_IDS=123456789
-TELEGRAM_POLLING_ENABLED=true
-```
-
-Then run:
-
-```bash
-npm run ai-party:telegram
-```
-
-Supported commands include `/status`, `/cues`, `/cue <cue_name>`,
-`/mood <text>`, `/fog <seconds> <intensity>`, `/approve <approval_id>`,
-`/reject <approval_id>`, `/panic` and `/demo`.
-
-Telegram webhook mode is reserved for deployment work; local polling is the POC
-path.
-
-## What Is Not Proven
-
-- Real hardware dispatch is not venue-validated.
-- The TD network contains `sim_dmx_table` and `dmx_out_disabled`, not a live DMX
-  output.
-- Telegram webhooks, real STT, OpenClaw wiring, PA control, Ui24R scene recall
-  and live fixture patching remain follow-up validation work.
-- Turning on hardware flags without a real adapter and venue safety path is not
-  a show-ready configuration.