@causyne/agent 1.10.18 → 1.10.20

package/docs/ARCHITECTURE.md

@@ -190,6 +190,26 @@ sliding window, and structural summaries. Full tool output is spilled before it
 is shortened. Material compaction is recorded in the audit log, and summary
 collapse emits a quiet terminal notice so context loss is visible.
 
+## Guided setup and creative applications
+
+`causyne setup` (also `causyne config`), `/setup` and the home setup menu share
+one TypeScript configuration transaction. Provider, endpoint, model, credentials
+and autonomy choices are reviewed before saving. Cancellation preserves the
+previous configuration. Credentials are not reused across providers or changed
+custom endpoints. Setup validates input locally; it does not claim provider
+connectivity or make a billable model request.
+
+The deferred `creative_apps` tool exposes installed-application status and native
+automation adapters for 3D and media creative applications. The file-based Blender batch
+adapter creates and manipulates 3D primitives (cubes, spheres, cylinders, planes, cameras,
+lights, materials, rendering) and saves a new `.blend` scene file without running model-supplied
+arbitrary code. Native ExtendScript adapters connect Photo Studio (image resizing,
+rotations, layer modifications, flattening to PSD/PNG/JPG/WEBP), Video Studio (sequence
+timelines, media import, markers, cuts to PRPROJ/XML), and Motion Studio (compositions,
+footage import, typography to AEP). It validates workspace paths, refuses overwrites,
+and applies timeout, output, and cancellation limits. No inbound control service or expensive
+remote protocols are opened.
+
 ## Removed boundaries
 
 No active pairing, relay, WebRTC/P2P, mobile command surface, background daemon

package/native-skills/media/creative-studio-pipeline/SKILL.md

@@ -0,0 +1,52 @@
+---
+name: creative-studio-pipeline
+description: Orchestrate end-to-end multi-stage creative production pipelines coordinating 3D modeling and rendering in Blender, Photo Studio grading and graphic compositing, and Video Studio sequence assembly. Use when a task spans multiple creative domains or requires chaining 3D outputs into video edits.
+---
+
+# Creative Studio Pipeline Orchestration
+
+Multi-stage creative production is most reliable when coordinated through a single declarative pipeline. Rather than switching manual applications or invoking separate tools out of sequence, Causyne Agent provides `creative_apps` with `action: "studio_pipeline"`.
+
+## Workflow
+
+1. Design the production stages.
+   - Stage A (3D Scene / Asset Generation): Blender batch (`3d_batch`) creates geometry, assigns PBR materials, positions lights/camera, and renders high-resolution stills or asset elements.
+   - Stage B (Still Grading / Composition): Photo Studio batch (`studio_batch`, host `photo-studio`) adjusts color grading, adds typographic overlays or branding, and exports optimized graphic assets.
+   - Stage C (Timeline Editing): Video Studio batch (`studio_batch`, host `video-studio`) constructs the sequence timeline, imports rendered stills/clips, cuts intervals, and saves the final project.
+   - **Done when:** stages are mapped in dependency order.
+
+2. Verify workspace asset directories.
+   - Create organized asset folders inside workspace (e.g. `assets/renders/`, `assets/graphics/`, `projects/`).
+   - Ensure all intermediate file paths point to future outputs that prior stages produce.
+   - **Done when:** directory structure is prepared.
+
+3. Configure pipeline stages.
+   - Stage configuration uses an array of `{ id, type, config }`.
+   - `id`: Unique identifier for logging and failure isolation (e.g., `stage_render_hero`, `stage_composite_badge`, `stage_assemble_timeline`).
+   - `type`: `"3d_batch"` for Blender operations, `"studio_batch"` for Photo/Video/Motion Studio operations.
+   - `config`: Complete batch payload conforming to respective engine requirements.
+   - **Done when:** stage definitions are fully structured.
+
+4. Run the coordinated pipeline.
+   - Invoke `creative_apps` with `action: "studio_pipeline"` and `stages`.
+   - The runtime executes stages sequentially.
+   - If any stage encounters an issue, execution halts immediately with full diagnostic reporting, leaving prior outputs intact for inspection.
+   - **Done when:** all stages return success.
+
+5. Inspect artifacts.
+   - Confirm generation of all planned project artifacts (.blend, .png/.psd, .prproj).
+   - Verify file sizes are greater than zero.
+   - **Done when:** all artifacts are present and verified.
+
+## Failure Modes Worth Naming
+
+- **Stage output collision:** Each stage must specify an uncreated file path; existing filenames cause immediate refusal.
+- **Missing intermediate files:** Stage N must only reference outputs created in Stage N-1 or earlier.
+- **Host unavailable:** If a host required by a stage is missing, check environment via `action: "status"` beforehand.
+
+## Verification
+
+- `action: "studio_pipeline"` completes with `success: true`.
+- `stagesCompleted` equals total stages count.
+- Every intermediate and final asset exists on disk in the workspace.
+- Diagnostic logs demonstrate clean execution across each host engine.

package/native-skills/media/photo-studio-automation/SKILL.md

@@ -0,0 +1,53 @@
+---
+name: photo-studio-automation
+description: Automate photo editing, graphic composition, text typography, and image format exports using native Photo Studio automation. Use when a task requires batch image edits, graphic layout creation, resizing, color adjustment, or PSD/PNG/JPG/WEBP asset generation.
+---
+
+# Photo Studio Automation
+
+Direct script execution beats screenshot clicking every time: it is exact, sub-second, costs zero vision tokens, and never misclicks because a window moved. Use the native Creative Studio bridge via `creative_apps` with `action: "studio_edit"` and `host: "photo-studio"`.
+
+## Workflow
+
+1. Verify environment and inspect installed creative hosts.
+   - Run `creative_apps` with `action: "status"`.
+   - Confirm Photo Studio availability before assembling jobs.
+   - **Done when:** host availability is confirmed.
+
+2. Establish workspace paths.
+   - All inputs and outputs must reside strictly within the active workspace.
+   - The output file must not already exist (the runtime enforces collision prevention).
+   - Supported output extensions: `.psd`, `.png`, `.jpg`, `.jpeg`, `.webp`.
+   - **Done when:** source asset and clean target output paths are defined.
+
+3. Plan layer hierarchy and typography before applying edits.
+   - If starting from scratch, omit `input` to let the engine create a canvas (1920x1080 standard).
+   - If transforming existing art, provide `input` path.
+   - Layer modifications should specify explicit `name`, `opacity`, `visible`, and `blendMode`.
+   - Typography actions should define `text`, `x`, `y`, `size`, and font family.
+   - **Done when:** sequence of edit actions is organized logically from background to foreground.
+
+4. Apply color adjustments and canvas sizing.
+   - Use `resize` or `crop` early in the action pipeline to establish composition geometry.
+   - Use `adjust_color` with `brightness` (-100..100) and `contrast` (-100..100) on target layers for grading.
+   - If delivering flat web assets (e.g., PNG or WEBP), include `flatten` at the end of the action chain.
+   - **Done when:** geometric and tonal parameters match design requirements.
+
+5. Execute batch and verify output.
+   - Invoke `creative_apps` with `action: "studio_edit"`, `host: "photo-studio"`, `output`, and `actions`.
+   - The runtime verifies execution receipt and confirms nonempty file generation.
+   - **Done when:** output file exists, has nonzero byte count, and matches requested specifications.
+
+## Failure Modes Worth Naming
+
+- **Overwriting existing file:** The engine halts if output exists. Specify a distinct or versioned filename.
+- **Outside workspace:** All paths must resolve inside the workspace root. Do not reference temporary directories or home root.
+- **Missing host:** If the host is not installed, the tool gracefully reports installation status. Propose alternative image processing or check system paths.
+- **Out-of-range layer index:** Layer indices start at 0. Ensure index matches existing layer count.
+
+## Verification
+
+- `action: "studio_edit"` returns `success: true`.
+- Output file exists on disk within workspace.
+- File size is non-zero.
+- Format matches target extension (.psd, .png, .jpg, or .webp).

package/native-skills/media/video-timeline-assembly/SKILL.md

@@ -0,0 +1,52 @@
+---
+name: video-timeline-assembly
+description: Assemble video timelines, import media clips to sequences, place markers, make precision edits, and export project files using native Video Studio automation. Use when a task requires automated video sequence building, cutting, multicam timeline layout, or PRPROJ/XML project generation.
+---
+
+# Video Timeline Assembly
+
+Timeline video editing is vastly faster and more dependable through direct project automation than GUI pixel clicking. Causyne Agent connects to the local Video Studio host via `creative_apps` (`action: "studio_edit"`, `host: "video-studio"`).
+
+## Workflow
+
+1. Check host installation.
+   - Run `creative_apps` with `action: "status"`.
+   - Ensure the Video Studio engine is discovered locally.
+   - **Done when:** host availability is confirmed.
+
+2. Prepare media assets in workspace.
+   - Collect source video, audio, and graphic clips inside the project workspace directory.
+   - Keep media paths relative to the current workspace root.
+   - Output must be a new `.prproj` or `.xml` file path that does not yet exist.
+   - **Done when:** all media files are present on disk and target project path is specified.
+
+3. Establish sequence settings.
+   - Begin with a `sequence` action specifying sequence `name`, `width`, `height`, and `fps`.
+   - Standard 1080p: `width: 1920`, `height: 1080`, `fps: 30` (or 24).
+   - Standard 4K UHD: `width: 3840`, `height: 2160`, `fps: 60` (or 30/24).
+   - **Done when:** sequence resolution and frame rate are locked.
+
+4. Import and place media clips along the timeline.
+   - Use `import` actions with `path`, `start` time in seconds, and `track` index (0 for primary video).
+   - Specify `duration` if clipping the source media to a specific length.
+   - Place voiceover, ambient sound, or music tracks with `mediaType: "audio"`.
+   - Add timeline markers with `marker` action (`start`, `name`, `comment`) for scene transitions or cue points.
+   - **Done when:** all media elements are arranged sequentially without unexpected track overlaps.
+
+5. Execute automation and inspect generated project.
+   - Invoke `creative_apps` with `action: "studio_edit"`, `host: "video-studio"`, `output`, and `actions`.
+   - Verify that the resulting `.prproj` or `.xml` is written cleanly.
+   - **Done when:** project file exists on disk with non-zero size.
+
+## Failure Modes Worth Naming
+
+- **Media file missing on disk:** Every imported file path must exist before running the automation batch.
+- **Output collision:** Target `.prproj` must be a fresh filename; the runtime refuses to overwrite existing project files.
+- **Timeline drift:** When laying clips sequentially, track cumulative duration so clips don't unintentionally overlap on the same track unless intended.
+
+## Verification
+
+- `action: "studio_edit"` returns `success: true`.
+- Output `.prproj` or `.xml` exists inside workspace.
+- Output file size is non-zero.
+- All imported media files exist and are referenced in the project structure.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@causyne/agent",
-  "version": "1.10.18",
+  "version": "1.10.20",
   "description": "Agentic AI for software, tools, and computer workflows",
   "main": "bin/causyne.js",
   "type": "module",

package/src/agentLoop.js

@@ -568,6 +568,7 @@ export class AgentLoop {
   }
 
   initializeProvider() {
+    if (!this.config?.provider) return null;
     return createProvider(this.config);
   }
 
@@ -808,6 +809,9 @@ export class AgentLoop {
     this.rearmAbortController();
     // Refresh tier limits on every task run
     this.updateTierLimits();
+    if (!this.llm) {
+      throw new Error("API Key Required. Causyne is a 100% Sovereign BYOK Engine. Please configure your API Provider and Key in Settings.");
+    }
 
     // 1. PERCEIVE — runs once per task so the agent starts with real workspace
     // awareness instead of burning steps rediscovering it. The result used to be

package/src/appChannel.js

@@ -17,10 +17,9 @@
 // ~200 tokens on every request to describe a decision the runtime can make on
 // its own, and a decision the model is allowed to skip is not a guard.
 //
-// EXTENDING IT: a Higgsfield-style plugin (Photoshop UXP, a Blender add-on) is
+// EXTENDING IT: A native automation bridge (ExtendScript bridge, Blender Python engine) is
 // one more CHANNEL_A row, not a parallel architecture. That is the whole point
-// of the table — the lesson from Higgsfield is "a native app API beats pixels",
-// not "build our own Higgsfield".
+// of the table — "a native app API beats pixels".
 
 /**
  * Apps that already have a Channel A tool, keyed by matcher.
@@ -38,6 +37,11 @@ const CHANNEL_A = [
     match: /\b(chrome|chromium|safari|firefox|edge|brave|arc)\b/,
     tool: 'browser_act',
     app: 'a web browser'
+  },
+  {
+    match: /\b(photo-studio|video-studio|motion-studio|photostudio|videostudio|motionstudio)\b/,
+    tool: 'creative_apps',
+    app: 'Creative Studio Bridge'
   }
 ];
 

package/src/cliMode.js

@@ -1,10 +1,11 @@
+import { runSetup } from './tui/setup.js';
 import { AgentLoop } from './agentLoop.js';
 import { loadConfig, saveConfig } from './config.js';
 import readline from 'readline';
 import { ChatPrompt, isInteractive } from './tui/prompt.js';
 import { handleInput, listCommands, isCommandInput } from './commands.js';
 import { ChatRenderer } from './tui/chatRenderer.js';
-import { paint, bold, badge, GLYPHS, terminalWidth } from './tui/theme.js';
+import { paint, bold, dim, badge, GLYPHS, terminalWidth } from './tui/theme.js';
 import { wordmark } from './tui/banner.js';
 import { createStageSurface } from './tui/stageSurface.js';
 import { wrapBlock } from './tui/text.js';
@@ -227,7 +228,7 @@ async function runHomeAction(action, { prompt, conn, finish }) {
       if (!prompt) { await interactiveConfig({}); return 'exit'; }
       const chosen = await configureInPrompt(prompt, { loadConfig, saveConfig });
       conn.renderer.write('\n' + (chosen
-        ? paint('success', `  Saved. Using ${chosen.provider} / ${chosen.model}.`)
+        ? paint('success', `  Saved. Using ${chosen.provider} / ${chosen.model}. Connection has not been tested.`)
         : dimNotice('  Cancelled. Nothing was changed.')) + '\n');
       reopen();
       return null;
@@ -547,6 +548,11 @@ export async function chatMode({ home = false } = {}) {
     const startedAt = Date.now();
     if (prompt) prompt.setBusy(true);
     try {
+      if (!loop.config.provider || (!loop.config.apiKey && loop.config.provider !== 'ollama')) {
+        conn.renderer.write('\n' + paint('warn', '  No API provider or key configured.') + '\n' +
+          paint('muted', '  Run /setup or select setup from the home menu to configure your provider.\n'));
+        continue;
+      }
       await loop.run();
     } finally {
       // Must be released on the error path too, or one failed turn leaves the
@@ -604,9 +610,14 @@ export async function execCommand(command) {
 export async function interactiveConfig(options) {
   const config = loadConfig();
 
-  console.log(logoLines() + '\n' + bold('Configuration') + '\n');
+  console.log(logoLines() + '\n' + bold('  Causyne Agent · Guided Setup & Configuration') + '\n'
+    + dim('  Standardized UI/UX · Models · Providers · Permissions · Creative Studio Bridge\n'));
 
   if (options.provider || options.apikey || options.model) {
+    if (options.provider && options.provider !== config.provider) {
+      config.apiKey = '';
+      delete config.baseUrl;
+    }
     if (options.provider) config.provider = options.provider;
     if (options.apikey) config.apiKey = options.apikey;
     if (options.model) config.model = options.model;
@@ -615,71 +626,26 @@ export async function interactiveConfig(options) {
     return;
   }
 
-  // Asked in the product's own panels rather than a second prompt framework.
-  //
-  // These were the last two surfaces on a second design system — different
-  // colours, different glyphs, no NO_COLOR handling, and a list widget nothing
-  // like the picker the same user sees inside `causyne`. askSelect/askText host a
-  // real dialog on a temporary ChatPrompt and fall back to readline wherever raw
-  // mode is unavailable. PROVIDERS is shared with the in-chat config flow, so
-  // the two can no longer offer different lists.
-  const provider = await askSelect({
-    title: 'AI provider',
-    items: PROVIDERS.map((entry) => ({ id: entry.id, label: entry.label, description: entry.hint }))
-  });
-  if (!provider) { console.log('  Cancelled. Nothing was changed.\n'); return; }
-
-  const preset = PROVIDERS.find((entry) => entry.id === provider.id) || { model: '' };
-  const suggested = (config.provider === provider.id && config.model) ? config.model : preset.model;
-
-  const model = await askText({
-    title: 'model',
-    lines: [
-      `${provider.label}`,
-      '',
-      'Model ids change often. Type the one you want, or accept the suggestion.'
-    ],
-    value: suggested,
-    placeholder: 'model id'
-  });
-  if (model === null) { console.log('  Cancelled. Nothing was changed.\n'); return; }
-
-  const apiKey = await askText({
-    title: 'API key',
-    lines: [
-      `${provider.label} ${GLYPHS.arrow} ${model || suggested}`,
-      '',
-      'Stored encrypted in ~/.causyne/config.json. It never leaves this machine.'
-    ],
-    mask: true,
-    placeholder: config.apiKey ? 'blank keeps the current key' : 'paste your key'
-  });
-  if (apiKey === null) { console.log('  Cancelled. Nothing was changed.\n'); return; }
-
-  // Only asked where it means something. Every provider used to be shown this,
-  // and for eight of the ten the right answer is always blank.
-  let baseUrl = config.baseUrl || '';
-  if (provider.id === 'openai-compatible' || provider.id === 'dashscope-byok' || provider.id === 'ollama') {
-    const answer = await askText({
-      title: 'base URL',
-      lines: ['Leave blank for the provider default.', 'Required for a custom DashScope workspace endpoint.'],
-      value: baseUrl,
-      placeholder: 'https://…'
-    });
-    if (answer === null) { console.log('  Cancelled. Nothing was changed.\n'); return; }
-    baseUrl = answer;
-  }
-
-  config.provider = provider.id;
-  config.model = (model || suggested || '').trim();
-  if (apiKey) config.apiKey = apiKey;
-  if (baseUrl && baseUrl.trim()) config.baseUrl = baseUrl.trim(); else delete config.baseUrl;
-
-  saveConfig(config);
-  console.log('\n' + paint('success', `  Saved. Using ${config.provider} / ${config.model}.`) + '\n');
+  const result = await runSetup({ select: askSelect, text: askText }, PROVIDERS, { loadConfig, saveConfig });
+  if (!result) { console.log('  Cancelled. Nothing was changed.\n'); return; }
+
+  let creativeSummary = '';
+  try {
+    const { creativeAppStatus } = await import('./creative/apps.js');
+    const apps = await creativeAppStatus();
+    const ready = apps.filter(a => a.available).map(a => a.label);
+    creativeSummary = ready.length > 0
+      ? `  ${paint('success', GLYPHS.bullet)} Creative Studio Bridge: ${ready.join(', ')} connected`
+      : `  ${dim(GLYPHS.bullet + ' Creative Studio Bridge: install Blender, Premiere, Photoshop, or After Effects to activate')}`;
+  } catch { /* ignore */ }
+
+  console.log('\n' + paint('success', `  ✔ Saved configuration successfully!`) + '\n'
+    + `  Active Provider: ${bold(result.provider)}\n`
+    + `  Active Model:    ${bold(result.model)}\n`
+    + (creativeSummary ? `${creativeSummary}\n` : '')
+    + `  Launch agent:    ${bold('causyne')} or ${bold('causyne chat')}\n`
+    + `  Diagnostics:     ${bold('causyne doctor')}\n`);
 
   // Seamlessly transition to the home screen.
-  //
-  // Finishing configuration returns to the local home screen.
   await chatMode({ home: true });
 }

package/src/commands.js

@@ -20,6 +20,7 @@ import {
 } from './userSkills.js';
 import { buildContextText, buildUsageText, buildCheckpointsText, buildUndoText, buildStatusText } from './reporting.js';
 import { loadConfig, saveConfig } from './config.js';
+import { resolveAvailableModels, getCuratedModels } from './dynamicCatalogue.js';
 import {
   serverStatus, grantServer, revokeServer, addServer, removeServer,
   connectServer, disconnectServer, refreshServer
@@ -44,6 +45,17 @@ function fail(output) {
 // ---------------------------------------------------------------------------
 
 export const COMMANDS = [
+  {
+    name: 'setup', aliases: ['config', 'connect'], scope: 'local', usage: '/setup',
+    description: 'Configure provider, endpoint, model and autonomy',
+    run: async (_argv, ctx) => {
+      if (!ctx.prompt) return ok('Run causyne setup in an interactive terminal.');
+      const { configureInPrompt } = await import('./tui/home.js');
+      const result = await configureInPrompt(ctx.prompt, { loadConfig, saveConfig });
+      return ok(result ? `Saved ${result.provider} / ${result.model}. Connection has not been tested.` : 'Cancelled. Nothing was changed.');
+    }
+  },
+
   {
     name: 'help',
     aliases: ['?', 'commands'],
@@ -632,22 +644,58 @@ export const COMMANDS = [
 
   {
     name: 'model',
+    aliases: ['models'],
     scope: 'local',
-    usage: '/model [provider/model | model]',
-    description: 'Show or change the model in use',
+    usage: '/model [list | refresh | provider/model | model]',
+    description: 'Show, list or change the model in use',
     dialogTitle: 'Model',
-    run: (argv) => {
+    complete: () => {
+      const config = loadConfig();
+      const currentProvider = config.provider || '';
+      const curated = getCuratedModels(currentProvider);
+      const items = [
+        { value: 'list', label: 'list', description: 'list models for active provider' },
+        { value: 'refresh', label: 'refresh', description: 'poll provider API for live models' }
+      ];
+      for (const m of curated) {
+        items.push({ value: m, label: m, description: currentProvider || 'model' });
+      }
+      return items;
+    },
+    run: async (argv) => {
       const config = loadConfig();
       if (argv.length === 0) {
-        return ok(`Provider : ${config.provider || 'default'}\nModel    : ${config.model || 'default'}`);
+        return ok(`Provider : ${config.provider || 'no provider configured'}\nModel    : ${config.model || 'none'}\n\nUse /model list to see models, or /model refresh to poll provider API.`);
+      }
+      const first = argv[0].toLowerCase();
+      if (first === 'list' || first === 'ls') {
+        const targetProvider = argv[1] || config.provider;
+        if (!targetProvider) return fail('No provider specified or configured. Run /setup first.');
+        const result = await resolveAvailableModels(targetProvider, config);
+        const active = config.model;
+        const lines = [
+          `Available models for ${targetProvider} (source: ${result.source}):`,
+          ''
+        ];
+        for (const m of result.models) {
+          const isCurrent = m === active;
+          lines.push(`  ${isCurrent ? '• ' : '  '}${m}${isCurrent ? '  (active)' : ''}`);
+        }
+        lines.push('');
+        lines.push('Use /model <name> to select, or /model refresh to poll the provider API.');
+        return ok(lines.join('\n'));
+      }
+      if (first === 'refresh') {
+        if (!config.provider) return fail('No provider configured. Run /setup first.');
+        const result = await resolveAvailableModels(config.provider, config, true);
+        return ok(`Refreshed ${config.provider} models (${result.models.length} available, source: ${result.source}). Run /model list to view.`);
       }
       const requested = argv.join(' ').trim();
       const separator = requested.indexOf('/');
-      // A bare name changes only the model; "provider/model" changes both. Any
-      // string is accepted — model ids move faster than any hardcoded list, and
-      // guessing from memory is exactly what .agents/AGENTS.md §3 forbids.
       if (separator > 0) {
-        config.provider = requested.slice(0, separator);
+        const provider = requested.slice(0, separator);
+        if (provider !== config.provider) { config.apiKey = ''; delete config.baseUrl; }
+        config.provider = provider;
         config.model = requested.slice(separator + 1);
       } else {
         config.model = requested;

package/src/compiledSkills.js

@@ -318,6 +318,21 @@ export const COMPILED_SKILLS = {
       "name": "cinematic-3d-shot",
       "description": "Build a 3D object in Blender and render it as a short cinematic clip. Use when a task asks for a 3D model, a product shot, a turntable, or a rendered animation of a few seconds.",
       "content": "---\nname: cinematic-3d-shot\ndescription: Build a 3D object in Blender and render it as a short cinematic clip. Use when a task asks for a 3D model, a product shot, a turntable, or a rendered animation of a few seconds.\n---\n\n# Cinematic 3D Shot\n\nA render looks amateur for four reasons, always the same four: default grey\nmaterial, one flat light, the camera where Blender left it, and nobody looked at\nthe result before shipping it. Fix those four and an ordinary model reads as a\nproduct shot.\n\n## Workflow\n\n1. Verify Blender before planning anything.\n   - Use the local CLI path. Budget a low-cost `renderFrame` whenever you need\n     to inspect the current visual result.\n   - **Done when:** the local Blender binary is available and the review loop is planned.\n\n2. Clear the scene.\n   - Blender opens with a cube, one lamp and a camera. Building around them is\n     how the default cube ends up in a client render.\n   - **Done when:** `clearScene` has run.\n\n3. Build the silhouette, then the detail.\n   - A shape is read as an outline before it is read as a surface. If the form\n     does not identify itself at thumbnail size, no amount of material work will\n     save it. Use `smartGeometry` — it applies bevel and subdivision, and a\n     bevel is what catches a highlight along an edge. Perfectly sharp edges are\n     the most common tell of untouched 3D: nothing in the real world has them.\n   - **Done when:** the object is recognisable as a flat grey silhouette.\n\n4. Give every visible object a material.\n   - `smartMaterial` on each one. One untouched object in a dressed scene reads\n     as unfinished more loudly than a whole scene of grey.\n   - Match the material to the object's story: a pen barrel is brushed metal or\n     lacquered resin, not plastic; the grip is rubber, and rubber is rough where\n     metal is not. Contrast between materials is what tells the eye there are\n     two parts.\n   - **Done when:** no object still carries the default surface.\n\n5. Light with three lights, in this order.\n   - `setupStudioLighting` gives key, fill and rim. The ratio does the work:\n     the key states the form, the fill keeps the shadow side from going black,\n     and the RIM is what separates the subject from the background. A render\n     that looks flat is almost always missing the rim.\n   - **Done when:** the subject's edge is visibly brighter than what is behind it.\n\n6. Add atmosphere only after the lights exist.\n   - `cinematicAtmosphere`. Fog scatters the light you already have, so adding\n     it first means tuning against something that will change. Keep the density\n     low — visible haze reads as a mistake unless the shot is about weather.\n   - **Done when:** there is depth between foreground and background, and the\n     subject is still crisp.\n\n7. Frame it deliberately.\n   - `autoFrameCamera`. A longer lens (65mm and up) flatters an object; a wide\n     lens distorts it and is why so many AI renders look like a fisheye photo of\n     a toy. Put the camera slightly above the eye line and the subject a little\n     off-centre.\n   - **Done when:** the camera is not where Blender put it.\n\n8. Look at it before you animate.\n   - `captureViewport`, or one low-sample `renderFrame`, then `view_image` or\n     `vision_inspect_ui`. Ask the specific question: is anything still grey, is\n     the subject separated from the background, does the silhouette read?\n   - **A render nobody has looked at is not a finished render.** This step is\n     the difference between the two outcomes, not a formality.\n   - **Done when:** you have described what you actually see, not what you\n     intended to build.\n\n9. One camera move, and do the arithmetic out loud.\n   - Seconds x fps = frames. 8s at 24fps is frames 1-192; set the range\n     explicitly or Blender renders its own 250-frame default.\n   - ONE move per short shot: a slow orbit, a push-in, or a rise. Three moves in\n     eight seconds reads as a screensaver. Ease the ends — constant velocity is\n     what makes a move feel mechanical.\n   - **Done when:** the frame range matches the requested duration exactly.\n\n10. Render with settings that fit the budget.\n    - EEVEE for anything short. Cycles at default samples will spend an entire\n      task budget on one frame and is almost never what a few seconds of\n      turntable needs.\n    - **Done when:** the file exists, and you have opened it to confirm it is\n      not zero bytes or a single black frame.\n\n## Failure modes worth naming\n\n- **Everything is grey.** A material step was skipped, or applied to one object.\n- **It looks flat.** No rim light, or fill too close to key.\n- **The object looks like a toy.** Lens too wide, or the camera is too close.\n- **The animation stutters.** Frame range and fps disagree with the duration.\n- **Blender is not installed.** Say so and offer the alternative. Do not loop\n  retrying the same missing binary — that spends budget to learn nothing.\n\n## Verification\n\n- The rendered file exists, is larger than a few kilobytes, and opens.\n- Open the first frame and the middle frame. Neither is black, neither is empty,\n  and the subject is in frame in both.\n- No object in shot still carries the default grey surface.\n- The subject's edge is brighter than the background directly behind it.\n- Frame count equals requested seconds x fps, exactly.\n- The camera position differs from Blender's default.\n"
+    },
+    "creative-studio-pipeline": {
+      "name": "creative-studio-pipeline",
+      "description": "Orchestrate end-to-end multi-stage creative production pipelines coordinating 3D modeling and rendering in Blender, Photo Studio grading and graphic compositing, and Video Studio sequence assembly. Use when a task spans multiple creative domains or requires chaining 3D outputs into video edits.",
+      "content": "---\nname: creative-studio-pipeline\ndescription: Orchestrate end-to-end multi-stage creative production pipelines coordinating 3D modeling and rendering in Blender, Photo Studio grading and graphic compositing, and Video Studio sequence assembly. Use when a task spans multiple creative domains or requires chaining 3D outputs into video edits.\n---\n\n# Creative Studio Pipeline Orchestration\n\nMulti-stage creative production is most reliable when coordinated through a single declarative pipeline. Rather than switching manual applications or invoking separate tools out of sequence, Causyne Agent provides `creative_apps` with `action: \"studio_pipeline\"`.\n\n## Workflow\n\n1. Design the production stages.\n   - Stage A (3D Scene / Asset Generation): Blender batch (`3d_batch`) creates geometry, assigns PBR materials, positions lights/camera, and renders high-resolution stills or asset elements.\n   - Stage B (Still Grading / Composition): Photo Studio batch (`studio_batch`, host `photo-studio`) adjusts color grading, adds typographic overlays or branding, and exports optimized graphic assets.\n   - Stage C (Timeline Editing): Video Studio batch (`studio_batch`, host `video-studio`) constructs the sequence timeline, imports rendered stills/clips, cuts intervals, and saves the final project.\n   - **Done when:** stages are mapped in dependency order.\n\n2. Verify workspace asset directories.\n   - Create organized asset folders inside workspace (e.g. `assets/renders/`, `assets/graphics/`, `projects/`).\n   - Ensure all intermediate file paths point to future outputs that prior stages produce.\n   - **Done when:** directory structure is prepared.\n\n3. Configure pipeline stages.\n   - Stage configuration uses an array of `{ id, type, config }`.\n   - `id`: Unique identifier for logging and failure isolation (e.g., `stage_render_hero`, `stage_composite_badge`, `stage_assemble_timeline`).\n   - `type`: `\"3d_batch\"` for Blender operations, `\"studio_batch\"` for Photo/Video/Motion Studio operations.\n   - `config`: Complete batch payload conforming to respective engine requirements.\n   - **Done when:** stage definitions are fully structured.\n\n4. Run the coordinated pipeline.\n   - Invoke `creative_apps` with `action: \"studio_pipeline\"` and `stages`.\n   - The runtime executes stages sequentially.\n   - If any stage encounters an issue, execution halts immediately with full diagnostic reporting, leaving prior outputs intact for inspection.\n   - **Done when:** all stages return success.\n\n5. Inspect artifacts.\n   - Confirm generation of all planned project artifacts (.blend, .png/.psd, .prproj).\n   - Verify file sizes are greater than zero.\n   - **Done when:** all artifacts are present and verified.\n\n## Failure Modes Worth Naming\n\n- **Stage output collision:** Each stage must specify an uncreated file path; existing filenames cause immediate refusal.\n- **Missing intermediate files:** Stage N must only reference outputs created in Stage N-1 or earlier.\n- **Host unavailable:** If a host required by a stage is missing, check environment via `action: \"status\"` beforehand.\n\n## Verification\n\n- `action: \"studio_pipeline\"` completes with `success: true`.\n- `stagesCompleted` equals total stages count.\n- Every intermediate and final asset exists on disk in the workspace.\n- Diagnostic logs demonstrate clean execution across each host engine.\n"
+    },
+    "photo-studio-automation": {
+      "name": "photo-studio-automation",
+      "description": "Automate photo editing, graphic composition, text typography, and image format exports using native Photo Studio automation. Use when a task requires batch image edits, graphic layout creation, resizing, color adjustment, or PSD/PNG/JPG/WEBP asset generation.",
+      "content": "---\nname: photo-studio-automation\ndescription: Automate photo editing, graphic composition, text typography, and image format exports using native Photo Studio automation. Use when a task requires batch image edits, graphic layout creation, resizing, color adjustment, or PSD/PNG/JPG/WEBP asset generation.\n---\n\n# Photo Studio Automation\n\nDirect script execution beats screenshot clicking every time: it is exact, sub-second, costs zero vision tokens, and never misclicks because a window moved. Use the native Creative Studio bridge via `creative_apps` with `action: \"studio_edit\"` and `host: \"photo-studio\"`.\n\n## Workflow\n\n1. Verify environment and inspect installed creative hosts.\n   - Run `creative_apps` with `action: \"status\"`.\n   - Confirm Photo Studio availability before assembling jobs.\n   - **Done when:** host availability is confirmed.\n\n2. Establish workspace paths.\n   - All inputs and outputs must reside strictly within the active workspace.\n   - The output file must not already exist (the runtime enforces collision prevention).\n   - Supported output extensions: `.psd`, `.png`, `.jpg`, `.jpeg`, `.webp`.\n   - **Done when:** source asset and clean target output paths are defined.\n\n3. Plan layer hierarchy and typography before applying edits.\n   - If starting from scratch, omit `input` to let the engine create a canvas (1920x1080 standard).\n   - If transforming existing art, provide `input` path.\n   - Layer modifications should specify explicit `name`, `opacity`, `visible`, and `blendMode`.\n   - Typography actions should define `text`, `x`, `y`, `size`, and font family.\n   - **Done when:** sequence of edit actions is organized logically from background to foreground.\n\n4. Apply color adjustments and canvas sizing.\n   - Use `resize` or `crop` early in the action pipeline to establish composition geometry.\n   - Use `adjust_color` with `brightness` (-100..100) and `contrast` (-100..100) on target layers for grading.\n   - If delivering flat web assets (e.g., PNG or WEBP), include `flatten` at the end of the action chain.\n   - **Done when:** geometric and tonal parameters match design requirements.\n\n5. Execute batch and verify output.\n   - Invoke `creative_apps` with `action: \"studio_edit\"`, `host: \"photo-studio\"`, `output`, and `actions`.\n   - The runtime verifies execution receipt and confirms nonempty file generation.\n   - **Done when:** output file exists, has nonzero byte count, and matches requested specifications.\n\n## Failure Modes Worth Naming\n\n- **Overwriting existing file:** The engine halts if output exists. Specify a distinct or versioned filename.\n- **Outside workspace:** All paths must resolve inside the workspace root. Do not reference temporary directories or home root.\n- **Missing host:** If the host is not installed, the tool gracefully reports installation status. Propose alternative image processing or check system paths.\n- **Out-of-range layer index:** Layer indices start at 0. Ensure index matches existing layer count.\n\n## Verification\n\n- `action: \"studio_edit\"` returns `success: true`.\n- Output file exists on disk within workspace.\n- File size is non-zero.\n- Format matches target extension (.psd, .png, .jpg, or .webp).\n"
+    },
+    "video-timeline-assembly": {
+      "name": "video-timeline-assembly",
+      "description": "Assemble video timelines, import media clips to sequences, place markers, make precision edits, and export project files using native Video Studio automation. Use when a task requires automated video sequence building, cutting, multicam timeline layout, or PRPROJ/XML project generation.",
+      "content": "---\nname: video-timeline-assembly\ndescription: Assemble video timelines, import media clips to sequences, place markers, make precision edits, and export project files using native Video Studio automation. Use when a task requires automated video sequence building, cutting, multicam timeline layout, or PRPROJ/XML project generation.\n---\n\n# Video Timeline Assembly\n\nTimeline video editing is vastly faster and more dependable through direct project automation than GUI pixel clicking. Causyne Agent connects to the local Video Studio host via `creative_apps` (`action: \"studio_edit\"`, `host: \"video-studio\"`).\n\n## Workflow\n\n1. Check host installation.\n   - Run `creative_apps` with `action: \"status\"`.\n   - Ensure the Video Studio engine is discovered locally.\n   - **Done when:** host availability is confirmed.\n\n2. Prepare media assets in workspace.\n   - Collect source video, audio, and graphic clips inside the project workspace directory.\n   - Keep media paths relative to the current workspace root.\n   - Output must be a new `.prproj` or `.xml` file path that does not yet exist.\n   - **Done when:** all media files are present on disk and target project path is specified.\n\n3. Establish sequence settings.\n   - Begin with a `sequence` action specifying sequence `name`, `width`, `height`, and `fps`.\n   - Standard 1080p: `width: 1920`, `height: 1080`, `fps: 30` (or 24).\n   - Standard 4K UHD: `width: 3840`, `height: 2160`, `fps: 60` (or 30/24).\n   - **Done when:** sequence resolution and frame rate are locked.\n\n4. Import and place media clips along the timeline.\n   - Use `import` actions with `path`, `start` time in seconds, and `track` index (0 for primary video).\n   - Specify `duration` if clipping the source media to a specific length.\n   - Place voiceover, ambient sound, or music tracks with `mediaType: \"audio\"`.\n   - Add timeline markers with `marker` action (`start`, `name`, `comment`) for scene transitions or cue points.\n   - **Done when:** all media elements are arranged sequentially without unexpected track overlaps.\n\n5. Execute automation and inspect generated project.\n   - Invoke `creative_apps` with `action: \"studio_edit\"`, `host: \"video-studio\"`, `output`, and `actions`.\n   - Verify that the resulting `.prproj` or `.xml` is written cleanly.\n   - **Done when:** project file exists on disk with non-zero size.\n\n## Failure Modes Worth Naming\n\n- **Media file missing on disk:** Every imported file path must exist before running the automation batch.\n- **Output collision:** Target `.prproj` must be a fresh filename; the runtime refuses to overwrite existing project files.\n- **Timeline drift:** When laying clips sequentially, track cumulative duration so clips don't unintentionally overlap on the same track unless intended.\n\n## Verification\n\n- `action: \"studio_edit\"` returns `success: true`.\n- Output `.prproj` or `.xml` exists inside workspace.\n- Output file size is non-zero.\n- All imported media files exist and are referenced in the project structure.\n"
     }
   },
   "mobile": {

package/src/config.js

@@ -78,30 +78,41 @@ export function loadConfig() {
         }
       }
 
-      // Self-heal a model id written before src/modelIds.js existed. The
-      // provider normalises too, but tokenTracker.getPricing() and the /model
-      // display both read config.model directly —
-      // they must agree with what is actually sent, not with what a stale
-      // Android build once wrote here.
+      // Self-heal a model id written before src/modelIds.js existed.
       const canonicalModel = canonicalModelId(parsed.model);
       if (canonicalModel !== parsed.model) {
         parsed.model = canonicalModel;
         saveConfig(parsed);
       }
 
+      // Self-heal legacy unconfigured DashScope bug:
+      // If config was written with provider 'dashscope' or 'dashscope-byok' but without an apiKey,
+      // it was an unintended default from older builds. Heal to an environment key or unconfigured.
+      if ((parsed.provider === 'dashscope' || parsed.provider === 'dashscope-byok') && !parsed.apiKey) {
+        const envDetected = detectEnvironmentProvider();
+        if (envDetected) {
+          parsed.provider = envDetected.provider;
+          parsed.model = envDetected.model;
+          parsed.apiKey = envDetected.apiKey;
+        } else {
+          parsed.provider = '';
+          parsed.model = '';
+        }
+        saveConfig(parsed);
+      }
+
       return parsed;
     } catch (e) {
       console.error('Error reading config.json:', e);
     }
   }
   
-    // Default config
+  // Default config: auto-detect from environment if available, otherwise unconfigured
+  const envDetected = detectEnvironmentProvider();
   const defaultConfig = {
-    provider: 'dashscope',
-    // From the catalogue, not repeated here — this had been left at the
-    // long-retired `qwen-plus` while the picker had moved to qwen3.7.
-    model: DEFAULT_MODEL_BY_PROVIDER.dashscope,
-    apiKey: '',
+    provider: envDetected ? envDetected.provider : '',
+    model: envDetected ? envDetected.model : '',
+    apiKey: envDetected ? envDetected.apiKey : '',
     autonomyMode: 'B',
     tokenBudget: {
       enabled: false,
@@ -116,6 +127,64 @@ export function loadConfig() {
   return defaultConfig;
 }
 
+/**
+ * Auto-detect available API keys from environment variables.
+ * Precedence matches frontier capability: Gemini, Anthropic, OpenAI, DeepSeek, xAI, Moonshot, Qwen/DashScope.
+ */
+export function detectEnvironmentProvider() {
+  const env = process.env;
+  if (env.GEMINI_API_KEY || env.GOOGLE_API_KEY) {
+    return {
+      provider: 'gemini',
+      apiKey: env.GEMINI_API_KEY || env.GOOGLE_API_KEY || '',
+      model: DEFAULT_MODEL_BY_PROVIDER.gemini
+    };
+  }
+  if (env.ANTHROPIC_API_KEY) {
+    return {
+      provider: 'anthropic',
+      apiKey: env.ANTHROPIC_API_KEY,
+      model: DEFAULT_MODEL_BY_PROVIDER.anthropic
+    };
+  }
+  if (env.OPENAI_API_KEY) {
+    return {
+      provider: 'openai',
+      apiKey: env.OPENAI_API_KEY,
+      model: DEFAULT_MODEL_BY_PROVIDER.openai
+    };
+  }
+  if (env.DEEPSEEK_API_KEY) {
+    return {
+      provider: 'deepseek',
+      apiKey: env.DEEPSEEK_API_KEY,
+      model: DEFAULT_MODEL_BY_PROVIDER.deepseek
+    };
+  }
+  if (env.XAI_API_KEY || env.GROK_API_KEY) {
+    return {
+      provider: 'xai',
+      apiKey: env.XAI_API_KEY || env.GROK_API_KEY || '',
+      model: DEFAULT_MODEL_BY_PROVIDER.xai
+    };
+  }
+  if (env.MOONSHOT_API_KEY) {
+    return {
+      provider: 'moonshot',
+      apiKey: env.MOONSHOT_API_KEY,
+      model: DEFAULT_MODEL_BY_PROVIDER.moonshot
+    };
+  }
+  if (env.DASHSCOPE_API_KEY || env.QWEN_API_KEY) {
+    return {
+      provider: 'dashscope-byok',
+      apiKey: env.DASHSCOPE_API_KEY || env.QWEN_API_KEY || '',
+      model: DEFAULT_MODEL_BY_PROVIDER['dashscope-byok']
+    };
+  }
+  return null;
+}
+
 export function saveConfig(config) {
   ensureConfigDir();
   // SECURITY: Deep copy to avoid mutating the in-memory config