[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_201":202,"_203":204},"routes/_app.x.$handle.$name",{"_11":12,"_13":14,"_101":-7,"_102":103,"_30":-7,"_104":-5,"_15":105,"_106":107,"_108":109,"_110":111,"_112":113,"_114":-5,"_115":-5,"_116":-5,"_117":46,"_118":119,"_120":58,"_121":-5,"_122":-5,"_123":124},"routes/_app.x.$handle.$name.branches",{"_11":12,"_13":14,"_15":16,"_17":18},"handle","xuhuanzhi32svg","name","infinite-canvas-mcp","project",{"_21":29,"_13":14,"_23":44,"_38":-5,"_45":46,"_47":48,"_49":48,"_50":51,"_52":-5,"_53":54,"_55":56,"_57":58,"_59":-5,"_60":-5,"_61":62,"_63":64,"_65":66,"_67":68,"_69":70,"_71":-5,"_72":73,"_74":73,"_75":76,"_77":76,"_78":76,"_79":73,"_80":73,"_81":82,"_83":84,"_85":86,"_87":46},"branches",[19,20],{"_21":33,"_23":39,"_25":40,"_13":41,"_28":29,"_30":42,"_32":-5,"_34":-5,"_36":43,"_38":-5},{"_21":22,"_23":24,"_25":26,"_13":27,"_28":29,"_30":31,"_32":33,"_34":35,"_36":37,"_38":-5},"id","75b14bbc-e602-4d5b-befa-fd2c642a1cf7","createdAt",["D",1789180137074],"updatedAt",["D",1789180137105],"backup-readonly-20260912","projectId","5f2940b9-497a-4c0e-b48a-7471214abdbb","version",1,"forkedBranchId","f7be5dbc-27c0-4d47-9c6a-f53b6506d92b","forkedBranchVersion",2,"lockfileKey","1789180137395-esm.town_v_xuhuanzhi32svg_infinite-canvas-mcp@1-backup-readonly-20260912","deletedAt",["D",1789176014798],["D",1789181312318],"main",16,"1789181312481-esm.town_v_xuhuanzhi32svg_infinite-canvas-mcp@16-main",["D",1789176014798],"legacyEnvVars",false,"privacy","public","httpPrivacy","description","Infinite Canvas remote MCP for ChatGPT","imageUrl","userId","aa4fc0c7-9a5e-471f-80f7-c417d4e86485","pinnedFileId","1b903a12-ae48-11f1-861e-1607ee4eb77e","pinPreview",true,"pinMigrated","screenshotVersion","likeCount",0,"orgId","01a0795e-e242-722a-a25b-204f9da61ffe","tursoName","5f2940b9497a4c0eb48a7471214abdbb","tursoHostname","5f2940b9497a4c0eb48a7471214abdbb-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789532458193],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_11":12,"_100":96},"org",{"_11":12,"_93":94,"_95":96,"_53":54,"_97":98,"_99":46},"branch",{"_30":42,"_13":41,"_21":33,"_28":29,"_32":-5,"_34":-5,"_88":-5,"_89":-5,"_90":-5,"_91":14,"_92":12},"isOwned","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","displayName","xuhuanzhi32-svg","avatar","https://avatars.githubusercontent.com/u/322119312?v=4","tier","free","canRestrictVals","profileImageUrl","branchName","path","","adoptOrgId",{"_21":29,"_13":14,"_23":196,"_38":-5,"_45":46,"_47":48,"_49":48,"_50":51,"_52":-5,"_53":54,"_55":56,"_57":58,"_59":-5,"_60":-5,"_61":62,"_63":64,"_65":66,"_67":68,"_69":197,"_71":-5,"_72":73,"_74":73,"_75":76,"_77":76,"_78":76,"_79":73,"_80":73,"_81":198,"_83":199,"_85":200,"_87":46},"tree",[179,180,181,182],"files",[179,180,181,182],"counts",{"_170":62,"_17":35,"_171":62,"_172":173,"_174":175},"contents",[125,126,127,128],"file","upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},{"_21":161,"_13":162,"_131":56,"_133":-5,"_134":33,"_135":163,"_137":58,"_23":164,"_30":165,"_140":166,"_142":167,"_144":46,"_145":168,"_102":169},{"_21":154,"_13":155,"_131":156,"_133":-5,"_134":33,"_135":157,"_137":58,"_23":158,"_30":62,"_140":141,"_142":159,"_144":46,"_145":-5,"_102":160},{"_21":147,"_13":148,"_131":149,"_133":-5,"_134":33,"_135":42,"_137":58,"_23":150,"_30":151,"_140":141,"_142":152,"_144":46,"_145":-5,"_102":153},{"_21":129,"_13":130,"_131":132,"_133":-5,"_134":33,"_135":136,"_137":58,"_23":138,"_30":139,"_140":141,"_142":143,"_144":46,"_145":-5,"_102":146},"e4ca8515-0a3a-42ef-ad94-90aa895a3bdd","workflow-core.js","fileId","1361f45c-ae52-11f1-a766-1607ee4eb77e","parentId","branchId","branchVersion",13,"latest",["D",1789181140473],4,"fileType","script","content","// Shared pure workflow contracts: imported by the MCP service and web canvas.\n// Provider calls are dispatched by the connected ChatGPT executor, not by this module.\nexport const VERSION = '0.3.0';\nexport const SOURCES = [\n {name: 'Weave · reusable I2V workflows', url: 'https://www.figma.com/solutions/ai-image-to-video-generator-weave/', lesson: 'Separate references, generation, edits and sequencing.'},\n {name: 'Weave · iterators', url: 'https://help.weavy.ai/en/articles/12343281-iterators', lesson: 'Keep batch inputs independent; review candidates before downstream use.'},\n {name: 'ComfyUI · Wan 2.2', url: 'https://docs.comfy.org/tutorials/video/wan/wan2_2', lesson: 'First/last frames are generation conditions, not optical-flow substitutes.'},\n {name: 'LTX · consistent characters', url: 'https://ltx.io/blog/how-to-create-a-consistent-character', lesson: 'Reuse an explicit character asset; treat wardrobe variants separately.'},\n {name: 'Weave · compositing', url: 'https://www.figma.com/solutions/ai-video-compositing/', lesson: 'Keep masks, color matching and compositing as inspectable stages.'}\n];\nconst spec = (id, name, output, inputs, executor, description) => ({id, name, output, inputs, executor, description});\nexport const NODE_TYPES = [\n spec('text', '创意 / 剧本', 'text', {}, 'input', '填写故事目标、受众和具体动作；空文本不会通过检查。'),\n spec('image', '角色 / 风格参考图', 'image', {}, 'input', '使用经过确认的参考图。身份、服装与风格分别记录。'),\n spec('video', '动作 / 实拍参考', 'video', {}, 'input', '提供真实视频，不用低帧率技术代理冒充参考表演。'),\n spec('audio', '对白 / 音乐参考', 'audio', {}, 'input', '提供拥有使用权限的对白、音乐或声音素材。'),\n spec('style', '美术设定锁定', 'text', {}, 'input', '锁定线条、材质、色彩、光线、服装和禁止变化项。'),\n spec('prompt', '提示词组装', 'text', {text: 'text'}, 'local', '合并上游文字与镜头描述，不声称经过大模型润色。'),\n spec('image_generate', '关键帧生成', 'image', {text: 'text'}, 'connector', '生成可独立审看的单张关键帧；计费前核对模型契约与余额。'),\n spec('image_edit', '参考图局部修改', 'image', {image: 'image'}, 'connector', '修改明确区域或服装，不静默更换角色身份。'),\n spec('text_to_video', '文生视频', 'video', {text: 'text'}, 'connector', '从文字生成真实连续视频，不降级为静帧拼接。'),\n spec('image_to_video', '图生视频', 'video', {start_image: 'image'}, 'connector', '首帧驱动主体表演；输出必须另外完成视觉审查。'),\n spec('first_last_video', '首尾帧艺术转场', 'video', {start_image: 'image', end_image: 'image'}, 'connector', '由支持首尾帧的模型生成过渡，不用交叉淡化代替。'),\n spec('motion_transfer', '视频动作迁移', 'video', {image: 'image', video: 'video'}, 'connector', '动作参考与角色参考独立；保留来源和实际任务 ID。'),\n spec('video_edit', '视频艺术重绘', 'video', {video: 'video', image: 'image'}, 'connector', '在保留时序和镜头运动的前提下重绘；先检查输入时长限制。'),\n spec('last_frame', '提取真实末帧', 'image', {video: 'video'}, 'media_worker', '从实际解码末帧提取，不用原参考图冒充上一镜末帧。'),\n spec('trim', '视频裁剪', 'video', {video: 'video'}, 'media_worker', '按真实时间区间裁剪已有视频，不生成缺失内容。'),\n spec('grade', '视频调色', 'video', {video: 'video'}, 'media_worker', '对已有视频调整曝光、对比和饱和度，不增加动作或细节。'),\n spec('speech', '对白配音', 'audio', {text: 'text'}, 'connector', '选定声音后生成对白；配音不等于口型同步。'),\n spec('sound', '环境 / 音效', 'audio', {text: 'text'}, 'connector', '生成独立环境声或音效，不替代人物表演。'),\n spec('technical_qc', '技术检测', 'video', {video: 'video'}, 'media_worker', '检查解码、尺寸、帧率、重复帧；不能据此自动判定美术合格。'),\n spec('review', '视觉验收关卡', 'video', {video: 'video'}, 'human', '检查自然度、身份、肢体、接触和连续性；未批准不得合成。'),\n spec('assemble', '批准镜头合成', 'video', {video: 'video'}, 'media_worker', '只合成已批准的视频，音频可选；禁止塞入静帧代替失败镜头。'),\n spec('mix', '对白 / 音效混音', 'video', {video: 'video', audio: 'audio'}, 'media_worker', '将真实音频按指定延迟混入视频，保留完整视频时长。')\n];\nconst BY_TYPE = new Map(NODE_TYPES.map(n => [n.id, n]));\nconst safeText = (x, max = 20000) => typeof x === 'string' && x.length <= max;\nconst ID = /^[A-Za-z0-9][A-Za-z0-9_-]{0,95}$/;\nexport function safeUrl(value) {\n if (typeof value !== 'string' || value.length > 8192) return false;\n try {\n const u = new URL(value);\n return u.protocol === 'https:' && !u.username && !u.password && !u.hash && u.hostname.includes('.') && !/^(localhost|127\\.|10\\.|192\\.168\\.|169\\.254\\.|0\\.|\\[)/.test(u.hostname) && !/\\.(local|internal)$/.test(u.hostname);\n } catch { return false; }\n}\nexport function operation(node) {\n return node.render?.operation || ({character: 'image', clip: 'video', audio: 'audio', review: 'review'}[node.type]) || 'text';\n}\nexport function mediaFor(node, jobs = {}) {\n const job = jobs[node.id];\n if (job && ['review', 'approved', 'completed'].includes(job.state) && job.output) return job.output;\n const type = BY_TYPE.get(operation(node))?.output;\n if (node.render?.source?.url && safeUrl(node.render.source.url)) return {kind: type, url: node.render.source.url, ...(node.render.source.media_id ? {media_id: node.render.source.media_id, provider: node.render.source.provider} : {})};\n if (node.render?.source?.media_id && /^[0-9a-f-]{36}$/i.test(node.render.source.media_id)) return {kind: type, media_id: node.render.source.media_id, provider: node.render.source.provider};\n if (type === 'text' && (node.prompt?.trim() || node.body?.trim())) return {kind: 'text', text: node.prompt || node.body};\n for (const kind of ['image', 'video', 'audio']) if (node[kind]) return {kind, url: safeUrl(node[kind]) ? node[kind] : undefined, local: !safeUrl(node[kind])};\n return null;\n}\nexport function compile(board, jobs = {}) {\n const errors = [], warnings = [], steps = [];\n if (!board || board.version !== 1 || !Array.isArray(board.nodes) || !Array.isArray(board.edges) || board.nodes.length > 2000 || board.edges.length > 5000) return {valid: false, errors: ['无效 v1 工程或超出节点上限'], warnings, steps};\n const map = new Map(), incoming = new Map(), outgoing = new Map();\n for (const n of board.nodes) {\n if (!n || !ID.test(n.id) || ['constructor','prototype','__proto__'].includes(n.id) || map.has(n.id) || !safeText(n.title, 200) || !safeText(n.body) || (n.prompt !== undefined && !safeText(n.prompt)) || !Number.isFinite(n.x) || !Number.isFinite(n.y)) {errors.push('节点 ID、文字或坐标无效'); continue;}\n map.set(n.id, n); incoming.set(n.id, []); outgoing.set(n.id, []);\n if (!BY_TYPE.has(operation(n))) errors.push(`${n.title}: 未知执行类型`);\n }\n const edgeIds = new Set();\n for (const e of board.edges) {\n if (!e || !ID.test(e.id) || edgeIds.has(e.id) || !map.has(e.from) || !map.has(e.to) || e.from === e.to) {errors.push('连线引用无效或重复'); continue;}\n edgeIds.add(e.id); incoming.get(e.to).push(e.from); outgoing.get(e.from).push(e.to);\n }\n const indegree = new Map([...incoming].map(([k,v]) => [k,v.length]));\n const queue = [...map.keys()].filter(id => indegree.get(id) === 0), order = [];\n for (let i = 0; i < queue.length; i++) {const id = queue[i]; order.push(id); for (const next of outgoing.get(id)) {indegree.set(next, indegree.get(next) - 1); if (indegree.get(next) === 0) queue.push(next);}}\n if (order.length !== map.size) errors.push('检测到循环依赖：工作流必须是无环图');\n for (const id of order) {\n const n = map.get(id), kind = operation(n), s = BY_TYPE.get(kind); if (!s) continue;\n const parents = incoming.get(id), bindings = {}, issues = [];\n const declared = n.render?.bindings || {};\n for (const [role, output] of Object.entries(s.inputs)) {\n const explicit = declared[role];\n const matches = explicit ? parents.filter(p => p === explicit) : parents.filter(p => BY_TYPE.get(operation(map.get(p)))?.output === output);\n if (matches.length === 0) issues.push(`缺少 ${role} 输入`);\n else if (matches.length > 1 && kind !== 'assemble' && role !== 'text') issues.push(`${role} 有多个来源，请明确绑定`);\n else if (matches.some(p => BY_TYPE.get(operation(map.get(p)))?.output !== output)) issues.push(`${role} 输入类型不匹配`);\n else bindings[role] = matches;\n }\n if (kind === 'first_last_video' && bindings.start_image?.[0] === bindings.end_image?.[0]) warnings.push(`${n.title}: 首尾引用相同，仅适用于明确设计的循环`);\n if (s.executor === 'input' && !mediaFor(n, jobs)) issues.push(s.output === 'text' ? '未填写内容' : '未绑定素材');\n if (s.executor === 'connector' && !n.prompt?.trim()) issues.push('未填写镜头提示词');\n if (n.duration_s !== undefined && (!Number.isFinite(n.duration_s) || n.duration_s <= 0 || n.duration_s > 3600)) issues.push('时长无效');\n const job = jobs[id];\n const waiting = parents.filter(p => {\n const ps = BY_TYPE.get(operation(map.get(p))); if (!ps) return true;\n if (kind === 'assemble') return jobs[p]?.state !== 'approved';\n if (ps.executor === 'input') return !mediaFor(map.get(p), jobs);\n const accepted = ['technical_qc','review'].includes(kind) ? ['review','approved','completed'] : ['approved','completed'];\n return !accepted.includes(jobs[p]?.state);\n });\n let state = job?.state || (issues.length ? 'needs_input' : waiting.length ? 'waiting_dependency' : s.executor === 'input' ? 'completed' : s.executor === 'human' ? 'needs_review' : 'ready');\n if (job?.state === 'review') state = 'review';\n steps.push({id, title: n.title, operation: kind, executor: s.executor, output: s.output, dependencies: parents, bindings, state, issues, waiting});\n }\n return {valid: errors.length === 0, ready: steps.filter(s => s.state === 'ready').map(s => s.id), errors, warnings, steps, execution_mode: 'chatgpt_connector_assisted', visual_quality_verified: false};\n}\nexport function connectorRequest(board, step, jobs = {}, billing = 'free_only') {\n const n = board.nodes.find(n => n.id === step.id);\n if (!n || step.executor !== 'connector') throw Error('不是模型生成节点');\n if (step.issues.length || step.waiting.length) throw Error('输入或依赖未就绪');\n const media = {};\n for (const [role, ids] of Object.entries(step.bindings)) {\n media[role] = ids.map(id => mediaFor(board.nodes.find(n => n.id === id), jobs));\n if (media[role].some(m => !m || m.local)) throw Error('本机素材尚未上传；先通过已连接素材工具上传，不能将本机路径作为模型输入');\n }\n const first = role => media[role]?.[0];\n const boundTextIds = new Set(step.bindings.text || []);\n const extraText = step.dependencies.filter(id => !boundTextIds.has(id)).map(id => board.nodes.find(x => x.id === id)).filter(x => BY_TYPE.get(operation(x))?.output === 'text').map(x => mediaFor(x, jobs)?.text || '');\n const prompt = [n.prompt, ...(media.text || []).map(x => x.text || ''), ...extraText].filter(Boolean).join('\\n\\n');\n if (step.operation === 'motion_transfer') {\n const value = role => first(role)?.media_id || first(role)?.url;\n if (!first('video')?.media_id) throw Error('Higgsfield 动作视频必须使用确认上传后的 media_id');\n const params = {model: 'hf_mult_motion_control', count: 1, resolution: n.render?.resolution || '720p', medias: [{role:'image', value:value('image')},{role:'video',value:value('video')}], prompt};\n if (billing === 'free_only') params.use_free_gens = true;\n return {provider: 'Higgsfield', tool: 'generate_video', arguments: {params}, preflight: ['balance', 'estimate_video_cost'], billing, duration_semantics: 'source_video_controlled', warning: '先核对真实余额与免费额度；绝不将免费失败改成付费重试'};\n }\n const args = {prompt, duration: n.duration_s || 5, aspect_ratio: n.render?.aspect_ratio || '16:9', resolution: n.render?.resolution || '720p', agentic_prompting: false};\n let tool = 'generate_video', model = 'video/seedance-2.0-mini';\n if (step.operation === 'image_to_video') args.image_url = first('start_image')?.url;\n if (step.operation === 'first_last_video') {model = 'video/minimax-h3-max'; args.resolution = '768P'; args.image_url = first('start_image')?.url; args.last_frame_url = first('end_image')?.url;}\n if (step.operation === 'video_edit') {model = 'video/gemini-omni-flash'; args.video_urls = [first('video')?.url]; args.image_urls = [first('image')?.url];}\n if (['image_generate','image_edit'].includes(step.operation)) {tool = 'generate_image'; model = 'image/nano-banana-2'; delete args.duration; delete args.resolution; delete args.aspect_ratio; args.size = '16:9'; if (step.operation === 'image_edit') args.image_url = first('image')?.url;}\n if (step.operation === 'speech') {tool = 'generate_speech'; model = 'speech/elevenlabs-v3'; for (const k of Object.keys(args)) delete args[k]; args.text = prompt; if (!n.render?.voice_id) throw Error('未选择并试听配音 voice_id'); args.voice_id = n.render.voice_id;}\n if (step.operation === 'sound') {tool = 'generate_audio'; model = 'sfx/elevenlabs-sound-v2'; for (const k of Object.keys(args)) delete args[k]; args.prompt = prompt; args.duration_seconds = n.duration_s || 5;}\n args.model = model;\n if (Object.values(args).some(x => x === undefined) || Object.values(args).some(x => Array.isArray(x) && x.includes(undefined))) throw Error('该后端需要永久 HTTPS 素材地址，不能混用另一后端的 media_id');\n return {provider:'Creative_Claw',tool,arguments:args,preflight:['get_model_params','credit_preflight'],billing,blocked: billing === 'free_only' ? 'NO_CONFIRMED_FREE_ALLOWANCE' : null, warning:'参数草案必须用当前 get_model_params 核对；不得静默替换模型、时长或参考图'};\n}\nfunction node(id, op, title, body, col, row, prompt = '') {\n const s = BY_TYPE.get(op);\n return {id, type: op === 'review' ? 'review' : s.output === 'image' && s.executor === 'input' ? 'character' : s.output === 'audio' ? 'audio' : s.output === 'text' ? 'note' : 'shot', title, body, prompt, x: col * 350, y: row * 240, w: 296, h: 188, color: s.executor === 'human' ? 'green' : s.executor === 'input' ? 'gold' : 'blue', status:'待处理', ...(s.output === 'video' && s.executor === 'connector' ? {duration_s:5} : {}), render:{operation:op,bindings:{}}};\n}\nexport const TEMPLATES = [\n {id:'single-shot',name:'单镜头角色表演',category:'先跑通一镜',description:'角色图 → 真实 I2V → 技术检测 → 视觉验收',build:()=>[\n [node('identity','image','角色身份锚点','确认脸、毛色/肤色、服装和身体比例。',0,0),node('shot','image_to_video','自然表演镜头','先做单角色、固定机位与一个明确动作。',1,0,'Same character and clothing as the first frame. Natural breathing, one gentle blink, a small head turn and believable hair or ear secondary motion. Stable camera, continuous real subject motion, no slideshow.'),node('qc','technical_qc','检测实际视频','全量解码与帧率检查不等于美术合格。',2,0),node('review','review','自然度人工验收','通过标准：没有畸形、滑步、换脸与僵硬表情。',3,0)], [['identity','shot'],['shot','qc'],['qc','review']]\n ]},\n {id:'manga-drama',name:'AI 漫剧 · 连续两镜',category:'叙事',description:'人物 / 美术锁定 → 第一镜 → 验收 → 真实末帧 → 第二镜 → 合成',build:()=>[\n [node('identity','image','人物与服装锁定','同一份批准的角色参考；不用无服装转面图替代。',0,0),node('style','style','美术设定','清晰线条、克制高光、统一光线和色彩；禁止中途换装。',0,1),node('s1','image_to_video','S1 · 动作建立','固定镜头，让角色完成一个清楚的动作意图。',1,0,'Cinematic anime, preserve identity and outfit. The character notices something offscreen and turns their head with natural anticipation. Continuous acting, no text.'),node('r1','review','S1 验收','批准后才提取末帧。',2,0),node('last','last_frame','S1 实际末帧','从通过的输出视频最后一个可解码帧提取。',3,0),node('s2','image_to_video','S2 · 动作承接','接续姿态，不能重置角色身体。',4,0,'Continue the action from the supplied first frame. Same identity, clothing, scene direction and lighting. Settle naturally and respond with a subtle expression.'),node('r2','review','S2 验收','核对动作承接和空间方向。',5,0),node('final','assemble','批准镜头成片','只接纳批准的真实视频。',6,0)], [['identity','s1'],['style','s1'],['s1','r1'],['r1','last'],['last','s2'],['style','s2'],['s2','r2'],['r1','final'],['r2','final']]\n ]},\n {id:'motion-transfer',name:'动作迁移 · 角色替换',category:'表演',description:'正式角色图 + 真实动作视频 → Genjutsu → 对照检测 → 人工验收',build:()=>[\n [node('identity','image','正式角色图','只使用一个确定的身份和服装版本。',0,0),node('driver','video','真实动作参考','输入时长就是动作迁移基准，不虚设输出长度。',0,1),node('motion','motion_transfer','角色动作迁移','严格免费模式；免费不适用时停止，不改成扣费。',1,0,'Transfer the source choreography to the exact reference character. Preserve appearance, clothing, source timing, formation, contact and camera. Natural articulated motion, no geometry proxy or slideshow.'),node('qc','technical_qc','技术 / 时序检查','来源和输出分别检查；全局帧差不能证明迁移正确。',2,0),node('review','review','表演与身份验收','核对动作节奏、重心、脚底接触、服装和面部。',3,0)], [['identity','motion'],['driver','motion'],['motion','qc'],['qc','review']]\n ]},\n {id:'art-transition',name:'艺术短片 · 首尾变形',category:'艺术',description:'首帧 + 目标尾帧 → 生成式过渡 → 调色 → 审看',build:()=>[\n [node('start','image','起始艺术画面','明确材质、主体和视觉中心。',0,0),node('end','image','目标尾帧','保持构图可衔接；不是交叉淡化素材。',0,1),node('transform','first_last_video','生成式材质转场','使用支持首尾帧的模型；不使用补帧伪造。',1,0,'A controlled material transformation from the first image into the last. Continuous physical unfolding and coherent silhouettes, deliberate motion and stable framing. No dissolve or slide transition.'),node('grade','grade','统一色彩','调色只改变色彩，不增加主体动作。',2,0),node('review','review','形变美术验收','拒绝脸部融化、纹理爬动和无动机闪烁。',3,0)], [['start','transform'],['end','transform'],['transform','grade'],['grade','review']]\n ]},\n {id:'music-visual',name:'音乐视觉 · 双段节奏',category:'艺术 / MV',description:'分段提示词 → 独立视频 → 候选审看 → 合成 → 音乐混音',build:()=>[\n [node('brief','text','节奏与视觉意图','第一段缓慢聚拢，第二段开放；时长与已获授权音乐匹配。',0,0),node('music','audio','已授权音乐','导入真实音频，禁止把声音模板当已生成音轨。',0,1),node('a','text_to_video','A · 蓄势','一个明确运动中心，不堆满噪声和特效。',1,0,'Abstract sculptural paper forms slowly gather under soft directional light. Genuine continuous deformation, tactile material and deliberate pacing, no slideshow.'),node('b','text_to_video','B · 释放','与 A 保持材质和色彩一致。',1,1,'The same abstract sculptural paper forms unfold outward gracefully. Cohesive tactile material, restrained palette and real continuous motion.'),node('ra','review','A 审看','单段先通过审美验收。',2,0),node('rb','review','B 审看','单段先通过审美验收。',2,1),node('cut','assemble','两段合成','连接批准视频，不自动添加未授权素材。',3,0),node('mix','mix','音乐混音','按时长保留视频；这一步不是口型同步。',4,0)], [['brief','a'],['brief','b'],['a','ra'],['b','rb'],['ra','cut'],['rb','cut'],['cut','mix'],['music','mix']]\n ]},\n {id:'art-restyle',name:'实拍转艺术 · 保留时序',category:'风格迁移',description:'真实视频 + 风格图 → 视频重绘 → 技术检测 → 视觉验收',build:()=>[\n [node('footage','video','原始视频','保留原片时序，确认人物与素材使用权限。',0,0),node('look','image','艺术风格参考','用线条、笔触和材质引导，不篡改动作。',0,1),node('restyle','video_edit','视频艺术重绘','先校验模型允许的输入长度与参考组合。',1,0,'Restyle the source video using the visual language of the reference image. Preserve source action timing, camera movement, geometry and subject identity. Stable temporal brushwork, no flicker.'),node('qc','technical_qc','时序与重复帧筛查','检测只作为筛查；需进一步审看片。',2,0),node('review','review','风格稳定验收','笔触稳定，动作节奏不失真，主体不融化。',3,0)], [['footage','restyle'],['look','restyle'],['restyle','qc'],['qc','review']]\n ]}\n];\nexport function template(id = 'single-shot') {\n const t = TEMPLATES.find(t => t.id === id); if (!t) throw Error('未知工作流模板');\n const [nodes, links] = t.build();\n if (id === 'art-transition') {nodes.find(n=>n.id==='transform').render.bindings={start_image:'start',end_image:'end'};}\n return {version:1,id:`workflow-${id}`,title:t.name,revision:0,view:{x:30,y:30,scale:1},nodes,edges:links.map(([from,to],i)=>({id:`e-${i}`,from,to}))};\n}\nexport const ADAPTERS = {\n mode:'connector-assisted',\n paid_submissions:'require a fresh provider estimate and explicit budget; no silent paid fallback',\n generation:['Higgsfield.generate_video', 'Creative_Claw.generate_video', 'Creative_Claw.generate_image', 'Creative_Claw.generate_speech', 'Creative_Claw.generate_audio'],\n media_worker:['last_frame','trim','grade','technical_qc','assemble','mix'],\n restrictions:['A provider job ID is not a playable video.', 'Do not retry a submitted or uncertain request blindly.', 'No automatic visual approval.', 'No frame-sequence or low-fps proxy fallback.', 'No independent background generator is deployed in this gateway.'],\n schema_rule:'Read the selected provider model schema before submission; preserve reference order and literal-frame semantics.'\n};\n","deleted","label",[130],"ac79a21e-5744-449b-9d41-08a5550b186e","run-store.js","5d6e656c-ae52-11f1-930a-1607ee4eb77e",["D",1789181312318],3,"import { sqlite } from 'https://esm.town/v/std/sqlite/main.ts';\nimport { compile, connectorRequest, mediaFor, operation, safeUrl, NODE_TYPES } from './workflow-core.js';\nconst MAX_BYTES = 192 * 1024;\nconst digest = async text => [...new Uint8Array(await crypto.subtle.digest('SHA-256', new TextEncoder().encode(text)))].map(x => x.toString(16).padStart(2, '0')).join('');\nconst token = () => [...crypto.getRandomValues(new Uint8Array(32))].map(x => x.toString(16).padStart(2, '0')).join('');\nlet initialized;\nasync function init() {\n if (!initialized) initialized = sqlite.execute('CREATE TABLE IF NOT EXISTS canvas_runs_v3 (id TEXT PRIMARY KEY, key_hash TEXT NOT NULL, revision INTEGER NOT NULL, document TEXT NOT NULL, created_at TEXT NOT NULL)');\n try { await initialized; } catch (e) { initialized = null; throw e; }\n}\nfunction bytes(value) { return new TextEncoder().encode(JSON.stringify(value)).length; }\nfunction clean(run) {\n const value = JSON.parse(JSON.stringify(run));\n for (const j of Object.values(value.jobs)) delete j.lease_token;\n value.plan = compile(value.board, value.jobs);\n return value;\n}\nasync function load(args) {\n if (!/^[0-9a-f-]{36}$/i.test(args.run_id || '') || !/^[0-9a-f]{64}$/.test(args.run_key || '')) throw Error('RUN_ACCESS_DENIED');\n await init();\n const result = await sqlite.execute({sql:'SELECT document, revision FROM canvas_runs_v3 WHERE id = ? AND key_hash = ?', args:[args.run_id, await digest(args.run_key)]});\n if (!result.rows.length) throw Error('RUN_ACCESS_DENIED');\n const run = JSON.parse(result.rows[0].document); run.revision = Number(result.rows[0].revision); return run;\n}\nasync function save(args, run) {\n if (!Number.isInteger(args.expected_revision) || args.expected_revision !== run.revision) throw Error('REVISION_CONFLICT: read run and retry');\n const expected = run.revision; run.revision++; run.updated_at = new Date().toISOString();\n if (bytes(run) > MAX_BYTES) throw Error('RUN_SIZE_LIMIT');\n const result = await sqlite.execute({sql:'UPDATE canvas_runs_v3 SET document = ?, revision = ? WHERE id = ? AND key_hash = ? AND revision = ?',args:[JSON.stringify(run),run.revision,run.id,await digest(args.run_key),expected]});\n if (result.rowsAffected !== 1) throw Error('REVISION_CONFLICT: concurrent update rejected');\n return clean(run);\n}\nexport async function createRun(args) {\n const plan = compile(args.board);\n if (!plan.valid || args.board.nodes.length > 128) throw Error(plan.errors.join('; ') || 'A run supports at most 128 nodes');\n if (args.consent !== true) throw Error('CONSENT_REQUIRED: workflow text and source URLs will be stored remotely');\n if (bytes(args.board) > 120 * 1024) throw Error('BOARD_TOO_LARGE: keep media at its provider and store HTTPS references, not base64');\n if (args.billing !== undefined && args.billing !== 'free_only') throw Error('This release only dispatches confirmed free model requests. Paid generation requires a separate reviewed adapter.');\n const id = crypto.randomUUID(), key = token(), now = new Date().toISOString();\n const run = {id,revision:0,created_at:now,updated_at:now,state:'active',mode:'chatgpt_connector_assisted',billing:'free_only',board:args.board,jobs:{},events:[{at:now,event:'run_created',note:'No provider job submitted by creation.'}]};\n await init();\n // Bound storage use on the public gateway; writes never overwrite another run.\n const inserted = await sqlite.execute({sql:'INSERT INTO canvas_runs_v3 (id,key_hash,revision,document,created_at) SELECT ?,?,?,?,? WHERE (SELECT COUNT(*) FROM canvas_runs_v3) < 40',args:[id,await digest(key),0,JSON.stringify(run),now]});\n if (inserted.rowsAffected !== 1) throw Error('RUN_CAPACITY_LIMIT: export existing runs before creating more');\n return {...clean(run),run_key:key,privacy:'Run key grants access. Keep it private; it is not a provider API key.'};\n}\nexport async function readRun(args) { return clean(await load(args)); }\nexport async function nextTasks(args) {\n const run = await load(args); if (run.state !== 'active') return {run_id:run.id,state:run.state,tasks:[]};\n const plan = compile(run.board, run.jobs), tasks = [];\n for (const step of plan.steps.filter(s => ['ready','needs_review','review','submitted','queued','in_progress','leased'].includes(s.state))) {\n let request = null, blocked = null;\n if (step.state === 'ready') {\n try {\n if (step.executor === 'connector') request = connectorRequest(run.board,step,run.jobs,run.billing);\n else if (['media_worker','local'].includes(step.executor)) request = {executor:'canvas_media_worker.py',source_tool:'canvas_get_media_worker',operation:step.operation,inputs:Object.fromEntries(Object.entries(step.bindings).map(([k,ids]) => [k,ids.map(id => ({...mediaFor(run.board.nodes.find(n=>n.id===id),run.jobs),...(run.jobs[id]?.review ? {review:{...run.jobs[id].review,decision:run.jobs[id].state}} : {})}))])),config:run.board.nodes.find(n=>n.id===step.id).render || {}};\n } catch(e) { blocked = e.message; }\n }\n tasks.push({...step,request,blocked,receipt:run.jobs[step.id]?.provider_job_id || null});\n }\n return {run_id:run.id,revision:run.revision,mode:run.mode,tasks,notice:'The ChatGPT executor must perform provider/worker calls and record actual receipts; this endpoint does not launch a background worker.'};\n}\nexport async function claimTask(args) {\n const run = await load(args); if (run.state !== 'active') throw Error('RUN_NOT_ACTIVE');\n if (!/^[A-Za-z0-9_-]{8,96}$/.test(args.request_id || '')) throw Error('REQUEST_ID_REQUIRED');\n const previous = run.jobs[args.node_id];\n if (previous) {\n if (previous.request_id === args.request_id) return {run_id:run.id,revision:run.revision,node_id:args.node_id,lease_token:previous.lease_token,provider_job_id:previous.provider_job_id || null,state:previous.state,idempotent:true,can_submit:false,action:previous.provider_job_id?'poll_existing_job_only':'reconcile_unknown_submission_before_any_retry'};\n throw Error('TASK_ALREADY_CLAIMED_OR_SUBMITTED: inspect its receipt; no blind retry');\n }\n const plan = compile(run.board,run.jobs), step = plan.steps.find(s=>s.id===args.node_id);\n if (!plan.valid || !step || step.state !== 'ready' || step.executor === 'human') throw Error('TASK_NOT_READY');\n let request;\n if (step.executor === 'connector') {request=connectorRequest(run.board,step,run.jobs,run.billing); if (request.blocked) throw Error(request.blocked);}\n const lease = token();\n run.jobs[args.node_id] = {state:'leased',request_id:args.request_id,lease_token:lease,claimed_at:new Date().toISOString(),request:request || null};\n run.events.push({at:new Date().toISOString(),event:'task_claimed',node_id:args.node_id});\n const saved = await save(args,run);\n return {run_id:run.id,revision:saved.revision,node_id:args.node_id,lease_token:lease,request:request || {executor:step.executor,operation:step.operation,bindings:step.bindings},notice:'Lease persisted before provider submission. A lost/unknown submission must be reconciled, never repeated automatically.'};\n}\nexport async function recordTask(args) {\n const run = await load(args), job = run.jobs[args.node_id];\n if (!job || !args.lease_token || job.lease_token !== args.lease_token) throw Error('LEASE_ACCESS_DENIED');\n if (run.state !== 'active') throw Error('RUN_NOT_ACTIVE: provider receipt remains external; no downstream execution');\n const states = ['submitted','queued','in_progress','review','completed','failed','blocked'];\n if (!states.includes(args.state)) throw Error('INVALID_TASK_STATE');\n const op = operation(run.board.nodes.find(n=>n.id===args.node_id)), spec = NODE_TYPES.find(s=>s.id===op);\n if (['approved','rejected','completed','failed'].includes(job.state)) throw Error('TERMINAL_TASK: create a new run for a deliberate retry');\n const allowed = {leased:['submitted','queued','in_progress','review','completed','failed','blocked'],submitted:['submitted','queued','in_progress','review','failed','blocked'],queued:['queued','in_progress','review','failed','blocked'],in_progress:['in_progress','review','failed','blocked'],review:['review','failed'],blocked:['blocked','failed']}[job.state] || [];\n if (!allowed.includes(args.state)) throw Error('INVALID_STATE_TRANSITION');\n if (spec.executor === 'connector' && args.state === 'completed') throw Error('VISUAL_REVIEW_REQUIRED: model output must enter review, never auto-complete');\n if (spec.executor === 'connector' && ['submitted','queued','in_progress','review'].includes(args.state) && (!args.provider_job_id || typeof args.provider_job_id !== 'string' || args.provider_job_id.length > 200)) throw Error('REAL_PROVIDER_JOB_ID_REQUIRED');\n if (job.provider_job_id && args.provider_job_id && job.provider_job_id !== args.provider_job_id) throw Error('PROVIDER_JOB_ID_MISMATCH');\n if (['review','completed'].includes(args.state)) {\n if (!args.output || args.output.kind !== spec.output || !(safeUrl(args.output.url) || spec.output === 'text' && typeof args.output.text === 'string')) throw Error('VALID_OUTPUT_REQUIRED');\n }\n if (args.output?.url && !safeUrl(args.output.url)) throw Error('UNSAFE_OUTPUT_URL');\n if (typeof args.note !== 'string' || args.note.length > 4000) throw Error('EVIDENCE_NOTE_REQUIRED');\n Object.assign(job,{state:args.state,checked_at:new Date().toISOString(),note:args.note,evidence_source:'executor_report'});\n if (args.provider_job_id) job.provider_job_id=args.provider_job_id;\n if (args.output) job.output=args.output;\n if (args.qa) {if (bytes(args.qa)>16000) throw Error('QA_TOO_LARGE'); job.qa=args.qa;}\n run.events.push({at:job.checked_at,event:'provider_or_worker_receipt',node_id:args.node_id,state:args.state});\n if (run.events.length > 100) run.events=run.events.slice(-100);\n return save(args,run);\n}\nexport async function reviewTask(args) {\n const run=await load(args); if (run.state!=='active') throw Error('RUN_NOT_ACTIVE');\n if (args.user_confirmed!==true || !['approved','rejected'].includes(args.decision) || typeof args.note!=='string' || args.note.trim().length<4) throw Error('EXPLICIT_USER_VISUAL_REVIEW_REQUIRED');\n const n=run.board.nodes.find(n=>n.id===args.node_id); if(!n) throw Error('NODE_NOT_FOUND');\n let job=run.jobs[n.id];\n if (operation(n)==='review' && !job) {\n const step=compile(run.board,run.jobs).steps.find(s=>s.id===n.id);\n if(step.waiting.length || step.issues.length) throw Error('REVIEW_INPUT_NOT_READY');\n const inputId=step.bindings.video?.[0], output=mediaFor(run.board.nodes.find(x=>x.id===inputId),run.jobs);\n if(!output?.url || !safeUrl(output.url)) throw Error('NO_PLAYABLE_VIDEO');\n job={state:'review',output}; run.jobs[n.id]=job;\n }\n if(!job || job.state!=='review') throw Error('TASK_NOT_AWAITING_REVIEW');\n job.state=args.decision; job.review={by:'user',note:args.note,at:new Date().toISOString()};\n run.events.push({at:job.review.at,event:'visual_review',node_id:n.id,decision:args.decision});\n return save(args,run);\n}\nexport async function cancelRun(args) {\n const run=await load(args); run.state='cancelled';\n const external=Object.values(run.jobs).filter(j=>j.provider_job_id&&!['approved','rejected','failed','completed'].includes(j.state)).map(j=>j.provider_job_id);\n run.events.push({at:new Date().toISOString(),event:'run_cancelled',note:'Does not cancel provider jobs or guarantee refunds.'});\n return {...await save(args,run),provider_jobs_not_cancelled:external};\n}\n",[148],"661001e2-f8a4-4e00-bbd3-3bd6f6b902d7","media-worker-source.js","c5ce5dbe-ae53-11f1-b6ce-1607ee4eb77e",9,["D",1789181024920],"// Runnable standard-library Python worker. No model inference or optical-flow substitution.\nexport const WORKER_SOURCE = String.raw`#!/usr/bin/env python3\n\"\"\"Infinite Canvas media worker: process existing video, never fabricate model output.\"\"\"\nimport argparse\nimport hashlib\nimport ipaddress\nimport json\nimport math\nfrom pathlib import Path\nimport shutil\nimport socket\nimport subprocess\nimport urllib.error\nimport urllib.parse\nimport urllib.request\n\nLIMIT = 256 * 1024 * 1024\nDEFAULT_HOSTS = {'cdn.creativeclaw.co', 'd2ol7oe51mr4n9.cloudfront.net'}\n\nclass NoRedirect(urllib.request.HTTPRedirectHandler):\n def redirect_request(self, req, fp, code, msg, headers, newurl):\n return None\n\ndef execute(args, timeout=95):\n result = subprocess.run(args, capture_output=True, timeout=timeout, check=False)\n if result.returncode:\n raise ValueError('Media operation failed: ' + result.stderr.decode(errors='replace')[-1200:])\n return result\n\ndef number(value, minimum, maximum):\n if isinstance(value, bool) or not isinstance(value, (int, float)) or not math.isfinite(value) or not minimum <= value <= maximum:\n raise ValueError('Invalid numeric media parameter')\n return value\n\ndef probe(path):\n result = execute(['ffprobe', '-v', 'error', '-protocol_whitelist', 'file,pipe', '-count_frames', '-show_streams', '-show_format', '-of', 'json', str(path)])\n data = json.loads(result.stdout)\n duration = float(data.get('format', {}).get('duration', 0))\n if not 0 < duration <= 300:\n raise ValueError('Source duration must be >0 and <=300 seconds')\n return data\n\ndef fetch_source(media, target, hosts):\n if media.get('path'):\n path = Path(media['path']).expanduser().resolve()\n if not path.is_file() or not 0 < path.stat().st_size <= LIMIT:\n raise ValueError('Invalid local source')\n shutil.copyfile(path, target)\n return target\n url = media.get('url', '')\n opener = urllib.request.build_opener(NoRedirect)\n for _ in range(6):\n parsed = urllib.parse.urlsplit(url)\n if parsed.scheme != 'https' or parsed.username or parsed.password or parsed.fragment or parsed.hostname not in hosts or parsed.port not in (None, 443):\n raise ValueError('Source must be an approved HTTPS media host')\n addresses = socket.getaddrinfo(parsed.hostname, 443, type=socket.SOCK_STREAM)\n if not addresses or any(not ipaddress.ip_address(item[4][0]).is_global for item in addresses):\n raise ValueError('Private or reserved network source rejected')\n try:\n response = opener.open(urllib.request.Request(url, headers={'User-Agent':'InfiniteCanvasMediaWorker/0.3'}), timeout=30)\n except urllib.error.HTTPError as error:\n if error.code in (301,302,303,307,308):\n url = urllib.parse.urljoin(url, error.headers.get('Location', ''))\n continue\n raise ValueError('Source download HTTP ' + str(error.code)) from None\n with response:\n if int(response.headers.get('Content-Length', '0')) > LIMIT:\n raise ValueError('Source exceeds 256 MiB')\n size = 0\n with target.open('wb') as out:\n while True:\n chunk = response.read(1024 * 1024)\n if not chunk:\n break\n size += len(chunk)\n if size > LIMIT:\n raise ValueError('Source exceeds 256 MiB')\n out.write(chunk)\n if not size:\n raise ValueError('Empty source')\n return target\n raise ValueError('Too many media redirects')\n\ndef video_stream(info):\n stream = next((s for s in info['streams'] if s['codec_type'] == 'video'), None)\n if not stream:\n raise ValueError('Input has no video stream')\n return stream\n\ndef fps(stream):\n pieces = stream.get('avg_frame_rate', '0/1').split('/')\n return float(pieces[0]) / float(pieces[1]) if len(pieces) == 2 and float(pieces[1]) else 0\n\ndef ffmpeg(inputs, tail):\n args = ['ffmpeg', '-v', 'error', '-nostdin', '-y', '-threads', '2']\n for path in inputs:\n args += ['-protocol_whitelist', 'file,pipe', '-i', str(path)]\n execute(args + tail)\n\ndef encode(out):\n return ['-c:v','libx264','-preset','fast','-crf','18','-pix_fmt','yuv420p','-c:a','aac','-b:a','160k','-movflags','+faststart',str(out)]\n\ndef quality(path):\n info = probe(path)\n stream = video_stream(info)\n decoded = execute(['ffmpeg','-v','error','-nostdin','-threads','2','-protocol_whitelist','file,pipe','-i',str(path),'-map','0:v:0','-vf','scale=64:36','-vsync','0','-pix_fmt','gray','-f','rawvideo','pipe:1']).stdout\n area = 64 * 36\n if len(decoded) % area:\n raise ValueError('Incomplete decoded frame')\n count = len(decoded) // area\n if count < 2:\n raise ValueError('Not a multi-frame video')\n differences = []\n for i in range(1, count):\n a = decoded[(i-1)*area:i*area]\n b = decoded[i*area:(i+1)*area]\n differences.append(sum(abs(x-y) for x,y in zip(a,b)) / (area*255))\n frozen = sum(x < .001 for x in differences)\n rate = fps(stream)\n return {'decoded_frames':count,'duration_s':float(info['format']['duration']),'width':stream['width'],'height':stream['height'],'fps':rate,'codec':stream.get('codec_name'),'audio_streams':sum(s['codec_type']=='audio' for s in info['streams']),'low_difference_pairs':frozen,'frame_pairs':len(differences),'mean_frame_difference':sum(differences)/len(differences),'minimum_frame_difference':min(differences),'technical_warnings':(['LOW_FPS_NOT_FORMAL_ANIMATION'] if rate<20 else []) + (['REPEATED_OR_LOW_MOTION_FRAMES'] if frozen else []),'visual_approval':False,'subject_motion_verified':False,'note':'Whole-frame differences are a screening metric only. They do not prove acting, identity, physical contact, or successful motion transfer.'}\n\ndef main():\n parser = argparse.ArgumentParser()\n parser.add_argument('--request', required=True)\n parser.add_argument('--out-dir', required=True)\n parser.add_argument('--allow-host', action='append', default=[])\n options = parser.parse_args()\n task = json.loads(Path(options.request).read_text())\n operation = task.get('operation')\n if operation not in {'prompt','last_frame','trim','grade','technical_qc','assemble','mix'}:\n raise ValueError('Unsupported operation; no frame-sequence or fake-video fallback exists')\n outdir = Path(options.out_dir).resolve()\n outdir.mkdir(parents=True, exist_ok=True)\n inputs = task.get('inputs', {})\n config = task.get('config', {})\n if operation == 'prompt':\n text = '\\n\\n'.join(v.get('text', '') for v in inputs.get('text', []))\n if not text.strip():\n raise ValueError('Text input required')\n report = {'operation':operation,'output':{'kind':'text','text':text},'generation':False}\n else:\n hosts = DEFAULT_HOSTS | set(options.allow_host)\n sources = inputs.get('video', [])\n if not sources or len(sources) > 8:\n raise ValueError('Provide 1–8 actual video inputs')\n if operation == 'assemble' and any(v.get('review', {}).get('decision') != 'approved' or v.get('review', {}).get('by') != 'user' for v in sources):\n raise ValueError('ASSEMBLY_REQUIRES_EXPLICIT_USER_APPROVAL_FOR_EVERY_CLIP')\n paths = [fetch_source(v, outdir/('source-'+str(i)+'.mp4'), hosts) for i,v in enumerate(sources)]\n infos = [probe(p) for p in paths]\n for info in infos:\n video_stream(info)\n duration = float(infos[0]['format']['duration'])\n out = outdir/'output.mp4'\n kind = 'video'\n if operation == 'last_frame':\n frames = int(video_stream(infos[0]).get('nb_read_frames', 0))\n if frames < 1:\n raise ValueError('Unable to count actual frames')\n out = outdir/'last-frame.png'\n ffmpeg([paths[0]], ['-vf','select=eq(n\\\\,'+str(frames-1)+')','-vsync','0','-frames:v','1',str(out)])\n kind = 'image'\n elif operation == 'trim':\n start = number(config.get('start_s',0),0,duration)\n length = number(config.get('duration_s',min(5,duration-start)),.04,duration)\n if start + length > duration + .001:\n raise ValueError('Trim exceeds source; no looping or padding with still frames')\n ffmpeg([paths[0]], ['-ss',str(start),'-t',str(length),'-map','0:v:0','-map','0:a?'] + encode(out))\n elif operation == 'grade':\n contrast=number(config.get('contrast',1.0),.5,2)\n brightness=number(config.get('brightness',0),-.3,.3)\n saturation=number(config.get('saturation',1.0),0,2)\n ffmpeg([paths[0]], ['-vf',f'eq=contrast={contrast}:brightness={brightness}:saturation={saturation}','-map','0:v:0','-map','0:a?']+encode(out))\n elif operation == 'technical_qc':\n shutil.copyfile(paths[0],out)\n elif operation == 'assemble':\n stream=video_stream(infos[0])\n width,height=stream['width'],stream['height']\n rate=fps(stream)\n if rate<20 or any(fps(video_stream(i))<20 for i in infos):\n raise ValueError('Low-frame-rate proxy clips are not accepted for formal assembly')\n if sum(float(i['format']['duration']) for i in infos)>300:\n raise ValueError('Assembly exceeds 300 seconds')\n filters=[]\n for i,info in enumerate(infos):\n filters.append(f'[{i}:v:0]setpts=PTS-STARTPTS,scale={width}:{height}:force_original_aspect_ratio=decrease,pad={width}:{height}:(ow-iw)/2:(oh-ih)/2,setsar=1,fps={rate}[v{i}]')\n if any(s['codec_type']=='audio' for s in info['streams']):\n filters.append(f'[{i}:a:0]asetpts=PTS-STARTPTS,aresample=48000,aformat=channel_layouts=stereo[a{i}]')\n else:\n seconds=float(info['format']['duration'])\n filters.append(f'anullsrc=r=48000:cl=stereo,atrim=duration={seconds}[a{i}]')\n filters.append(''.join(f'[v{i}][a{i}]' for i in range(len(paths)))+f'concat=n={len(paths)}:v=1:a=1[v][a]')\n ffmpeg(paths,['-filter_complex',';'.join(filters),'-map','[v]','-map','[a]']+encode(out))\n elif operation == 'mix':\n audio=inputs.get('audio',[])\n if len(audio)!=1:\n raise ValueError('Exactly one audio input required')\n track=fetch_source(audio[0],outdir/'track.audio',hosts)\n audio_info=probe(track)\n if not any(s['codec_type']=='audio' for s in audio_info['streams']):\n raise ValueError('Source has no audio stream')\n delay=round(number(config.get('delay_s',0),0,duration)*1000)\n gain=number(config.get('gain',1),0,2)\n filters=f'[1:a:0]volume={gain},adelay={delay}:all=1,apad[voice]'\n if any(s['codec_type']=='audio' for s in infos[0]['streams']):\n filters+=';[0:a:0][voice]amix=inputs=2:duration=first:normalize=0[mix]'\n label='[mix]'\n else:\n label='[voice]'\n ffmpeg([paths[0],track],['-filter_complex',filters,'-map','0:v:0','-map',label,'-t',str(duration)]+encode(out))\n if not out.is_file() or not out.stat().st_size:\n raise ValueError('No media output produced')\n report={'operation':operation,'generation':False,'output':{'kind':kind,'local_path':str(out),'sha256':hashlib.sha256(out.read_bytes()).hexdigest(),'bytes':out.stat().st_size},'provenance':'Existing-video processing; not a newly AI-generated clip','visual_approval':False}\n if kind=='video':\n report['qa']=quality(out)\n (outdir/'result.json').write_text(json.dumps(report,ensure_ascii=False,indent=2))\n print(json.dumps(report,ensure_ascii=False))\n\nif __name__ == '__main__':\n try:\n main()\n except Exception as error:\n print(json.dumps({'ok':False,'error':str(error)},ensure_ascii=False))\n raise SystemExit(1)\n`;\nexport async function workerGuide() {\n const bytes = new TextEncoder().encode(WORKER_SOURCE);\n const sha256 = [...new Uint8Array(await crypto.subtle.digest('SHA-256', bytes))].map(x=>x.toString(16).padStart(2,'0')).join('');\n return {filename:'canvas_media_worker.py',sha256,source:WORKER_SOURCE,requires:['python3','ffmpeg','ffprobe'],operations:['prompt','last_frame','trim','grade','technical_qc','assemble','mix'],command:'python3 canvas_media_worker.py --request request.json --out-dir output',boundary:'Processes existing media. Does not create AI footage, call a paid API or approve visual quality.'};\n}\n",[155],"ed251f7c-64ad-4260-a7b8-9cf344deb548","main.ts",12,["D",1789181061626],5,"http","import { VERSION, NODE_TYPES, TEMPLATES, SOURCES, ADAPTERS, template, compile } from './workflow-core.js';\nimport { createRun, readRun, nextTasks, claimTask, recordTask, reviewTask, cancelRun } from './run-store.js';\nimport { workerGuide } from './media-worker-source.js';\nconst WEB_ENTRY = 'https://app-cl8wbg.v2.appdeploy.ai/';\nconst PROTOCOLS = new Set(['2025-06-18','2025-03-26','2024-11-05']);\nconst SERVER = {name:'Infinite Canvas · 无限画布',version:VERSION};\nconst string = {type:'string'};\nconst object = {type:'object'};\nconst access = {run_id:string,run_key:{type:'string',minLength:64,maxLength:64,description:'Private per-run access key, returned only by canvas_create_run. Never put provider credentials here.'}};\nconst revision = {expected_revision:{type:'integer',minimum:0}};\nconst tool = (name:string,description:string,properties:object={},required:string[]=[],read=true) => ({name,description,inputSchema:{type:'object',properties,required,additionalProperties:false},annotations:{readOnlyHint:read,destructiveHint:name==='canvas_cancel_run',idempotentHint:read,openWorldHint:false}});\nconst tools = [\n tool('canvas_probe','Check real service capabilities. Does not imply that a video provider is funded or available.'),\n tool('canvas_get_template','Return a usable version-1 workflow; choose template_id from canvas_list_templates.',{template_id:string}),\n tool('canvas_validate','Validate the portable board and its execution graph; reports cycle, binding and missing-input errors.',{board:object},['board']),\n tool('canvas_open_web','Return the editable browser canvas entry. This does not read the user browser storage.'),\n tool('canvas_list_templates','List six manga, art and motion workflow recipes and their reference sources.'),\n tool('canvas_list_nodes','List typed input, generation, media-processing and human-review nodes.'),\n tool('canvas_compile_workflow','Compile a board into ordered executable tasks without submitting or charging.',{board:object},['board']),\n tool('canvas_get_adapter_guides','Read executor contracts, provider preflight requirements and quality boundaries.'),\n tool('canvas_get_media_worker','Retrieve the actual Python/FFmpeg worker with its SHA-256; supports real trim, last-frame extraction, grade, QA, assembly and mixing. No AI footage fabrication.'),\n tool('canvas_create_run','With user consent, persist a workflow snapshot and private run key. Creation does NOT submit model jobs.',{board:object,consent:{type:'boolean'},billing:{type:'string',enum:['free_only']}},['board','consent'],false),\n tool('canvas_read_run','Read one private run, actual recorded receipts and freshness timestamps.',access,['run_id','run_key']),\n tool('canvas_next_tasks','Return ready connector or media-worker tasks. The ChatGPT executor must perform them and record real receipts.',access,['run_id','run_key']),\n tool('canvas_claim_task','Reserve a ready task before external execution. Never blindly re-submit an existing or uncertain provider request.',{...access,...revision,node_id:string,request_id:string},['run_id','run_key','expected_revision','node_id','request_id'],false),\n tool('canvas_record_task','Record a real provider/worker receipt using the private run and task keys. Model outputs enter review, never automatic approval.',{...access,...revision,node_id:string,lease_token:string,state:{type:'string',enum:['submitted','queued','in_progress','review','completed','failed','blocked']},provider_job_id:string,output:object,qa:object,note:string},['run_id','run_key','expected_revision','node_id','lease_token','state','note'],false),\n tool('canvas_review_task','Record explicit USER visual approval or rejection. Do not call approval merely because metadata or pixel tests pass.',{...access,...revision,node_id:string,user_confirmed:{type:'boolean'},decision:{type:'string',enum:['approved','rejected']},note:string},['run_id','run_key','expected_revision','node_id','user_confirmed','decision','note'],false),\n tool('canvas_cancel_run','Stop downstream execution of one private run. Does not cancel already submitted provider jobs or promise a refund.',{...access,...revision},['run_id','run_key','expected_revision'],false)\n];\nconst handlers:Recordany> = {\n canvas_probe:()=>({ok:true,server:SERVER,toolCount:tools.length,templateCount:TEMPLATES.length,nodeTypeCount:NODE_TYPES.length,webEntry:WEB_ENTRY,transport:'streamable-http-json',execution_mode:'chatgpt_connector_assisted',provider_credentials_stored:false,unattended_worker:false,privacy:'Public recipes; per-run 256-bit capability keys and revision-guarded private state.'}),\n canvas_get_template:a=>template(a.template_id),\n canvas_validate:a=>compile(a.board),\n canvas_open_web:()=>({url:WEB_ENTRY,note:'Browser editor and execution console; model calls are performed by the connected ChatGPT executor.'}),\n canvas_list_templates:()=>({templates:TEMPLATES.map(({build,...t})=>({...t,node_count:build()[0].length})),sources:SOURCES}),\n canvas_list_nodes:()=>({nodes:NODE_TYPES}),\n canvas_compile_workflow:a=>compile(a.board),\n canvas_get_adapter_guides:()=>ADAPTERS,\n canvas_get_media_worker:workerGuide,\n canvas_create_run:createRun,canvas_read_run:readRun,canvas_next_tasks:nextTasks,canvas_claim_task:claimTask,canvas_record_task:recordTask,canvas_review_task:reviewTask,canvas_cancel_run:cancelRun\n};\nconst headers = {'content-type':'application/json; charset=utf-8','access-control-allow-origin':'*','cache-control':'no-store','x-content-type-options':'nosniff'};\nfunction json(value:unknown,status=200) {return new Response(JSON.stringify(value),{status,headers});}\nfunction error(id:unknown,code:number,message:string,status=200) {return json({jsonrpc:'2.0',id,error:{code,message}},status);}\nasync function body(req:Request) {\n if (Number(req.headers.get('content-length') || 0)>262144) throw Error('REQUEST_TOO_LARGE');\n const reader=req.body?.getReader(); if(!reader) return '';\n const chunks:Uint8Array[]=[]; let total=0;\n while(true) {const {done,value}=await reader.read(); if(done) break; total+=value.byteLength; if(total>262144){await reader.cancel();throw Error('REQUEST_TOO_LARGE');} chunks.push(value);}\n const data=new Uint8Array(total); let at=0; for(const part of chunks){data.set(part,at);at+=part.length;} return new TextDecoder().decode(data);\n}\nexport default async function(req:Request):Promise {\n const url=new URL(req.url);\n if(req.method==='OPTIONS') return new Response(null,{status:204,headers:{...headers,'access-control-allow-methods':'POST,GET,OPTIONS','access-control-allow-headers':'content-type,accept,mcp-protocol-version,mcp-session-id'}});\n if(req.method==='GET') {\n if(url.searchParams.get('health')==='1') return json(handlers.canvas_probe({}));\n if(req.headers.get('accept')?.includes('text/event-stream')) return new Response(null,{status:405,headers:{allow:'POST'}});\n return json({service:SERVER,mcp:'POST JSON-RPC to this URL',web:WEB_ENTRY,mode:'connector-assisted, not an unattended renderer'});\n }\n if(req.method!=='POST') return new Response(null,{status:405,headers:{allow:'POST,GET,OPTIONS'}});\n if(!req.headers.get('content-type')?.includes('application/json')) return error(null,-32600,'Content-Type must be application/json',415);\n let m:any;\n try {m=JSON.parse(await body(req));} catch(e) {return error(null,-32700,(e as Error).message==='REQUEST_TOO_LARGE'?'REQUEST_TOO_LARGE':'Parse error',(e as Error).message==='REQUEST_TOO_LARGE'?413:400);}\n if(!m||Array.isArray(m)||m.jsonrpc!=='2.0'||typeof m.method!=='string'||(m.id!==undefined&&m.id!==null&&typeof m.id!=='string'&&typeof m.id!=='number')) return error(null,-32600,'Invalid Request',400);\n if(m.id===undefined) return new Response(null,{status:202,headers});\n const params=m.params||{}; if(typeof params!=='object'||Array.isArray(params)) return error(m.id,-32602,'Invalid params');\n if(m.method==='initialize') return json({jsonrpc:'2.0',id:m.id,result:{protocolVersion:PROTOCOLS.has(params.protocolVersion)?params.protocolVersion:'2025-06-18',capabilities:{tools:{listChanged:true}},serverInfo:SERVER,instructions:'Video workflow orchestration, not an autonomous generator. Read canvas_get_adapter_guides. Use run keys for private state. Claim before provider submission; record genuine job IDs and output URLs. Never confuse queued jobs, technical QA, or proxy clips with visually approved video.'}});\n if(m.method==='ping') return json({jsonrpc:'2.0',id:m.id,result:{}});\n if(m.method==='tools/list') return json({jsonrpc:'2.0',id:m.id,result:{tools}});\n if(m.method!=='tools/call') return error(m.id,-32601,'Method not found');\n const entry=tools.find(t=>t.name===params.name); if(!entry) return error(m.id,-32602,'Unknown tool');\n const args=params.arguments||{}; if(typeof args!=='object'||Array.isArray(args)) return error(m.id,-32602,'Invalid arguments');\n if(entry.inputSchema.required.some(k=>!Object.hasOwn(args,k))||Object.keys(args).some(k=>!Object.hasOwn(entry.inputSchema.properties,k))) return error(m.id,-32602,'Missing or unknown arguments');\n try {\n const data=await handlers[params.name](args);\n return json({jsonrpc:'2.0',id:m.id,result:{content:[{type:'text',text:JSON.stringify(data)}],structuredContent:data,isError:false}});\n } catch(e) {\n const message=(e as Error).message.slice(0,600);\n return json({jsonrpc:'2.0',id:m.id,result:{content:[{type:'text',text:message}],isError:true}});\n }\n}\n","infinite-canvas-mcp-xh32",[162],"pulls","forks","envVars",{"_176":62,"_178":62},"connections",{"_176":62,"_177":62},"count","unfilledCount","emptyCount",{"_21":161,"_13":162,"_131":56,"_133":-5,"_134":33,"_135":163,"_137":58,"_23":193,"_30":165,"_140":166,"_142":167,"_144":46,"_145":168,"_184":194,"_102":195},{"_21":154,"_13":155,"_131":156,"_133":-5,"_134":33,"_135":157,"_137":58,"_23":190,"_30":62,"_140":141,"_142":159,"_144":46,"_145":-5,"_184":191,"_102":192},{"_21":147,"_13":148,"_131":149,"_133":-5,"_134":33,"_135":42,"_137":58,"_23":187,"_30":151,"_140":141,"_142":152,"_144":46,"_145":-5,"_184":188,"_102":189},{"_21":129,"_13":130,"_131":132,"_133":-5,"_134":33,"_135":136,"_137":58,"_23":183,"_30":139,"_140":141,"_142":143,"_144":46,"_145":-5,"_184":185,"_102":186},["D",1789181140473],"children",[],[130],["D",1789181312318],[],[148],["D",1789181024920],[],[155],["D",1789181061626],[],[162],["D",1789176014798],["D",1789532458193],{"_11":12,"_100":96},{"_11":12,"_93":94,"_95":96,"_53":54,"_97":98,"_99":46},{"_30":42,"_13":41,"_21":33,"_28":29,"_32":-5,"_34":-5,"_88":-5,"_89":-5,"_90":-5,"_91":14,"_92":12},"ctx",{"_81":-5,"_122":234,"_235":236,"_237":238,"_239":46,"_240":46,"_241":46,"_242":58,"_243":-5,"_244":-7,"_245":-7,"_246":62,"_247":46,"_248":46,"_249":-5,"_250":46,"_251":46,"_252":46,"_253":-5},"ENV",{"_205":206,"_207":208,"_209":41,"_210":211,"_212":213,"_214":215,"_216":217,"_218":219,"_220":221,"_222":223,"_224":225,"_226":227,"_228":229,"_230":231,"_232":233},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","fdb67791413f4386b4dbf18f354029a617f6160f","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

Branches2

Branches
2