TownCar

Where can we take you? Describe your destination.

\n\n

\n\n \n \n \n `;\n}\n\n// --- BACKEND: The Hono API Server ---\nconst app = new Hono();\n\nasync function callOpenAI(c, messages, response_format = { type: \"text\" }) {\n const openai = new OpenAI();\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: messages,\n response_format: response_format,\n });\n\n if (!completion || !completion.choices || completion.choices.length === 0) {\n console.error(\"Invalid response from OpenAI API:\", JSON.stringify(completion, null, 2));\n throw new Error(\"The AI service returned an invalid or empty response.\");\n }\n\n return completion.choices[0].message.content;\n}\n\napp.post(\"/api/plan-workflow\", async (c) => {\n const { goal } = await c.req.json();\n try {\n const messages = [{ role: \"system\", content: META_PROMPTS.ROUTE_PLANNER.replace(\"{{goal}}\", goal) }];\n const planJson = await callOpenAI(c, messages, { type: \"json_object\" });\n return c.json(JSON.parse(planJson));\n } catch (e) {\n console.warn(\"Initial route planning failed. Retrying...\", e);\n try {\n const retryMessages = [{\n role: \"system\",\n content: META_PROMPTS.ROUTE_PLANNER_RETRY.replace(\"{{goal}}\", goal),\n }];\n const retryPlanJson = await callOpenAI(c, retryMessages, { type: \"json_object\" });\n return c.json(JSON.parse(retryPlanJson));\n } catch (retryError) {\n console.error(\"Route planning retry failed:\", retryError);\n return c.json(\n { details: \"The AI failed to create a valid workflow plan. Please try rephrasing your goal.\" },\n 500,\n );\n }\n }\n});\n\napp.post(\"/api/generate-step-ui\", async (c) => {\n try {\n const { task, context } = await c.req.json();\n const contextString = JSON.stringify(context, null, 2);\n\n const architectMessages = [{\n role: \"system\",\n content: META_PROMPTS.STOP_ARCHITECT.replace(\"{{task}}\", task).replace(\"{{context}}\", contextString),\n }];\n const specJson = await callOpenAI(c, architectMessages, { type: \"json_object\" });\n const spec = JSON.parse(specJson);\n\n let uiJson = \"{\\\"inputs\\\":[]}\";\n if (spec.inputs && spec.inputs.length > 0) {\n const uiBuilderMessages = [{ role: \"system\", content: META_PROMPTS.UI_BUILDER.replace(\"{{spec}}\", specJson) }];\n uiJson = await callOpenAI(c, uiBuilderMessages, { type: \"json_object\" });\n }\n\n return c.json({ taskSpecification: spec, uiSchema: JSON.parse(uiJson) });\n } catch (e) {\n console.error(\"Step UI generation failed:\", e);\n return c.json({ details: `The AI failed to build the UI for the current step. Error: ${e.message}` }, 500);\n }\n});\n\napp.post(\"/api/execute-step\", async (c) => {\n const { readable, writable } = new TransformStream();\n const writer = writable.getWriter();\n const textEncoder = new TextEncoder();\n\n new Promise(async (resolve, reject) => {\n try {\n const { task, context, inputs } = await c.req.json();\n const contextString = JSON.stringify(context, null, 2);\n\n const specJson = await callOpenAI(c, [{\n role: \"system\",\n content: META_PROMPTS.STOP_ARCHITECT.replace(\"{{task}}\", task).replace(\"{{context}}\", contextString),\n }], { type: \"json_object\" });\n let executionPrompt = await callOpenAI(c, [{\n role: \"system\",\n content: META_PROMPTS.LOGIC_WEAVER.replace(\"{{spec}}\", specJson).replace(\"{{context}}\", contextString),\n }]);\n\n for (const [key, value] of Object.entries(inputs)) {\n executionPrompt = executionPrompt.replace(new RegExp(`\\\\{\\\\{\\\\s*${key}\\\\s*\\\\}\\\\}`, \"g\"), value);\n }\n\n const openai = new OpenAI();\n const aiStream = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [{ role: \"user\", content: executionPrompt }],\n stream: true,\n });\n\n for await (const chunk of aiStream) {\n const content = chunk.choices[0]?.delta?.content || \"\";\n if (content) {\n await writer.write(textEncoder.encode(content));\n }\n }\n\n await writer.close();\n resolve();\n } catch (e) {\n console.error(\"Step execution stream failed:\", e);\n const errorPayload = { error: true, details: `The AI failed to execute this step. Details: ${e.message}` };\n try {\n await writer.write(textEncoder.encode(JSON.stringify(errorPayload)));\n } catch (writeError) {\n console.error(\"Failed to write error to stream:\", writeError);\n }\n await writer.close();\n reject(e);\n }\n });\n\n return new Response(readable, {\n headers: { \"Content-Type\": \"text/plain; charset=utf-8\" },\n });\n});\n\napp.get(\"/\", (c) => c.html(generateHtml(c.req.url)));\n\nexport default app.fetch;","label","towncar","children",[71],"contentHash","18ef47a7cb8dee53c1b864c6da6435f8d842b7c8cc769e781c95b8d34256ae6f","b3db9246-678e-11f0-91a4-0224a6c84d84","8a6ecc4a-677d-11f0-872a-0224a6c84d84",["D",1753252304890],24,"// @ts-ignore\nimport { Hono } from \"npm:hono@4.4.12\";\n// @ts-ignore\nimport { OpenAI } from \"https://esm.town/v/std/openai?v=4\";\n\n// --- Emergent: The JIT Workflow Engine v3.0 (Autonomous Task Architect) ---\n// --- Banks 7/22/2025. For Leo & Niko. Architect now correctly identifies autonomous tasks.\n// --- josh@dereticular.com\n// --- jabanks@gmail.com\n//\n\n// --- META-PROMPTS: The Engine's Core Logic (REVISED TASK_ARCHITECT) ---\nconst META_PROMPTS = {\n WORKFLOW_PLANNER:\n `You are an expert AI system architect designing a workflow for an autonomous Large Language Model (LLM) agent. Your task is to deconstruct a user's goal into a logical sequence of discrete, objectively performable tasks for the LLM.\n\n **CRITICAL INSTRUCTIONS:**\n 1. **AI-Centric Tasks:** Each task must be something an LLM can execute independently within a software environment. Focus on analysis, structuring data, planning content, generating drafts, and refining text.\n 2. **NO EXTERNAL ACTIONS:** Do NOT generate steps that involve human intervention (e.g., \"review the draft,\" \"send a test email\") or physical/external actions (e.g., \"publish the article,\" \"send the final email\"). The final output is the deliverable, not the action of delivering it.\n 3. **Consolidate Steps:** Combine trivial micro-tasks. Instead of separate steps for \"Determine audience,\" \"Choose tone,\" and \"Draft content,\" create a single, more comprehensive task like \"Draft the content, adopting a [professional/casual] tone for an audience of [experts/beginners].\"\n\n Return your plan as a JSON object with a single key, \"steps\", which is an array of objects. Each object must have an \"id\" and a \"task\" description. Your entire output must be the JSON object itself.\n\n User Goal: \"{{goal}}\"`,\n WORKFLOW_PLANNER_RETRY:\n `The first plan was flawed. Simplify the workflow for an autonomous LLM agent. Break the original goal into just 2 or 3 high-level, essential, AI-performable tasks. Focus only on content generation and refinement. Do not include steps requiring human actions or external system interactions. Return the plan in the same JSON format as before: a JSON object with a \"steps\" key. Your entire output must be only the JSON object. Original User Goal: \"{{goal}}\"`,\n TASK_ARCHITECT:\n `You are an expert system architect. Your role is to analyze a task for an autonomous LLM agent and define its operational parameters in a structured JSON format.\n\n **Task Analysis:**\n Analyze the 'Current Task' below. Your primary goal is to determine if the task can be completed autonomously by the LLM using its own knowledge and the provided 'Context', or if it absolutely requires new information from the human user.\n\n **CRITICAL RULES FOR DEFINING INPUTS:**\n 1. **DEFAULT TO AUTONOMY:** Assume the task is autonomous by default. If the task is an instruction for the LLM to research, write, create, analyze, or generate something (e.g., \"Research market trends,\" \"Draft a blog post about dogs,\" \"Analyze the provided text\"), the LLM must perform the action without user input. In this case, the \"inputs\" array in your JSON output MUST be empty (\\`[]\\`).\n 2. **REQUIRE INPUTS ONLY WHEN ESSENTIAL:** Only define \"inputs\" if the task is impossible to complete without a specific piece of information that the AI cannot know and that is not in the 'Context' (e.g., \"Incorporate the user's specific company name,\" \"Summarize the attached, unseen document,\" \"Translate the following user-provided text\").\n\n **JSON Output:**\n Convert the task into a JSON object with three keys:\n - \"objective\": A brief, one-sentence summary of the task's goal.\n - \"inputs\": An array of input objects. This array MUST be empty if the task is autonomous.\n - \"output\": A description of the final deliverable for this specific step.\n\n Context: {{context}}\n Current Task: \"{{task}}\"`,\n UI_BUILDER:\n `You are a UI/UX specialist. Convert the provided task specification JSON into a JSON form schema. The output must be a JSON object with a single key, \"inputs\", containing an array of objects. Each input object must have a \"name\", \"label\", \"type\", \"required\", and a \"placeholder\". Task Specification: {{spec}}`,\n LOGIC_WEAVER:\n `You are a senior prompt engineer. Write a new, expert-level system prompt that can execute the provided task specification. The prompt must be a complete set of instructions for another AI. CRITICAL: The final output of the AI executing this prompt MUST be well-structured, clear Markdown. Incorporate the provided \"context\" from previous steps directly into the prompt's instructions. Use double-curly-braced variables for any \"inputs\" that the user still needs to provide. Context: {{context}}. Task Specification: {{spec}}`,\n};\n\n// --- FRONTEND: The Emergent UI (Unchanged) ---\nfunction generateHtml(sourceUrl) {\n return `\n \n \n \n \n \n Emergent\n \n \n \n \n \n \n \n \n \n \n

Emergent

Describe the workflow you want to build.

\n \n

Tip: Be descriptive for the best results! Aim for at least 5 words.

\n \n \n \n `;\n}\n\n// --- BACKEND: The Hono API Server (Unchanged) ---\nconst app = new Hono();\n\nasync function callOpenAI(c, messages, response_format = { type: \"text\" }) {\n const openai = new OpenAI();\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: messages,\n response_format: response_format,\n });\n\n if (!completion || !completion.choices || completion.choices.length === 0) {\n console.error(\"Invalid response from OpenAI API:\", JSON.stringify(completion, null, 2));\n throw new Error(\"The AI service returned an invalid or empty response.\");\n }\n\n return completion.choices[0].message.content;\n}\n\napp.post(\"/api/plan-workflow\", async (c) => {\n const { goal } = await c.req.json();\n try {\n const messages = [{ role: \"system\", content: META_PROMPTS.WORKFLOW_PLANNER.replace(\"{{goal}}\", goal) }];\n const planJson = await callOpenAI(c, messages, { type: \"json_object\" });\n return c.json(JSON.parse(planJson));\n } catch (e) {\n console.warn(\"Initial workflow planning failed. Retrying...\", e);\n try {\n const retryMessages = [{\n role: \"system\",\n content: META_PROMPTS.WORKFLOW_PLANNER_RETRY.replace(\"{{goal}}\", goal),\n }];\n const retryPlanJson = await callOpenAI(c, retryMessages, { type: \"json_object\" });\n return c.json(JSON.parse(retryPlanJson));\n } catch (retryError) {\n console.error(\"Workflow planning retry failed:\", retryError);\n return c.json(\n { details: \"The AI failed to create a valid workflow plan. Please try rephrasing your goal.\" },\n 500,\n );\n }\n }\n});\n\napp.post(\"/api/generate-step-ui\", async (c) => {\n try {\n const { task, context } = await c.req.json();\n const contextString = JSON.stringify(context, null, 2);\n\n // The architect now has smarter logic to decide if inputs are needed\n const architectMessages = [{\n role: \"system\",\n content: META_PROMPTS.TASK_ARCHITECT.replace(\"{{task}}\", task).replace(\"{{context}}\", contextString),\n }];\n const specJson = await callOpenAI(c, architectMessages, { type: \"json_object\" });\n const spec = JSON.parse(specJson);\n\n // The UI builder only runs if the architect actually defines inputs\n let uiJson = \"{\\\"inputs\\\":[]}\";\n if (spec.inputs && spec.inputs.length > 0) {\n const uiBuilderMessages = [{ role: \"system\", content: META_PROMPTS.UI_BUILDER.replace(\"{{spec}}\", specJson) }];\n uiJson = await callOpenAI(c, uiBuilderMessages, { type: \"json_object\" });\n }\n\n return c.json({ taskSpecification: spec, uiSchema: JSON.parse(uiJson) });\n } catch (e) {\n console.error(\"Step UI generation failed:\", e);\n return c.json({ details: `The AI failed to build the UI for the current step. Error: ${e.message}` }, 500);\n }\n});\n\napp.post(\"/api/execute-step\", async (c) => {\n const { readable, writable } = new TransformStream();\n const writer = writable.getWriter();\n const textEncoder = new TextEncoder();\n\n new Promise(async (resolve, reject) => {\n try {\n const { task, context, inputs } = await c.req.json();\n const contextString = JSON.stringify(context, null, 2);\n\n const specJson = await callOpenAI(c, [{\n role: \"system\",\n content: META_PROMPTS.TASK_ARCHITECT.replace(\"{{task}}\", task).replace(\"{{context}}\", contextString),\n }], { type: \"json_object\" });\n let executionPrompt = await callOpenAI(c, [{\n role: \"system\",\n content: META_PROMPTS.LOGIC_WEAVER.replace(\"{{spec}}\", specJson).replace(\"{{context}}\", contextString),\n }]);\n\n for (const [key, value] of Object.entries(inputs)) {\n executionPrompt = executionPrompt.replace(new RegExp(`\\\\{\\\\{\\\\s*${key}\\\\s*\\\\}\\\\}`, \"g\"), value);\n }\n\n const openai = new OpenAI();\n const aiStream = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [{ role: \"user\", content: executionPrompt }],\n stream: true,\n });\n\n for await (const chunk of aiStream) {\n const content = chunk.choices[0]?.delta?.content || \"\";\n if (content) {\n await writer.write(textEncoder.encode(content));\n }\n }\n\n await writer.close();\n resolve();\n } catch (e) {\n console.error(\"Step execution stream failed:\", e);\n const errorPayload = { error: true, details: `The AI failed to execute this step. Details: ${e.message}` };\n try {\n await writer.write(textEncoder.encode(JSON.stringify(errorPayload)));\n } catch (writeError) {\n console.error(\"Failed to write error to stream:\", writeError);\n }\n await writer.close();\n reject(e);\n }\n });\n\n return new Response(readable, {\n headers: { \"Content-Type\": \"text/plain; charset=utf-8\" },\n });\n});\n\napp.get(\"/\", (c) => c.html(generateHtml(c.req.url)));\n\nexport default app.fetch;",[71],"6df8326ec27390b4e9b12a3c1d9f3e7fddcfd32a6777dd44bbeeca4cc29fa601",["D",1753266566240],"legacyEnvVars","httpPrivacy","description","Take your dreams for a ride around town with TownCar.","imageUrl","userId","63f04ac0-562b-11f0-a849-2f0fd58d5bdc","pinnedFileId","pinPreview","pinMigrated","screenshotVersion","likeCount","0abcaa5c-2531-4e34-967b-6f68bd9904a8","tursoName","tursoHostname","tursoMeasuredAt","pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_11":12,"_139":135},"org",{"_11":12,"_133":-5,"_134":135,"_107":108,"_136":137,"_138":48},"branch","displayName","avatar","https://img.clerk.com/eyJ0eXBlIjoiZGVmYXVsdCIsImlpZCI6Imluc18yTm10cGhQUklwU0tqZHNiOHNnNGd2ZjJURFUiLCJyaWQiOiJ1c2VyXzJ6Rng5NEh2b3dPMmVwbThwVm1keXA3aTk5bCJ9","tier","free","canRestrictVals","profileImageUrl","","adoptOrgId",{"_32":59,"_13":14,"_34":185,"_46":-5,"_102":48,"_49":50,"_103":50,"_104":105,"_106":-5,"_107":108,"_109":78,"_110":26,"_111":26,"_112":-5,"_113":22,"_53":114,"_115":-5,"_116":-5,"_117":-5,"_118":-5,"_119":120,"_121":120,"_122":123,"_124":123,"_125":123,"_126":120,"_127":120,"_128":186,"_130":187,"_132":188,"_47":48},"tree",[182],"files",[182],"counts",{"_171":22,"_172":173,"_174":173,"_175":176,"_177":178},"contents",[168],"file","upstreamProject",{"_32":39,"_13":56,"_104":-5,"_106":-5,"_49":50,"_128":166},"upstreamBranch",{"_32":33,"_34":164,"_36":165,"_13":16,"_38":39,"_40":41,"_42":-5,"_43":-5,"_44":45,"_46":-5,"_47":48,"_49":50,"_51":52,"_53":54,"_55":56},"emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},["D",1753244933938],["D",1753252304890],{"_11":52,"_139":167},"https://img.clerk.com/eyJ0eXBlIjoiZGVmYXVsdCIsImlpZCI6Imluc18yTm10cGhQUklwU0tqZHNiOHNnNGd2ZjJURFUiLCJyaWQiOiJ1c2VyXzJ4RElTYm91QXRtUWhBeEdMM1RYYUVmMVVwYSJ9",{"_32":76,"_13":71,"_77":78,"_79":-5,"_80":58,"_81":57,"_82":26,"_34":169,"_40":20,"_84":85,"_86":87,"_65":48,"_88":89,"_70":170},["D",1753268002976],[71],"pulls","branches",1,"forks","envVars",{"_179":22,"_181":22},"connections",{"_179":22,"_180":22},"count","unfilledCount","emptyCount",{"_32":76,"_13":71,"_77":78,"_79":-5,"_80":58,"_81":57,"_82":26,"_34":83,"_40":20,"_84":85,"_86":87,"_65":48,"_88":89,"_90":183,"_70":184},[],[71],["D",1753266566240],{"_11":12,"_139":135},{"_11":12,"_133":-5,"_134":135,"_107":108,"_136":137,"_138":48},{"_40":57,"_13":16,"_32":58,"_38":59,"_42":33,"_43":41,"_60":50,"_61":39,"_62":54,"_55":14,"_51":12},"ctx",{"_128":-5,"_161":222,"_223":224,"_225":226,"_227":48,"_228":48,"_229":48,"_230":26,"_231":-5,"_232":-7,"_233":-7,"_234":22,"_235":48,"_236":48,"_237":-5,"_238":48,"_239":48,"_240":48,"_241":-5},"ENV",{"_193":194,"_195":196,"_197":16,"_198":199,"_200":201,"_202":203,"_204":205,"_206":207,"_208":209,"_210":211,"_212":213,"_214":215,"_216":217,"_218":219,"_220":221},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","71c1b20804b3232a7e513be260cfffc1932a9e97","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"]

Compare

Mergereal/mainintojoin/main

Hmain.tsx+409-255