Agentive

\n\n\n\n`;\n}\n// --- HONO APPLICATION SETUP ---\nconst app = new Hono();\n\n// --- API ROUTES ---\n\n// Serves the main HTML file\napp.get(\"/\", (c: Context) => c.html(generateHtml(c.req.url)));\n\n// Serves the static list of industries for the sidebar\napp.get(\"/api/industries\", (c: Context) => c.json(industries));\n\n// New endpoint for the conversational information gathering agent\napp.post(\"/api/gather\", async (c: Context) => {\n const { task, domain, history } = await c.req.json();\n\n const historyString = history.map((h) =>\n `${h.role === \"assistant\" ? \"AI\" : \"User\"}: ${h.content}`\n ).join(\"\\n\");\n\n const prompt = config.prompts.INFORMATION_GATHERER\n .replace(\"{{task}}\", task)\n .replace(\"{{domain}}\", domain)\n .replace(\n \"{{history}}\",\n historyString ||\n \"No history yet. Start with your most important question.\",\n );\n\n try {\n const completion = await services.callOpenAI(\n \"You are a helpful assistant following strict output rules.\",\n prompt,\n { model: config.models.minor },\n );\n const responseText = completion.choices[0].message.content.trim();\n\n if (responseText.startsWith(\"__COMPLETE__\")) {\n const jsonPart = responseText.replace(\"__COMPLETE__\", \"\");\n const data = JSON.parse(jsonPart);\n return c.json({ status: \"complete\", data: data });\n } else {\n return c.json({ status: \"in-progress\", question: responseText });\n }\n } catch (e) {\n console.error(\"Gathering agent failed:\", e);\n return c.json({\n status: \"in-progress\",\n question:\n \"I'm having a little trouble with that. Could you please rephrase your last answer?\",\n }, 500);\n }\n});\n\n// Generates specialized occupations or tasks based on company context\napp.post(\"/api/generate/:type\", async (c: Context) => {\n const { type } = c.req.param();\n if (type !== \"occupations\" && type !== \"tasks\") {\n return c.json({ error: \"Invalid generation type\" }, 400);\n }\n const body = await c.req.json();\n const userContent = { type, ...body };\n return services.callOpenAI(\n config.prompts.DYNAMIC_LIST_GENERATOR,\n userContent,\n { c, isJson: true },\n );\n});\n\n// Creates the master prompt template for a given task\napp.post(\"/api/prompt/dynamic\", async (c: Context) => {\n const { occupation_title, task, company_context } = await c.req.json<\n DynamicPromptBody\n >();\n const userContent = `Company Context: ${\n JSON.stringify(company_context)\n }\\n\\nOccupation: ${occupation_title}\\n\\nTask: ${task}`;\n return services.callOpenAI(config.prompts.PROMPT_REFINER, userContent, {\n c,\n isJson: true,\n });\n});\n\n// Extracts form fields from the master prompt template\napp.post(\"/api/inputs\", async (c: Context) => {\n const { refined_prompt } = await c.req.json();\n return services.callOpenAI(config.prompts.INPUT_EXTRACTOR, refined_prompt, {\n c,\n isJson: true,\n });\n});\n\n// Executes the main agent task, with support for tool calls\napp.post(\"/api/execute-stream\", async (c: Context) => {\n const { refined_prompt, user_inputs, company_context } = await c.req.json<\n ExecuteBody\n >();\n if (!refined_prompt || !user_inputs || !company_context) {\n return c.json({ error: \"Required fields missing\" }, 400);\n }\n\n let finalUserPrompt = refined_prompt;\n for (const [key, value] of Object.entries(user_inputs)) {\n finalUserPrompt = finalUserPrompt.replace(\n new RegExp(`\\\\{\\\\{\\\\s*${key}\\\\s*\\\\}\\\\}`, \"g\"),\n String(value),\n );\n }\n\n const systemPrompt = `# YOUR CONTEXT\n- Company Name: ${company_context.company_name}\n- Company Description: ${company_context.company_description}\n- Primary Goal: ${company_context.use_case}\n- Language for Output: ${company_context.language}\n---\n# YOUR TASK BRIEFING\n${finalUserPrompt}\nYou MUST write your entire response in the requested language: ${company_context.language}.`;\n\n const messages: any[] = [\n { role: \"system\", content: systemPrompt },\n {\n role: \"user\",\n content:\n \"Proceed with the task. If you need to research current information, use the web_search tool.\",\n },\n ];\n\n return streamText(c, async (stream) => {\n try {\n const openai = new OpenAI();\n const initialResponse = await openai.chat.completions.create({\n model: config.models.major,\n messages,\n tools: config.toolDefinitions,\n tool_choice: \"auto\",\n });\n const responseMessage = initialResponse.choices[0].message;\n\n if (responseMessage.tool_calls) {\n messages.push(responseMessage);\n for (const toolCall of responseMessage.tool_calls) {\n const toolResult = await services.executeTool(toolCall);\n messages.push({\n tool_call_id: toolCall.id,\n role: \"tool\",\n name: toolCall.function.name,\n content: toolResult,\n });\n }\n\n const finalStream = await openai.chat.completions.create({\n model: config.models.major,\n messages,\n stream: true,\n });\n\n for await (const part of finalStream) {\n await stream.write(part.choices[0]?.delta?.content || \"\");\n }\n } else {\n if (responseMessage.content) {\n await stream.write(responseMessage.content);\n }\n }\n } catch (e) {\n console.error(\"Execution failed:\", e);\n await stream.write(\"Error: AI execution failed.\");\n }\n });\n});\n\n// --- AGENTIVE LOOP ROUTES (IMPLEMENTED) ---\napp.post(\"/api/generate-criteria\", async (c: Context) => {\n const { refined_prompt, company_context } = await c.req.json<\n GenerateCriteriaBody\n >();\n const userContent = `Company Context:\\n${\n JSON.stringify(company_context)\n }\\n\\nTask Briefing:\\n${refined_prompt}`;\n return services.callOpenAI(\n config.prompts.EVALUATION_CRITERIA_GENERATOR,\n userContent,\n { c, model: config.models.minor, isJson: true },\n );\n});\n\napp.post(\"/api/evaluate-output\", async (c: Context) => {\n const { refined_prompt, criteria, raw_output, language } = await c.req.json<\n EvaluateBody\n >();\n const userContent =\n `Task Briefing:\\n${refined_prompt}\\n\\nSuccess Criteria:\\n${\n JSON.stringify(criteria)\n }\\n\\nOutput to Evaluate:\\n${raw_output}`;\n const systemPrompt =\n `${config.prompts.EVALUATOR_AGENT}\\n\\nOutput Language: ${language}`;\n\n const messages: any[] = [\n { role: \"system\", content: systemPrompt },\n { role: \"user\", content: userContent },\n ];\n\n return streamText(c, async (stream) => {\n try {\n const openai = new OpenAI();\n const initialResponse = await openai.chat.completions.create({\n model: config.models.major, // Use the powerful model for evaluation\n messages,\n tools: config.toolDefinitions,\n tool_choice: \"auto\",\n });\n const responseMessage = initialResponse.choices[0].message;\n\n if (responseMessage.tool_calls) {\n messages.push(responseMessage);\n for (const toolCall of responseMessage.tool_calls) {\n const toolResult = await services.executeTool(toolCall);\n messages.push({\n tool_call_id: toolCall.id,\n role: \"tool\",\n name: toolCall.function.name,\n content: toolResult,\n });\n }\n\n const finalStream = await openai.chat.completions.create({\n model: config.models.major,\n messages,\n stream: true,\n });\n\n for await (const part of finalStream) {\n await stream.write(part.choices[0]?.delta?.content || \"\");\n }\n } else {\n if (responseMessage.content) {\n await stream.write(responseMessage.content);\n }\n }\n } catch (e) {\n console.error(\"Evaluation failed:\", e);\n // Ensure the output is still a parsable JSON object on error\n const errorPayload = {\n scores: [],\n critique:\n `AI evaluation failed: The model could not complete the request. Details: ${e.message}`,\n };\n await stream.write(JSON.stringify(errorPayload));\n }\n });\n});\n\napp.post(\"/api/reevaluate-output\", async (c: Context) => {\n const { refined_prompt, criteria, raw_output, language } = await c.req.json<\n EvaluateBody\n >();\n const userContent =\n `Task Briefing:\\n${refined_prompt}\\n\\nSuccess Criteria:\\n${\n JSON.stringify(criteria)\n }\\n\\nOutput to Evaluate:\\n${raw_output}`;\n const systemPrompt =\n `${config.prompts.EVALUATOR_AGENT}\\n\\nOutput Language: ${language}`;\n\n const messages: any[] = [\n { role: \"system\", content: systemPrompt },\n { role: \"user\", content: userContent },\n ];\n\n return streamText(c, async (stream) => {\n try {\n const openai = new OpenAI();\n const initialResponse = await openai.chat.completions.create({\n model: config.models.major,\n messages,\n tools: config.toolDefinitions,\n tool_choice: \"auto\",\n });\n const responseMessage = initialResponse.choices[0].message;\n\n if (responseMessage.tool_calls) {\n messages.push(responseMessage);\n for (const toolCall of responseMessage.tool_calls) {\n const toolResult = await services.executeTool(toolCall);\n messages.push({\n tool_call_id: toolCall.id,\n role: \"tool\",\n name: toolCall.function.name,\n content: toolResult,\n });\n }\n\n const finalStream = await openai.chat.completions.create({\n model: config.models.major,\n messages,\n stream: true,\n });\n\n for await (const part of finalStream) {\n await stream.write(part.choices[0]?.delta?.content || \"\");\n }\n } else {\n if (responseMessage.content) {\n await stream.write(responseMessage.content);\n }\n }\n } catch (e) {\n console.error(\"Re-evaluation failed:\", e);\n const errorPayload = {\n scores: [],\n critique:\n `AI re-evaluation failed: The model could not complete the request. Details: ${e.message}`,\n };\n await stream.write(JSON.stringify(errorPayload));\n }\n });\n});\n\napp.post(\"/api/refine-output\", async (c: Context) => {\n const { refined_prompt, raw_output, evaluation, language } = await c.req.json<\n RefineBody\n >();\n const userContent =\n `Original Task Briefing:\\n${refined_prompt}\\n\\nMy First Draft:\\n${raw_output}\\n\\nQA Critique:\\n${\n JSON.stringify(evaluation)\n }`;\n const systemPrompt =\n `${config.prompts.REFINER_AGENT}\\n\\nOutput Language: ${language}`;\n return services.callOpenAI(systemPrompt, userContent, { c, stream: true });\n});\n\n// --- EXPORT THE VAL ---\nexport default app.fetch;","deleted","label",[98],"permalink","https://www.val.town/x/join/auto/code/main.tsx","extension","actionData","errors"]

Welcome to Agentive