O*NET Agents with LaTeX Viewer

\n \n\n

1. Choose a Cluster

2. Select an Occupation

3. Pick a Task

\n\n

4. Generated Agent Prompt

\n \n

5. Configure Agent Inputs

\n \n

\n\n

✅ Agent Output (LaTeX Source)

\n\n

Rendered Preview

\n\n \n \n \n `;\n}\n\n// --- BACKEND (API) ---\nconst app = new Hono();\n\n// --- API UTILITIES ---\n// Memoize finding occupations to avoid re-calculating on every API call\nconst getAllOccupations = (() => {\n let allOccupations = null;\n return () => {\n if (allOccupations === null) {\n allOccupations = db.clusters.flatMap(c => c.occupations);\n }\n return allOccupations;\n };\n})();\n\n// --- API ROUTES ---\n\n// Serve Frontend\napp.get(\"/\", (c) => c.html(generateHtml(c.req.url)));\n\n// Get top-level clusters (name and id only)\napp.get(\"/api/clusters\", (c) => {\n return c.json(db.clusters.map(({ id, name }) => ({ id, name })));\n});\n\n// Get occupations for a given cluster\napp.get(\"/api/occupations/:cluster_id\", (c) => {\n const { cluster_id } = c.req.param();\n const cluster = db.clusters.find(c => c.id == parseInt(cluster_id, 10));\n const occupations = cluster ? cluster.occupations.map(({ onet_soc_code, title }) => ({ onet_soc_code, title })) : [];\n return c.json(occupations);\n});\n\n// Get tasks for a given occupation\napp.get(\"/api/tasks/:onet_soc_code\", (c) => {\n const { onet_soc_code } = c.req.param();\n const occupation = getAllOccupations().find(o => o.onet_soc_code === onet_soc_code);\n const tasks = occupation ? occupation.tasks.map(({ id, task }) => ({ task_id: id, task })) : [];\n return c.json(tasks);\n});\n\n// Generate a refined prompt for the AI agent\napp.get(\"/api/prompt/:onet_soc_code/:task_id\", async (c) => {\n const { onet_soc_code, task_id } = c.req.param();\n const occupation = getAllOccupations().find(o => o.onet_soc_code === onet_soc_code);\n const task = occupation?.tasks.find(t => t.id == parseInt(task_id, 10));\n\n if (!task || !occupation) {\n return c.json({ error: \"Task or Occupation not found.\" }, 404);\n }\n\n const userInput = `Occupation: ${occupation.title}, Task: ${task.task}`;\n\n try {\n const openai = new OpenAI();\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [{ role: \"system\", content: PROMPT_REFINER_SYSTEM_PROMPT }, { role: \"user\", content: userInput }],\n response_format: { type: \"json_object\" },\n });\n return c.json(JSON.parse(completion.choices[0].message.content));\n } catch (e) {\n console.error(\"Error in /api/prompt:\", e);\n return c.json({ error: \"AI failed to generate a valid prompt.\" }, 500);\n }\n});\n\n// Extract input fields from the generated prompt\napp.post(\"/api/inputs\", async (c) => {\n const { refined_prompt } = await c.req.json();\n if (!refined_prompt) return c.json({ error: \"refined_prompt is required\" }, 400);\n\n try {\n const openai = new OpenAI();\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [{ role: \"system\", content: INPUT_EXTRACTOR_SYSTEM_PROMPT }, { role: \"user\", content: refined_prompt }],\n response_format: { type: \"json_object\" },\n });\n return c.json(JSON.parse(completion.choices[0].message.content));\n } catch (e) {\n console.error(\"Error in /api/inputs:\", e);\n return c.json({ error: \"AI failed to extract inputs.\" }, 500);\n }\n});\n\n// Execute the final prompt and format the output as LaTeX\napp.post(\"/api/execute\", async (c) => {\n const { refined_prompt, user_inputs } = await c.req.json();\n if (!refined_prompt || !user_inputs) return c.json({ error: \"refined_prompt and user_inputs are required\" }, 400);\n\n let finalUserPrompt = refined_prompt;\n for (const [key, value] of Object.entries(user_inputs)) {\n const placeholder = new RegExp(`\\\\{\\\\{\\\\s*${key}\\\\s*\\\\}\\\\}`, \"g\");\n finalUserPrompt = finalUserPrompt.replace(placeholder, String(value));\n }\n\n try {\n const openai = new OpenAI();\n\n const agentCompletion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [\n { role: \"system\", content: finalUserPrompt },\n { role: \"user\", content: \"Proceed with the task as instructed.\" },\n ],\n });\n const agentOutput = agentCompletion.choices[0].message.content;\n\n if (!agentOutput) {\n throw new Error(\"The agent returned no content.\");\n }\n\n const latexCompletion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [\n { role: \"system\", content: LATEX_FORMATTER_SYSTEM_PROMPT },\n { role: \"user\", content: agentOutput },\n ],\n });\n const latexOutput = latexCompletion.choices[0].message.content;\n\n return c.json({ output: latexOutput });\n } catch (e) {\n console.error(\"Error during execution:\", e);\n return c.json({ error: `AI execution failed: ${e.message}` }, 500);\n }\n});\n\nexport default app.fetch;","deleted","label",[98],"permalink","https://www.val.town/x/real/onet2/code/main.tsx","extension","actionData","errors"]