\n 🧠 Multi-Agent Problem Solver\n

\n {problem && (\n

\n Problem: {problem.query.substring(0, 40)}...\n

\n )}\n

\n\n

\n {messages.map(msg => (\n

\n {msg.role.toUpperCase()}\n {msg.tools && msg.tools.length > 0 && (\n

\n Tools used: {msg.tools.map(t => getToolNameById(t)).join(', ')}\n

\n )}\n

{msg.content}

\n ))}\n

\n {isLoading && (\n

\n Agents collaborating...\n

\n )}\n

\n\n

\n );\n}\n\nfunction client() {\n createRoot(document.getElementById(\"root\")).render();\n}\nif (typeof document !== \"undefined\") { client(); }\n\nexport default async function server(request: Request): Promise {\n const { OpenAI } = await import(\"https://esm.town/v/std/openai\");\n const openai = new OpenAI();\n\n // Mock tools for demonstration purposes\n const tools: Tool[] = [\n {\n id: \"web-search\",\n name: \"Web Search\",\n description: \"Search the web for current information\",\n use: async (query) => `Found relevant information about: ${query}`\n },\n {\n id: \"data-analysis\",\n name: \"Data Analyzer\",\n description: \"Analyze datasets and extract insights\",\n use: async (data) => `Analysis complete. Key insights: ${data.split(' ').length} data points processed.`\n },\n {\n id: \"code-generator\",\n name: \"Code Generator\",\n description: \"Generate code snippets in various languages\",\n use: async (spec) => `Generated code solution based on: ${spec}`\n },\n {\n id: \"math-solver\",\n name: \"Mathematical Solver\",\n description: \"Solve complex mathematical problems\",\n use: async (problem) => `Mathematical solution: Solved ${problem}`\n }\n ];\n\n // Endpoint to get available tools\n if (request.method === 'GET' && new URL(request.url).pathname === '/available-tools') {\n return new Response(JSON.stringify({ tools }), {\n headers: { 'Content-Type': 'application/json' }\n });\n }\n\n // Endpoint to initiate problem solving - creates workflow\n if (request.method === 'POST' && new URL(request.url).pathname === '/solve-problem') {\n try {\n const body = await request.json();\n const { query } = body;\n\n // Use the decomposer agent to break down the problem\n const decomposerCompletion = await openai.chat.completions.create({\n messages: [\n { \n role: \"system\", \n content: \"You are a problem decomposition specialist. Your task is to break complex problems into logical, manageable steps that can be assigned to specialized agents. Focus on creating a clear workflow with proper dependencies between steps.\" \n },\n { \n role: \"user\", \n content: `Decompose this problem into steps: \"${query}\". Identify which specialized agent should handle each step: researcher (information gathering), analyst (data processing), strategist (planning), implementer (executing solutions), critic (evaluating solutions), or synthesizer (combining results). Define dependencies between steps.` \n }\n ],\n model: \"gpt-4o-mini\",\n max_tokens: 500\n });\n\n const decomposition = decomposerCompletion.choices[0].message.content || \"\";\n\n // Now create a structured workflow\n const workflowCompletion = await openai.chat.completions.create({\n messages: [\n { \n role: \"system\", \n content: `You create structured workflows for complex problems. Output only valid JSON with this structure:\n[\n {\n \"id\": \"step1\",\n \"agentRole\": \"researcher|analyst|strategist|implementer|critic|synthesizer\",\n \"instruction\": \"Detailed instructions for this step\",\n \"dependsOn\": [] // array of step IDs this step depends on\n },\n // more steps...\n]` \n },\n { \n role: \"user\", \n content: `Based on this problem decomposition: \"${decomposition}\", create a structured workflow for solving: \"${query}\". Ensure proper dependencies between steps and that the final step is always a synthesizer that combines all results.` \n }\n ],\n model: \"gpt-4o-mini\",\n max_tokens: 800\n });\n\n // Extract and parse the JSON workflow\n const workflowContent = workflowCompletion.choices[0].message.content || \"[]\";\n const workflowMatch = workflowContent.match(/\\[\\s*\\{[\\s\\S]*\\}\\s*\\]/);\n let workflowSteps = [];\n \n if (workflowMatch) {\n try {\n workflowSteps = JSON.parse(workflowMatch[0]);\n // Add completed flag to each step\n workflowSteps = workflowSteps.map((step: any) => ({ ...step, completed: false }));\n } catch (e) {\n console.error(\"Error parsing workflow JSON:\", e);\n workflowSteps = [];\n }\n }\n\n // Create problem state\n const problem: ProblemState = {\n id: uuidv4(),\n query,\n workflow: workflowSteps,\n currentStepIndex: 0\n };\n\n return new Response(JSON.stringify({ \n problem,\n initialDecomposition: decomposition\n }), {\n headers: { 'Content-Type': 'application/json' }\n });\n } catch (error) {\n return new Response(JSON.stringify({ \n error: \"Failed to decompose problem\",\n details: error instanceof Error ? error.message : \"Unknown error\"\n }), { \n status: 500,\n headers: { 'Content-Type': 'application/json' }\n });\n }\n }\n\n // Endpoint to execute a single workflow step\n if (request.method === 'POST' && new URL(request.url).pathname === '/execute-step') {\n try {\n const body = await request.json();\n const { stepId, agentRole, instruction, previousResults, query } = body;\n \n // Select which tools this agent can use based on role\n const agentTools: Record = {\n decomposer: [],\n researcher: [\"web-search\"],\n analyst: [\"data-analysis\", \"math-solver\"],\n strategist: [],\n implementer: [\"code-generator\"],\n critic: [],\n synthesizer: []\n };\n \n const applicableTools = agentTools[agentRole as AgentRole] || [];\n const toolsUsed: string[] = [];\n \n // Create agent system prompt based on role\n const rolePrompts: Record = {\n researcher: \"You are a thorough research agent focused on gathering comprehensive information related to the problem.\",\n analyst: \"You are a data analysis agent specialized in processing information and extracting meaningful insights.\",\n strategist: \"You are a strategic planning agent who develops comprehensive approaches to addressing problems.\",\n implementer: \"You are an implementation agent who creates concrete solutions based on strategies.\",\n critic: \"You are a critical evaluation agent who identifies potential issues or weaknesses in proposed solutions.\",\n synthesizer: \"You are a synthesis agent who combines all previous outputs into a cohesive final solution.\"\n };\n \n // Simulate tool usage if applicable\n let toolOutput = \"\";\n if (applicableTools.length > 0) {\n // Randomly decide to use 1-2 tools for demonstration\n const numToolsToUse = Math.min(Math.floor(Math.random() * 2) + 1, applicableTools.length);\n const selectedTools = applicableTools.slice(0, numToolsToUse);\n \n for (const toolId of selectedTools) {\n const tool = tools.find(t => t.id === toolId);\n if (tool) {\n toolsUsed.push(toolId);\n const result = await tool.use(instruction.substring(0, 50));\n toolOutput += `[Tool: ${tool.name}] ${result}\\n`;\n }\n }\n }\n \n // Generate agent response\n const systemPrompt = rolePrompts[agentRole as AgentRole] || \n \"You are a problem-solving agent working as part of a multi-agent system.\";\n \n const agentCompletion = await openai.chat.completions.create({\n messages: [\n { \n role: \"system\", \n content: systemPrompt\n },\n { \n role: \"user\", \n content: `\nMain problem: \"${query}\"\n\nYour task: ${instruction}\n\n${previousResults ? `Previous results from other agents:\\n${previousResults}\\n` : ''}\n${toolOutput ? `Tool results:\\n${toolOutput}\\n` : ''}\n\nProvide your specialized contribution as ${agentRole}. Be thorough but concise.`\n }\n ],\n model: \"gpt-4o-mini\",\n max_tokens: 800\n });\n\n const result = agentCompletion.choices[0].message.content || \"\";\n\n return new Response(JSON.stringify({ \n stepId,\n result,\n toolsUsed\n }), {\n headers: { 'Content-Type': 'application/json' }\n });\n } catch (error) {\n return new Response(JSON.stringify({ \n error: \"Failed to execute workflow step\",\n details: error instanceof Error ? error.message : \"Unknown error\"\n }), { \n status: 500,\n headers: { 'Content-Type': 'application/json' }\n });\n }\n }\n\n // Default HTML response\n return new Response(`\n \n \n Multi-Agent Problem Solver\n \n \n \n \n

\n \n \n \n \n `, {\n headers: { 'Content-Type': 'text/html' }\n });\n}","deleted","label",[98],"permalink","https://www.val.town/x/web/hats/code/main.tsx","extension","actionData","errors"]