🔗 Policy Analysis Tool

\n JavaScript is required to run this application. Please enable JavaScript.\n

Loading Application...

\n\n \n \n \n `;\n}\n\n// --- Main Server Request Handler (Val Town Entry Point) ---\nexport default async function(req: Request): Promise {\n // --- Dynamic Import of OpenAI Library ---\n const { OpenAI } = await import(\"https://esm.town/v/std/openai\");\n\n // Server-side logging utility\n function logServer(level: \"INFO\" | \"WARN\" | \"ERROR\", message: string, context?: any) {\n const timestamp = new Date().toISOString();\n console.log(JSON.stringify({ timestamp, level, message, context, environment: \"server\" }));\n }\n\n // --- OpenAI Initialization ---\n let openai: OpenAI;\n try {\n // Assumes OPENAI_API_KEY is set as a Val Town secret named 'openai'\n openai = new OpenAI();\n // Perform a simple test call or validation if needed, but often just instantiating is enough.\n } catch (e: any) {\n logServer(\"ERROR\", \"Failed to initialize OpenAI Client\", { error: e.message });\n // If OpenAI fails to init, we can still potentially serve the HTML,\n // but API calls will fail later. For API calls, we return an immediate error.\n // We'll handle API call errors within the analysis function.\n }\n\n // --- Core Analysis Tool: Extract Policy Entities and Relationships ---\n const analyzePolicyDocument = async (\n policyText: string,\n ): Promise<{ success: boolean; data?: AnalysisResult; error?: string }> => {\n logServer(\"INFO\", \"Executing Tool: Analyze Policy Document\", { textLength: policyText.length });\n\n if (!policyText || policyText.trim().length < 30) { // Slightly shorter min length\n return { success: false, error: \"Input text is too short or empty to analyze meaningfully.\" };\n }\n if (policyText.length > 20000) { // Set a reasonable limit\n logServer(\"WARN\", \"Input text truncated for analysis\", { originalLength: policyText.length });\n policyText = policyText.substring(0, 20000);\n }\n\n // Check if OpenAI client initialized successfully *before* making the call\n if (!openai) {\n return {\n success: false,\n error: \"OpenAI client failed to initialize. Check server logs and API key configuration.\",\n };\n }\n\n // ** CRITICAL: This prompt needs careful design and testing! **\n // Adjusted prompt for React Flow format (explicit position: {x:0, y:0})\n const analyzePolicyPrompt = `\n You are an expert legal and policy analyst AI. Your task is to read the provided policy text and identify key entities (policies, laws, regulations, specific sections, acts, standards) and the relationships between them (e.g., cites, references, amends, is amended by, supersedes, defines term from, related to).\n\n Input Text:\n \"\"\"\n \\${policyText}\n \"\"\"\n\n Instructions:\n 1. **Identify Entities:** Find mentions of specific policies, laws, regulations, acts, standards, or distinct sections (e.g., \"Public Law 117-103\", \"Section 508\", \"GDPR Article 30\", \"ADA\", \"ISO 27001\", \"OMB A-130\"). Treat the main input document itself as an entity if possible.\n 2. **Identify Relationships:** Determine how entities relate based *only* on the text. Look for verbs/phrases like \"cites\", \"references\", \"pursuant to\", \"as defined in\", \"amends\", \"is amended by\", \"supersedes\", \"incorporates\", \"harmonized with\", \"see also\".\n 3. **Create Nodes:** For each unique entity, create a node object for React Flow.\n * 'id': Use a concise, unique identifier string (e.g., \"PL-117-103\", \"Sec508\", \"GDPR-Art30\"). NO SPACES or special chars. Use hyphens or camelCase.\n * 'data': An object containing 'label' (the full name/title for display) and optionally 'type' (e.g., \"Law\", \"Regulation\", \"Section\").\n * 'position': An object \\`{ \"x\": 0, \"y\": 0 }\\` (Client will handle actual layout).\n 4. **Create Edges:** For each relationship (A -> B), create an edge object for React Flow.\n * 'id': A unique edge identifier string (e.g., \"e1\", \"edge-Sec508-cites-PL117\"). Generate simple IDs like \"e1\", \"e2\"...\n * 'source': The 'id' of the source node (A).\n * 'target': The 'id' of the target node (B).\n * 'label': (Optional) The relationship verb/phrase (e.g., \"cites\", \"amends\"). Keep labels concise.\n 5. **Consolidate:** Ensure node IDs are consistent. Use judgment for granularity.\n 6. **Summary:** Briefly summarize the main topic and the most significant relationships found (1-2 sentences).\n\n **Output Format:** Respond ONLY with a valid JSON object containing 'nodes', 'edges', and 'analysisSummary' keys, strictly adhering to the React Flow structure described above. Example:\n {\n \"nodes\": [\n { \"id\": \"Node1ID\", \"data\": { \"label\": \"Full Name 1\", \"type\": \"Type1\" }, \"position\": { \"x\": 0, \"y\": 0 } },\n { \"id\": \"Node2ID\", \"data\": { \"label\": \"Full Name 2\", \"type\": \"Type2\" }, \"position\": { \"x\": 0, \"y\": 0 } }\n ],\n \"edges\": [\n { \"id\": \"e1\", \"source\": \"Node1ID\", \"target\": \"Node2ID\", \"label\": \"Relationship\" }\n ],\n \"analysisSummary\": \"Brief summary text here...\"\n }\n Ensure node 'id' fields are unique strings. Ensure 'source' and 'target' in edges correspond to valid node 'id's.\n `;\n\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o\", // Recommend powerful model for this task\n response_format: { type: \"json_object\" },\n messages: [\n {\n role: \"system\",\n content: \"You are an expert legal/policy analyst AI outputting structured JSON for a network graph.\",\n },\n { role: \"user\", content: analyzePolicyPrompt },\n ],\n temperature: 0.15, // Low temperature for deterministic extraction\n max_tokens: 2500, // Adjust based on expected output complexity\n });\n\n const content = completion.choices[0]?.message?.content;\n if (!content) {\n throw new Error(\"LLM returned empty content for policy analysis.\");\n }\n\n logServer(\"INFO\", \"Received analysis JSON from LLM\");\n let parsedData: AnalysisResult;\n try {\n parsedData = JSON.parse(content);\n } catch (parseError: any) {\n logServer(\"ERROR\", \"Failed to parse JSON response from analysis LLM\", {\n responsePreview: content.substring(0, 200),\n error: parseError.message,\n });\n throw new Error(\n `AI response was not valid JSON: \\${parseError.message}. Response start: \\${content.substring(0,100)}`,\n );\n }\n\n // --- Server-Side Validation ---\n if (!Array.isArray(parsedData.nodes) || !Array.isArray(parsedData.edges)) {\n logServer(\"WARN\", \"Analysis LLM JSON response missing expected arrays (nodes/edges)\", { parsedData });\n throw new Error(\"AI response missing 'nodes' or 'edges' array.\");\n }\n // Basic validation of node/edge structure\n parsedData.nodes.forEach((n, i) => {\n if (\n !n.id || typeof n.id !== \"string\" || !n.data || typeof n.data.label !== \"string\"\n || typeof n.position !== \"object\"\n ) {\n throw new Error(`Invalid node structure at index \\${i}: \\${JSON.stringify(n)}`);\n }\n });\n parsedData.edges.forEach((e, i) => {\n if (\n !e.id || typeof e.id !== \"string\" || !e.source || typeof e.source !== \"string\" || !e.target\n || typeof e.target !== \"string\"\n ) {\n throw new Error(`Invalid edge structure at index \\${i}: \\${JSON.stringify(e)}`);\n }\n // Optionally ensure source/target IDs exist in nodes list\n });\n\n return { success: true, data: parsedData };\n } catch (error: any) {\n logServer(\"ERROR\", \"Policy Analysis tool execution failed\", { errorMessage: error.message, stack: error.stack });\n let specificError = error.message;\n // Check for common OpenAI API errors\n if (error.status === 401) specificError = \"OpenAI Authentication failed. Check API key secret.\";\n if (error.status === 429) specificError = \"OpenAI Rate limit or quota exceeded.\";\n if (error.status >= 500) specificError = \"OpenAI server error. Please try again later.\";\n\n return { success: false, error: `Analysis failed: \\${specificError}` };\n }\n };\n\n // --- Request Handling Logic ---\n const url = new URL(req.url);\n const format = url.searchParams.get(\"format\");\n const sourceUrl = import.meta.url; // Used for link back in HTML shell\n\n // --- API Request Branch (POST /?format=json) ---\n if (req.method === \"POST\" && format === \"json\") {\n let requestBody: AnalysisRequestPayload;\n let responsePayload: ServerResponse;\n let status = 200;\n\n try {\n // --- Check OpenAI init status again specifically for API calls ---\n if (!openai) {\n throw new Error(\"OpenAI client is not available. Check server configuration/secrets.\");\n }\n\n requestBody = await req.json();\n const { policyText } = requestBody;\n logServer(\"INFO\", \"Processing POST /?format=json request\", { textLength: policyText?.length });\n\n if (!policyText) {\n throw new Error(\"Missing 'policyText' in request body.\");\n }\n\n // === Execute the Analysis ===\n const analysisResult = await analyzePolicyDocument(policyText);\n\n if (!analysisResult.success || !analysisResult.data) {\n // Use the error from the analysis function\n responsePayload = { nodes: [], edges: [], error: analysisResult.error || \"Policy analysis tool failed.\" };\n status = 400; // Bad request or processing error\n logServer(\"WARN\", \"Analysis failed, returning error to client\", { error: responsePayload.error });\n } else {\n responsePayload = {\n nodes: analysisResult.data.nodes,\n edges: analysisResult.data.edges,\n analysisSummary: analysisResult.data.analysisSummary,\n };\n logServer(\"INFO\", \"Analysis successful, returning data\", {\n nodeCount: responsePayload.nodes.length,\n edgeCount: responsePayload.edges.length,\n });\n }\n } catch (error: any) {\n // Catch errors from JSON parsing, validation, or unexpected issues\n logServer(\"ERROR\", \"Error in POST /?format=json handler\", { errorMessage: error.message });\n status = (error.message.includes(\"Missing 'policyText'\")) ? 400 : 500; // Bad Request or Internal Server Error\n responsePayload = {\n nodes: [],\n edges: [],\n error: error.message || \"An internal server error occurred.\",\n };\n }\n\n // Return the JSON response\n return new Response(JSON.stringify(responsePayload), {\n status: status,\n headers: {\n \"Content-Type\": \"application/json\",\n \"Access-Control-Allow-Origin\": \"*\", // Allow CORS for API usage if needed\n },\n });\n }\n // --- HTML Page Load Branch (GET /) ---\n else if (req.method === \"GET\" && url.pathname === \"/\") {\n // Generate the full HTML page\n const html = generateHtmlShell(sourceUrl);\n return new Response(html, {\n status: 200,\n headers: { \"Content-Type\": \"text/html; charset=utf-8\" },\n });\n }\n // --- Favicon / Other Routes ---\n else if (req.method === \"GET\" && url.pathname === \"/favicon.ico\") {\n return new Response(null, { status: 204 }); // No Content\n }\n // Default 404 for any other unhandled routes/methods\n else {\n logServer(\"WARN\", \"Unhandled route requested\", { path: url.pathname, method: req.method });\n return new Response(\n JSON.stringify({ error: \"Not Found\", message: `Endpoint \\${req.method} \\${url.pathname} not found.` }),\n { status: 404, headers: { \"Content-Type\": \"application/json\" } },\n );\n }\n}","deleted","label",[98],"permalink","https://www.val.town/x/salon/polity/code/main.tsx","extension","actionData","errors"]