Entscheidsuche MCP Server

import { McpServer, ResourceTemplate } from "@modelcontextprotocol/sdk/server/mcp.js"; import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js"; import { z } from "zod"; import fetch from "node-fetch"; import { setTimeout } from "timers/promises"; // Constants const API_BASE_URL = "https://entscheidsuche.ch"; const SEARCH_ENDPOINT = "/_search.php"; const ELASTIC_ENDPOINT = "https://entscheidsuche.pansoft.de:9200/entscheidsuche-*/_search"; const DOCS_BASE_URL = `${API_BASE_URL}/docs`; const STATUS_URL = `${API_BASE_URL}/status`; // Rate limiting constants const REQUEST_DELAY_MS = 500; // Be kind to their server with a 500ms delay between requests // Create the MCP server const server = new McpServer({ name: "entscheidsuche-swiss-legal-server", version: "1.0.0", }); // Utility function to handle rate limiting async function rateLimitedFetch(url: string, options: any = {}) { try { await setTimeout(REQUEST_DELAY_MS); // Add delay to be kind to their server const response = await fetch(url, options); if (!response.ok) { throw new Error(`HTTP error! Status: ${response.status} for URL: ${url}`); } return response; } catch (error) { console.error(`Error fetching ${url}:`, error); throw error; } } // Utility function to parse JSON safely async function safeJsonParse(response: Response) { try { return await response.json(); } catch (error) { console.error("Error parsing JSON:", error); return null; } } // Format document metadata for display function formatDocumentMetadata(doc: any) { try { const metadata = { signature: doc.Signatur || "Unknown", caseNumber: doc.Num || "Unknown", date: doc.Datum || "Unknown", court: doc.Meta?.DE || doc.Kopfzeile?.DE || "Unknown", language: doc.Sprache || "Unknown", abstract: doc.Abstract?.DE || doc.Abstract?.FR || doc.Abstract?.IT || "", pdfUrl: doc.PDFFile ? `${DOCS_BASE_URL}/${doc.PDFFile}` : null, htmlUrl: doc.HTMLFile ? `${DOCS_BASE_URL}/${doc.HTMLFile}` : null, originalUrl: doc.OriginalURL || null, }; return metadata; } catch (error) { console.error("Error formatting document metadata:", error); return { error: "Could not parse document metadata" }; } } // Helper to format Elasticsearch results function formatSearchResults(results: any) { if (!results || !results.hits || !results.hits.hits) { return "No results found."; } const totalHits = results.hits.total?.value || 0; const formattedResults = results.hits.hits.map((hit: any) => { const source = hit._source; return formatDocumentMetadata(source); }); return { totalResults: totalHits, results: formattedResults, }; } // ===== RESOURCES ===== // Resource for listing all available scrapers/collections server.resource( "scrapers-list", "entscheidsuche://scrapers", async (uri) => { const response = await rateLimitedFetch(`${API_BASE_URL}/status`); const html = await response.text(); // Extract scraper information from HTML (simplified - in production use a proper HTML parser) const scraperMatches = html.match(/\/docs\/Index\/([A-Z0-9_]+)\/last/g) || []; const scrapers = scraperMatches.map(match => match.split('/')[3]); return { contents: [{ uri: uri.href, text: JSON.stringify({ scrapers, note: "For details about each scraper, query the scraper-details resource.", statusPageUrl: STATUS_URL }, null, 2) }] }; } ); // Resource for getting details about a specific scraper server.resource( "scraper-details", new ResourceTemplate("entscheidsuche://scraper/{scraperId}", { list: undefined }), async (uri, { scraperId }) => { // Fetch the last index file const indexUrl = `${DOCS_BASE_URL}/Index/${scraperId}/last`; const response = await rateLimitedFetch(indexUrl); const indexData = await safeJsonParse(response); return { contents: [{ uri: uri.href, text: JSON.stringify({ scraperId, lastRunDate: indexData?.Zeit || "Unknown", documentCount: indexData?.Dokumente?.length || 0, jobType: indexData?.Jobtyp || "Unknown", details: indexData || "No data available", indexUrl, documentsUrl: `${DOCS_BASE_URL}/${scraperId}/` }, null, 2) }] }; } ); // Resource for accessing a specific document by signature server.resource( "document-metadata", new ResourceTemplate("entscheidsuche://document/{documentId}", { list: undefined }), async (uri, { documentId }) => { try { // Fetch document metadata JSON const documentUrl = `${DOCS_BASE_URL}/${documentId}.json`; const response = await rateLimitedFetch(documentUrl); const documentData = await safeJsonParse(response); if (!documentData) { return { contents: [{ uri: uri.href, text: JSON.stringify({ error: `Document ${documentId} not found.` }, null, 2) }] }; } const metadata = formatDocumentMetadata(documentData); return { contents: [{ uri: uri.href, text: JSON.stringify(metadata, null, 2) }] }; } catch (error) { return { contents: [{ uri: uri.href, text: JSON.stringify({ error: `Error fetching document ${documentId}: ${error}` }, null, 2) }] }; } } ); // ===== TOOLS ===== // Tool for searching court decisions server.tool( "search-decisions", "Search for court decisions using Elasticsearch query syntax", { query: z.string().describe("Elasticsearch query string (e.g., 'copyright AND music')"), size: z.number().min(1).max(100).default(10).describe("Number of results to return (1-100)"), from: z.number().min(0).default(0).describe("Starting index for pagination"), sort: z.string().optional().describe("Optional sort field and direction (e.g., 'Datum:desc')"), }, async ({ query, size, from, sort }) => { try { // Construct Elasticsearch query const esQuery = { query: { query_string: { query: query } }, size, from, }; if (sort) { const [field, direction] = sort.split(':'); esQuery["sort"] = [{ [field]: { order: direction || 'desc' } }]; } // Make the search request const response = await rateLimitedFetch(ELASTIC_ENDPOINT, { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify(esQuery) }); const searchResults = await safeJsonParse(response); const formattedResults = formatSearchResults(searchResults); return { content: [{ type: "text", text: JSON.stringify(formattedResults, null, 2) }] }; } catch (error) { return { content: [{ type: "text", text: `Error performing search: ${error}` }], isError: true }; } } ); // Tool for getting document content (text/html/pdf) server.tool( "get-document-content", "Retrieve the content of a specific document", { documentId: z.string().describe("Document ID/signature"), format: z.enum(["text", "html"]).default("text").describe("Content format to retrieve (text or html)") }, async ({ documentId, format }) => { try { // First get the document metadata to know what files are available const metadataUrl = `${DOCS_BASE_URL}/${documentId}.json`; const metadataResponse = await rateLimitedFetch(metadataUrl); const metadata = await safeJsonParse(metadataResponse); if (!metadata) { return { content: [{ type: "text", text: `Document ${documentId} not found.` }], isError: true }; } let contentUrl; let contentType; if (format === "html" && metadata.HTMLFile) { contentUrl = `${DOCS_BASE_URL}/${metadata.HTMLFile}`; contentType = "HTML"; } else { // If HTML is not available or text format is requested, use the JSON metadata // This simulates text extraction from the document contentUrl = metadataUrl; contentType = "Metadata"; } const contentResponse = await rateLimitedFetch(contentUrl); let content; if (contentType === "HTML") { content = await contentResponse.text(); // Here you could add additional processing to clean up the HTML if needed } else { // For text format, just return formatted metadata content = JSON.stringify(formatDocumentMetadata(metadata), null, 2); } return { content: [{ type: "text", text: content }] }; } catch (error) { return { content: [{ type: "text", text: `Error retrieving document content: ${error}` }], isError: true }; } } ); // Tool for listing courts by canton server.tool( "list-courts", "List available courts by canton", { canton: z.string().optional().describe("Optional two-letter canton code (e.g., 'ZH' for Zürich, 'BE' for Bern)") }, async ({ canton }) => { try { // Fetch status page which contains information about courts const response = await rateLimitedFetch(STATUS_URL); const html = await response.text(); // In a real implementation, properly parse the HTML to extract court information // This is a simplified version const cantonPattern = canton ? new RegExp(`${canton}[^<]*?</a>.*?<ul>(.*?)</ul>`, 's') : /<h3>.*?<a[^>]*>(.*?)<\/a>.*?<ul>(.*?)<\/ul>/g; let courts = {}; if (canton) { const match = html.match(cantonPattern); if (match) { const courtMatches = match[1].match(/<li>(.*?)<\/li>/g) || []; courts[canton] = courtMatches.map(m => m.replace(/<[^>]*>/g, '').trim()); } } else { let match; const fullPattern = /<h3>.*?<a[^>]*>(.*?)<\/a>.*?<ul>(.*?)<\/ul>/gs; while ((match = fullPattern.exec(html)) !== null) { const cantonName = match[1].trim(); const courtMatches = match[2].match(/<li>(.*?)<\/li>/g) || []; courts[cantonName] = courtMatches.map(m => m.replace(/<[^>]*>/g, '').trim()); } } return { content: [{ type: "text", text: JSON.stringify(courts, null, 2) }] }; } catch (error) { return { content: [{ type: "text", text: `Error listing courts: ${error}` }], isError: true }; } } ); // Tool for getting document URLs server.tool( "get-document-urls", "Get direct URLs for a document's PDF and HTML versions", { documentId: z.string().describe("Document ID/signature") }, async ({ documentId }) => { try { const metadataUrl = `${DOCS_BASE_URL}/${documentId}.json`; const response = await rateLimitedFetch(metadataUrl); const metadata = await safeJsonParse(response); if (!metadata) { return { content: [{ type: "text", text: `Document ${documentId} not found.` }], isError: true }; } const urls = { documentId, pdfUrl: metadata.PDFFile ? `${DOCS_BASE_URL}/${metadata.PDFFile}` : null, htmlUrl: metadata.HTMLFile ? `${DOCS_BASE_URL}/${metadata.HTMLFile}` : null, originalUrl: metadata.OriginalURL || null, jsonUrl: metadataUrl }; return { content: [{ type: "text", text: JSON.stringify(urls, null, 2) }] }; } catch (error) { return { content: [{ type: "text", text: `Error getting document URLs: ${error}` }], isError: true }; } } ); // ===== PROMPTS ===== // Prompt for searching precedents on a legal topic server.prompt( "search-legal-precedents", "Find relevant precedents on a specific legal topic", { topic: z.string().describe("Legal topic or keywords to search for"), jurisdiction: z.string().optional().describe("Optional: Canton code or court to restrict search") }, ({ topic, jurisdiction }) => { let queryString = topic; if (jurisdiction) { queryString = `(${topic}) AND (${jurisdiction})`; } return { messages: [ { role: "user", content: { type: "text", text: `Please search for relevant Swiss legal precedents about "${topic}"${jurisdiction ? ` in jurisdiction ${jurisdiction}` : ""} and analyze the most significant findings. For each relevant case, provide a summary of the facts, the legal reasoning, and the outcome.` } } ] }; } ); // Prompt for comparing legal rulings across jurisdictions server.prompt( "compare-jurisdictions", "Compare rulings on a specific legal issue across different cantons", { legalIssue: z.string().describe("Legal issue to compare"), cantons: z.string().describe("Comma-separated list of canton codes to compare (e.g., 'ZH,BE,GE')") }, ({ legalIssue, cantons }) => { const cantonList = cantons.split(',').map(c => c.trim()).join('", "'); return { messages: [ { role: "user", content: { type: "text", text: `I'd like to understand how different Swiss cantons have ruled on the issue of "${legalIssue}". Please search for relevant court decisions in these cantons: "${cantonList}". Then compare and contrast the approaches, highlighting any significant differences in legal interpretation or application.` } } ] }; } ); // Prompt for retrieving decisions from a specific court server.prompt( "court-decisions", "Retrieve recent decisions from a specific court", { court: z.string().describe("Court name or identifier"), timeframe: z.string().default("1 year").describe("Timeframe to search within (e.g., '6 months', '2 years')") }, ({ court, timeframe }) => { return { messages: [ { role: "user", content: { type: "text", text: `Please find and list the most important decisions from ${court} within the last ${timeframe}. For each decision, provide the case number, date, a brief summary of the legal issue, and outcome.` } } ] }; } ); // Start the server with a stdio transport const transport = new StdioServerTransport(); server.connect(transport).then(() => { console.error("Entscheidsuche MCP server started"); }).catch(error => { console.error("Error starting server:", error); });