Shopify Liquid MCP Server

Instructions

Implementation Reference

• shopify_liquid_mcp/server.py:22-54 (handler)

  The handler function for the 'search_liquid_docs' tool, decorated with @mcp.tool() which also serves as registration. It processes the input queries, calls the helper search_documentation, and formats the search results into a readable string output.

  @mcp.tool()
  def search_liquid_docs(queries: List[str]) -> str:
      """Search Shopify Liquid documentation using full-text search.
  
      Args:
          queries: List of search terms (maximum 3)
  
      Returns:
          Formatted search results with snippets
      """
      if not queries:
          return "Error: Please provide at least one search query"
  
      # Limit to 3 queries
      queries = queries[:3]
  
      logger.info(f"Searching for: {queries}")
      results = search_documentation(queries, limit=10)
  
      if not results:
          return f"No results found for: {', '.join(queries)}"
  
      # Format results
      output = [f"Found {len(results)} results for: {', '.join(queries)}\n"]
  
      for i, doc in enumerate(results, 1):
          output.append(f"{i}. **{doc['title']}** ({doc['category']})")
          output.append(f"   Path: {doc['path']}")
          if doc.get("snippet"):
              output.append(f"   {doc['snippet']}")
          output.append("")
  
      return "\n".join(output)

• shopify_liquid_mcp/ingest.py:201-244 (helper)

  Helper utility that implements the core full-text search logic using SQLite FTS5 virtual table, generating snippets with highlights, and returning structured results used by the tool handler.

  def search_documentation(queries: List[str], limit: int = 10) -> List[Dict[str, str]]:
      """Search documentation using FTS5.
  
      Args:
          queries: List of search terms
          limit: Maximum number of results to return
  
      Returns:
          List of matching documents with metadata
      """
      conn = sqlite3.connect(DB_PATH)
      cursor = conn.cursor()
  
      # Build FTS5 query
      search_query = " OR ".join(queries[:3])  # Limit to 3 queries like Gemini example
  
      cursor.execute(
          f"""
          SELECT d.name, d.title, d.category, d.content, d.path,
                 snippet({FTS_TABLE}, 3, '<mark>', '</mark>', '...', 64) as snippet
          FROM {FTS_TABLE} fts
          JOIN {DOCS_TABLE} d ON fts.rowid = d.id
          WHERE {FTS_TABLE} MATCH ?
          ORDER BY rank
          LIMIT ?
      """,
          (search_query, limit),
      )
  
      results = []
      for row in cursor.fetchall():
          results.append(
              {
                  "name": row[0],
                  "title": row[1],
                  "category": row[2],
                  "content": row[3],
                  "path": row[4],
                  "snippet": row[5],
              }
          )
  
      conn.close()
      return results