nia-mcp-server 1.0.1__tar.gz → 1.0.3__tar.gz

{nia_mcp_server-1.0.1 → nia_mcp_server-1.0.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nia-mcp-server
-Version: 1.0.1
+Version: 1.0.3
 Summary: NIA Knowledge Agent - MCP server for intelligent codebase search
 Project-URL: Homepage, https://trynia.ai
 Project-URL: Documentation, https://docs.trynia.ai
@@ -20,7 +20,6 @@ Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 Requires-Python: >=3.8
-Requires-Dist: exa-py>=1.0.8
 Requires-Dist: httpx>=0.24.0
 Requires-Dist: mcp>=0.1.0
 Requires-Dist: pydantic>=2.0.0

{nia_mcp_server-1.0.1 → nia_mcp_server-1.0.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "nia-mcp-server"
-version = "1.0.1"
+version = "1.0.3"
 description = "NIA Knowledge Agent - MCP server for intelligent codebase search"
 readme = "README.md"
 requires-python = ">=3.8"
@@ -30,7 +30,6 @@ dependencies = [
     "httpx>=0.24.0",
     "pydantic>=2.0.0",
     "python-dotenv>=1.0.0",
-    "exa-py>=1.0.8",
 ]
 
 [project.urls]

{nia_mcp_server-1.0.1 → nia_mcp_server-1.0.3}/src/nia_mcp_server/__init__.py

@@ -2,4 +2,4 @@
 NIA MCP Server - Proxy server for NIA Knowledge Agent
 """
 
-__version__ = "1.0.1"
+__version__ = "1.0.3"

{nia_mcp_server-1.0.1 → nia_mcp_server-1.0.3}/src/nia_mcp_server/api_client.py

@@ -31,7 +31,7 @@ class NIAApiClient:
                 "User-Agent": "nia-mcp-server/1.0.0",
                 "Content-Type": "application/json"
             },
-            timeout=300.0  # 5 minute timeout for long operations
+            timeout=720.0  # 12 minute timeout for deep research operations
         )
     
     async def close(self):
@@ -474,4 +474,65 @@ class NIAApiClient:
         except httpx.HTTPStatusError as e:
             raise self._handle_api_error(e)
         except Exception as e:
-            raise APIError(f"Query failed: {str(e)}")
+            raise APIError(f"Query failed: {str(e)}")
+    
+    async def web_search(
+        self,
+        query: str,
+        num_results: int = 5,
+        category: Optional[str] = None,
+        days_back: Optional[int] = None,
+        find_similar_to: Optional[str] = None
+    ) -> Dict[str, Any]:
+        """Perform AI-powered web search."""
+        try:
+            payload = {
+                "query": query,
+                "num_results": min(num_results, 10),
+            }
+            
+            # Add optional parameters
+            if category:
+                payload["category"] = category
+            if days_back:
+                payload["days_back"] = days_back
+            if find_similar_to:
+                payload["find_similar_to"] = find_similar_to
+            
+            response = await self.client.post(
+                f"{self.base_url}/v2/web-search",
+                json=payload
+            )
+            response.raise_for_status()
+            return response.json()
+            
+        except httpx.HTTPStatusError as e:
+            raise self._handle_api_error(e)
+        except Exception as e:
+            raise APIError(f"Web search failed: {str(e)}")
+    
+    async def deep_research(
+        self,
+        query: str,
+        output_format: Optional[str] = None
+    ) -> Dict[str, Any]:
+        """Perform deep research using AI agent."""
+        try:
+            payload = {
+                "query": query,
+            }
+            
+            if output_format:
+                payload["output_format"] = output_format
+            
+            response = await self.client.post(
+                f"{self.base_url}/v2/deep-research",
+                json=payload
+            )
+            response.raise_for_status()
+            return response.json()
+            
+        except httpx.HTTPStatusError as e:
+            raise self._handle_api_error(e)
+        except Exception as e:
+            raise APIError(f"Deep research failed: {str(e)}")

{nia_mcp_server-1.0.1 → nia_mcp_server-1.0.3}/src/nia_mcp_server/server.py

@@ -755,215 +755,132 @@ async def nia_web_search(
         Search results with actionable next steps
     """
     try:
-        # Check for API key
-        api_key = os.getenv("EXA_API_KEY")
-        if not api_key:
-            return [TextContent(
-                type="text",
-                text="❌ NIA Web Search unavailable.\n\n"
-                     "This feature requires additional configuration. "
-                     "Please contact support for access to advanced search features."
-            )]
-        
-        # Import client
-        try:
-            from exa_py import Exa
-        except ImportError:
-            return [TextContent(
-                type="text",
-                text="❌ NIA Web Search unavailable. Please update the NIA MCP server."
-            )]
-        
-        # Initialize client
-        client = Exa(api_key)
-        
-        # Limit results to reasonable number
-        num_results = min(num_results, 10)
+        client = await ensure_api_client()
         
         logger.info(f"Searching content for query: {query}")
         
-        # Build search parameters
-        search_params = {
-            "num_results": num_results * 2,  # Get more to filter
-            "type": "auto",  # Automatically choose best search type
-            "text": True,
-            "highlights": True
-        }
-        
-        # Add category filter if specified
-        if category:
-            # Map user-friendly categories to Exa categories
-            category_map = {
-                "github": "github",
-                "company": "company", 
-                "research": "research paper",
-                "news": "news",
-                "tweet": "tweet",
-                "pdf": "pdf",
-                "blog": "personal site"
-            }
-            if category.lower() in category_map:
-                search_params["category"] = category_map[category.lower()]
-        
-        # Add date filter for trending content
-        if days_back:
-            from datetime import datetime, timedelta
-            start_date = (datetime.now() - timedelta(days=days_back)).strftime("%Y-%m-%d")
-            search_params["start_published_date"] = start_date
-        
-        # Use similarity search if URL provided
-        if find_similar_to:
-            results = client.find_similar_and_contents(
-                find_similar_to,
-                **search_params
-            )
-        else:
-            results = client.search_and_contents(
-                query,
-                **search_params
-            )
-        
-        # Separate results by type
-        github_repos = []
-        documentation = []
-        other_content = []
+        # Use the API client method instead of direct HTTP call
+        result = await client.web_search(
+            query=query,
+            num_results=num_results,
+            category=category,
+            days_back=days_back,
+            find_similar_to=find_similar_to
+        )
         
-        for result in results.results:
-            url = result.url
-            
-            # Categorize based on URL patterns
-            if "github.com" in url and "/tree/" not in url and "/blob/" not in url:
-                # It's a GitHub repo (not a specific file)
-                # Extract owner/repo from URL
-                try:
-                    parsed = urlparse(url)
-                    # Ensure we have a valid GitHub URL
-                    if parsed.hostname == "github.com":
-                        # Remove leading/trailing slashes and split the path
-                        path_parts = parsed.path.strip("/").split("/")
-                        # Verify we have at least owner and repo in the path
-                        if len(path_parts) >= 2 and path_parts[0] and path_parts[1]:
-                            # Extract only the owner and repo, ignoring any additional path components
-                            owner_repo = f"{path_parts[0]}/{path_parts[1]}"
-                            github_repos.append({
-                                "url": url,
-                                "owner_repo": owner_repo,
-                                "title": result.title or owner_repo,
-                                "summary": result.text[:200] if result.text else "",
-                                "highlights": result.highlights[:2] if result.highlights else [],
-                                "published_date": getattr(result, 'published_date', None)
-                            })
-                except:
-                    pass
-            elif any(doc_pattern in url for doc_pattern in ["docs.", "documentation", ".readthedocs.", "/docs/"]):
-                # It's documentation
-                documentation.append({
-                    "url": url,
-                    "title": result.title or "Documentation",
-                    "summary": result.text[:200] if result.text else "",
-                    "highlights": result.highlights[:2] if result.highlights else []
-                })
-            else:
-                # Other content
-                other_content.append({
-                    "url": url,
-                    "title": result.title or "Content",
-                    "summary": result.text[:200] if result.text else ""
-                })
+        # Extract results
+        github_repos = result.get("github_repos", [])
+        documentation = result.get("documentation", [])
+        other_content = result.get("other_content", [])
         
         # Format response to naturally guide next actions
-        response = f"## 🔍 NIA Web Search Results for: \"{query}\"\n\n"
+        response_text = f"## 🔍 NIA Web Search Results for: \"{query}\"\n\n"
         
         if days_back:
-            response += f"*Showing results from the last {days_back} days*\n\n"
+            response_text += f"*Showing results from the last {days_back} days*\n\n"
         
         if find_similar_to:
-            response += f"*Finding content similar to: {find_similar_to}*\n\n"
+            response_text += f"*Finding content similar to: {find_similar_to}*\n\n"
         
         # GitHub Repositories Section
         if github_repos:
-            response += f"### 📦 GitHub Repositories ({len(github_repos)} found)\n\n"
+            response_text += f"### 📦 GitHub Repositories ({len(github_repos)} found)\n\n"
             
             for i, repo in enumerate(github_repos[:num_results], 1):
-                response += f"**{i}. {repo['title']}**\n"
-                response += f"   📍 `{repo['url']}`\n"
+                response_text += f"**{i}. {repo['title']}**\n"
+                response_text += f"   📍 `{repo['url']}`\n"
                 if repo.get('published_date'):
-                    response += f"   📅 Updated: {repo['published_date']}\n"
+                    response_text += f"   📅 Updated: {repo['published_date']}\n"
                 if repo['summary']:
-                    response += f"   📝 {repo['summary']}...\n"
+                    response_text += f"   📝 {repo['summary']}...\n"
                 if repo['highlights']:
-                    response += f"   ✨ Key features: {', '.join(repo['highlights'])}\n"
-                response += "\n"
+                    response_text += f"   ✨ Key features: {', '.join(repo['highlights'])}\n"
+                response_text += "\n"
             
             # Be more aggressive based on query specificity
             if len(github_repos) == 1 or any(specific_word in query.lower() for specific_word in ["specific", "exact", "particular", "find me", "looking for"]):
-                response += "**🚀 RECOMMENDED ACTION - Index this repository with NIA:**\n"
-                response += f"```\nIndex {github_repos[0]['owner_repo']}\n```\n"
-                response += "✨ This will enable AI-powered code search, understanding, and analysis!\n\n"
+                response_text += "**🚀 RECOMMENDED ACTION - Index this repository with NIA:**\n"
+                response_text += f"```\nIndex {github_repos[0]['owner_repo']}\n```\n"
+                response_text += "✨ This will enable AI-powered code search, understanding, and analysis!\n\n"
             else:
-                response += "**🚀 Make these repositories searchable with NIA's AI:**\n"
-                response += f"- **Quick start:** Say \"Index {github_repos[0]['owner_repo']}\"\n"
-                response += "- **Index multiple:** Say \"Index all repositories\"\n"
-                response += "- **Benefits:** AI-powered code search, architecture understanding, implementation details\n\n"
+                response_text += "**🚀 Make these repositories searchable with NIA's AI:**\n"
+                response_text += f"- **Quick start:** Say \"Index {github_repos[0]['owner_repo']}\"\n"
+                response_text += "- **Index multiple:** Say \"Index all repositories\"\n"
+                response_text += "- **Benefits:** AI-powered code search, architecture understanding, implementation details\n\n"
         
         # Documentation Section
         if documentation:
-            response += f"### 📚 Documentation ({len(documentation)} found)\n\n"
+            response_text += f"### 📚 Documentation ({len(documentation)} found)\n\n"
             
             for i, doc in enumerate(documentation[:num_results], 1):
-                response += f"**{i}. {doc['title']}**\n"
-                response += f"   📍 `{doc['url']}`\n"
+                response_text += f"**{i}. {doc['title']}**\n"
+                response_text += f"   📍 `{doc['url']}`\n"
                 if doc['summary']:
-                    response += f"   📝 {doc['summary']}...\n"
+                    response_text += f"   📝 {doc['summary']}...\n"
                 if doc.get('highlights'):
-                    response += f"   ✨ Key topics: {', '.join(doc['highlights'])}\n"
-                response += "\n"
+                    response_text += f"   ✨ Key topics: {', '.join(doc['highlights'])}\n"
+                response_text += "\n"
             
             # Be more aggressive for documentation too
             if len(documentation) == 1 or any(specific_word in query.lower() for specific_word in ["docs", "documentation", "guide", "tutorial", "reference"]):
-                response += "**📖 RECOMMENDED ACTION - Index this documentation with NIA:**\n"
-                response += f"```\nIndex documentation {documentation[0]['url']}\n```\n"
-                response += "✨ NIA will make this fully searchable with AI-powered Q&A!\n\n"
+                response_text += "**📖 RECOMMENDED ACTION - Index this documentation with NIA:**\n"
+                response_text += f"```\nIndex documentation {documentation[0]['url']}\n```\n"
+                response_text += "✨ NIA will make this fully searchable with AI-powered Q&A!\n\n"
             else:
-                response += "**📖 Make this documentation AI-searchable with NIA:**\n"
-                response += f"- **Quick start:** Say \"Index documentation {documentation[0]['url']}\"\n"
-                response += "- **Index all:** Say \"Index all documentation\"\n"
-                response += "- **Benefits:** Instant answers, smart search, code examples extraction\n\n"
+                response_text += "**📖 Make this documentation AI-searchable with NIA:**\n"
+                response_text += f"- **Quick start:** Say \"Index documentation {documentation[0]['url']}\"\n"
+                response_text += "- **Index all:** Say \"Index all documentation\"\n"
+                response_text += "- **Benefits:** Instant answers, smart search, code examples extraction\n\n"
         
         # Other Content Section
         if other_content and not github_repos and not documentation:
-            response += f"### 🌐 Other Content ({len(other_content)} found)\n\n"
+            response_text += f"### 🌐 Other Content ({len(other_content)} found)\n\n"
             
             for i, content in enumerate(other_content[:num_results], 1):
-                response += f"**{i}. {content['title']}**\n"
-                response += f"   📍 `{content['url']}`\n"
+                response_text += f"**{i}. {content['title']}**\n"
+                response_text += f"   📍 `{content['url']}`\n"
                 if content['summary']:
-                    response += f"   📝 {content['summary']}...\n"
-                response += "\n"
+                    response_text += f"   📝 {content['summary']}...\n"
+                response_text += "\n"
         
         # No results found
         if not github_repos and not documentation and not other_content:
-            response = f"No results found for '{query}'. Try:\n"
-            response += "- Using different keywords\n"
-            response += "- Being more specific (e.g., 'Python RAG implementation')\n"
-            response += "- Including technology names (e.g., 'LangChain', 'TypeScript')\n"
+            response_text = f"No results found for '{query}'. Try:\n"
+            response_text += "- Using different keywords\n"
+            response_text += "- Being more specific (e.g., 'Python RAG implementation')\n"
+            response_text += "- Including technology names (e.g., 'LangChain', 'TypeScript')\n"
         
         # Add prominent call-to-action if we found indexable content
         if github_repos or documentation:
-            response += "\n## 🎯 **Ready to unlock NIA's AI capabilities?**\n"
-            response += "The repositories and documentation above can be indexed for:\n"
-            response += "- 🤖 AI-powered code understanding and search\n"
-            response += "- 💡 Instant answers to technical questions\n"
-            response += "- 🔍 Deep architectural insights\n"
-            response += "- 📚 Smart documentation Q&A\n\n"
-            response += "**Just copy and paste the index commands above!**\n"
+            response_text += "\n## 🎯 **Ready to unlock NIA's AI capabilities?**\n"
+            response_text += "The repositories and documentation above can be indexed for:\n"
+            response_text += "- 🤖 AI-powered code understanding and search\n"
+            response_text += "- 💡 Instant answers to technical questions\n"
+            response_text += "- 🔍 Deep architectural insights\n"
+            response_text += "- 📚 Smart documentation Q&A\n\n"
+            response_text += "**Just copy and paste the index commands above!**\n"
         
         # Add search metadata
-        response += f"\n---\n"
-        response += f"*Searched {len(results.results)} sources using NIA Web Search*"
+        response_text += f"\n---\n"
+        response_text += f"*Searched {result.get('total_results', 0)} sources using NIA Web Search*"
         
-        return [TextContent(type="text", text=response)]
+        return [TextContent(type="text", text=response_text)]
         
+    except APIError as e:
+        logger.error(f"API Error in web search: {e}")
+        if e.status_code == 403 or "free tier limit" in str(e).lower() or "free api requests" in str(e).lower():
+            if e.detail and "25 free API requests" in e.detail:
+                return [TextContent(
+                    type="text", 
+                    text=f"❌ {e.detail}\n\n💡 Tip: Upgrade to Pro at https://trynia.ai/billing for unlimited API access."
+                )]
+            else:
+                return [TextContent(
+                    type="text",
+                    text=f"❌ {str(e)}\n\n💡 Tip: You've reached the free tier limit. Upgrade to Pro for unlimited access."
+                )]
+        else:
+            return [TextContent(type="text", text=f"❌ {str(e)}")]
     except Exception as e:
         logger.error(f"Error in NIA web search: {str(e)}")
         return [TextContent(
@@ -1011,154 +928,138 @@ async def nia_deep_research_agent(
         Comprehensive research results with citations
     """
     try:
-        # Check for API key
-        api_key = os.getenv("EXA_API_KEY")
-        if not api_key:
-            return [TextContent(
-                type="text",
-                text="❌ Deep research unavailable. This advanced feature requires additional configuration."
-            )]
+        client = await ensure_api_client()
         
-        # Import client
+        logger.info(f"Starting deep research for: {query}")
+        
+        # Use the API client method with proper timeout handling
         try:
-            from exa_py import Exa
-            import json
-        except ImportError:
+            result = await asyncio.wait_for(
+                client.deep_research(query=query, output_format=output_format),
+                timeout=720.0  # 12 minutes to allow for longer research tasks
+            )
+        except asyncio.TimeoutError:
+            logger.error(f"Deep research timed out after 12 minutes for query: {query}")
             return [TextContent(
                 type="text",
-                text="❌ Research service unavailable. Please update the NIA MCP server."
+                text="❌ Research timed out. The query may be too complex. Try:\n"
+                     "- Breaking it into smaller questions\n"  
+                     "- Using more specific keywords\n"
+                     "- Trying the nia_web_search tool for simpler queries"
             )]
         
-        # Initialize client
-        client = Exa(api_key)
-        
-        logger.info(f"Starting deep research for: {query}")
+        # Format the research results
+        response_text = f"## 🔬 NIA Deep Research Agent Results\n\n"
+        response_text += f"**Query:** {query}\n\n"
         
-        # Create a research task
-        try:
-            # Let the AI infer the schema based on the query
-            task = client.research.create_task(
-                instructions=query,
-                infer_schema=True,
-                model="exa-research-pro"  # Use the pro model
-            )
+        if result.get("data"):
+            response_text += "### 📊 Research Findings:\n\n"
             
-            logger.info(f"Research task created: {task.id}")
+            # Pretty print the JSON data
+            import json
+            formatted_data = json.dumps(result["data"], indent=2)
+            response_text += f"```json\n{formatted_data}\n```\n\n"
             
-            # Poll for completion (max 2 minutes)
-            result = client.research.poll_task(
-                task.id,
-                poll_interval=3,
-                max_wait_time=120
-            )
+            # Add citations if available
+            if result.get("citations"):
+                response_text += "### 📚 Sources & Citations:\n\n"
+                citation_num = 1
+                for field, citations in result["citations"].items():
+                    if citations:
+                        response_text += f"**{field}:**\n"
+                        for citation in citations[:3]:  # Limit to 3 citations per field
+                            response_text += f"{citation_num}. [{citation.get('title', 'Source')}]({citation.get('url', '#')})\n"
+                            if citation.get('snippet'):
+                                response_text += f"   > {citation['snippet'][:150]}...\n"
+                            citation_num += 1
+                        response_text += "\n"
             
-            if result.status == "failed":
-                return [TextContent(
-                    type="text",
-                    text=f"❌ Research failed. Please try rephrasing your question."
-                )]
+            response_text += "### 💡 RECOMMENDED NEXT ACTIONS WITH NIA:\n\n"
             
-            # Format the research results
-            response = f"## 🔬 NIA Deep Research Agent Results\n\n"
-            response += f"**Query:** {query}\n\n"
+            # Extract potential repos and docs from the research data
+            repos_found = []
+            docs_found = []
             
-            if result.data:
-                response += "### 📊 Research Findings:\n\n"
-                
-                # Pretty print the JSON data
-                formatted_data = json.dumps(result.data, indent=2)
-                response += f"```json\n{formatted_data}\n```\n\n"
+            # Helper function to extract URLs from nested data structures
+            def extract_urls_from_data(data, urls_list=None):
+                if urls_list is None:
+                    urls_list = []
                 
-                # Add citations if available
-                if result.citations:
-                    response += "### 📚 Sources & Citations:\n\n"
-                    citation_num = 1
-                    for field, citations in result.citations.items():
-                        if citations:
-                            response += f"**{field}:**\n"
-                            for citation in citations[:3]:  # Limit to 3 citations per field
-                                response += f"{citation_num}. [{citation.get('title', 'Source')}]({citation.get('url', '#')})\n"
-                                if citation.get('snippet'):
-                                    response += f"   > {citation['snippet'][:150]}...\n"
-                                citation_num += 1
-                            response += "\n"
+                if isinstance(data, dict):
+                    for value in data.values():
+                        extract_urls_from_data(value, urls_list)
+                elif isinstance(data, list):
+                    for item in data:
+                        extract_urls_from_data(item, urls_list)
+                elif isinstance(data, str):
+                    # Check if this string is a URL
+                    if data.startswith(('http://', 'https://')):
+                        urls_list.append(data)
                 
-                response += "### 💡 RECOMMENDED NEXT ACTIONS WITH NIA:\n\n"
-                
-                # Extract potential repos and docs from the research data
-                repos_found = []
-                docs_found = []
-                
-                # Helper function to extract URLs from nested data structures
-                def extract_urls_from_data(data, urls_list=None):
-                    if urls_list is None:
-                        urls_list = []
-                    
-                    if isinstance(data, dict):
-                        for value in data.values():
-                            extract_urls_from_data(value, urls_list)
-                    elif isinstance(data, list):
-                        for item in data:
-                            extract_urls_from_data(item, urls_list)
-                    elif isinstance(data, str):
-                        # Check if this string is a URL
-                        if data.startswith(('http://', 'https://')):
-                            urls_list.append(data)
-                    
-                    return urls_list
-                
-                # Extract all URLs from the data
-                all_urls = extract_urls_from_data(result.data)
-                
-                # Filter for GitHub repos and documentation
-                import re
-                github_pattern = r'github\.com/([a-zA-Z0-9-]+/[a-zA-Z0-9-_.]+)'
-                
-                for url in all_urls:
-                    # Check for GitHub repos
-                    github_match = re.search(github_pattern, url)
-                    if github_match and '/tree/' not in url and '/blob/' not in url:
-                        repos_found.append(github_match.group(1))
-                    # Check for documentation URLs
-                    elif any(doc_indicator in url.lower() for doc_indicator in ['docs', 'documentation', '.readthedocs.', '/guide', '/tutorial']):
-                        docs_found.append(url)
-                
-                # Remove duplicates and limit results
-                repos_found = list(set(repos_found))[:3]
-                docs_found = list(set(docs_found))[:3]
-                
-                if repos_found:
-                    response += "**🚀 DISCOVERED REPOSITORIES - Index with NIA for deep analysis:**\n"
-                    for repo in repos_found:
-                        response += f"```\nIndex {repo}\n```\n"
-                    response += "✨ Enable AI-powered code search and architecture understanding!\n\n"
-                
-                if docs_found:
-                    response += "**📖 DISCOVERED DOCUMENTATION - Index with NIA for smart search:**\n"
-                    for doc in docs_found[:2]:  # Limit to 2 for readability
-                        response += f"```\nIndex documentation {doc}\n```\n"
-                    response += "✨ Make documentation instantly searchable with AI Q&A!\n\n"
-                
-                if not repos_found and not docs_found:
-                    response += "**🔍 Manual indexing options:**\n"
-                    response += "- If you see any GitHub repos mentioned: Say \"Index [owner/repo]\"\n"
-                    response += "- If you see any documentation sites: Say \"Index documentation [url]\"\n"
-                    response += "- These will unlock NIA's powerful AI search capabilities!\n\n"
-                
-                response += "**📊 Other actions:**\n"
-                response += "- Ask follow-up questions about the research\n"
-                response += "- Request a different analysis format\n"
-                response += "- Search for more specific information\n"
-            else:
-                response += "No structured data returned. The research may need a more specific query."
+                return urls_list
+            
+            # Extract all URLs from the data
+            all_urls = extract_urls_from_data(result["data"])
             
-            return [TextContent(type="text", text=response)]
+            # Filter for GitHub repos and documentation
+            import re
+            github_pattern = r'github\.com/([a-zA-Z0-9-]+/[a-zA-Z0-9-_.]+)'
             
-        except Exception as e:
-            # Fallback to regular search if research API fails
-            logger.warning(f"Research API failed, falling back to search: {e}")
-            return await nia_web_search(query, num_results=10)
+            for url in all_urls:
+                # Check for GitHub repos
+                github_match = re.search(github_pattern, url)
+                if github_match and '/tree/' not in url and '/blob/' not in url:
+                    repos_found.append(github_match.group(1))
+                # Check for documentation URLs
+                elif any(doc_indicator in url.lower() for doc_indicator in ['docs', 'documentation', '.readthedocs.', '/guide', '/tutorial']):
+                    docs_found.append(url)
             
+            # Remove duplicates and limit results
+            repos_found = list(set(repos_found))[:3]
+            docs_found = list(set(docs_found))[:3]
+            
+            if repos_found:
+                response_text += "**🚀 DISCOVERED REPOSITORIES - Index with NIA for deep analysis:**\n"
+                for repo in repos_found:
+                    response_text += f"```\nIndex {repo}\n```\n"
+                response_text += "✨ Enable AI-powered code search and architecture understanding!\n\n"
+            
+            if docs_found:
+                response_text += "**📖 DISCOVERED DOCUMENTATION - Index with NIA for smart search:**\n"
+                for doc in docs_found[:2]:  # Limit to 2 for readability
+                    response_text += f"```\nIndex documentation {doc}\n```\n"
+                response_text += "✨ Make documentation instantly searchable with AI Q&A!\n\n"
+            
+            if not repos_found and not docs_found:
+                response_text += "**🔍 Manual indexing options:**\n"
+                response_text += "- If you see any GitHub repos mentioned: Say \"Index [owner/repo]\"\n"
+                response_text += "- If you see any documentation sites: Say \"Index documentation [url]\"\n"
+                response_text += "- These will unlock NIA's powerful AI search capabilities!\n\n"
+            
+            response_text += "**📊 Other actions:**\n"
+            response_text += "- Ask follow-up questions about the research\n"
+            response_text += "- Request a different analysis format\n"
+            response_text += "- Search for more specific information\n"
+        else:
+            response_text += "No structured data returned. The research may need a more specific query."
+        
+        return [TextContent(type="text", text=response_text)]
+        
+    except APIError as e:
+        logger.error(f"API Error in deep research: {e}")
+        if e.status_code == 403 or "free tier limit" in str(e).lower() or "free api requests" in str(e).lower():
+            if e.detail and "25 free API requests" in e.detail:
+                return [TextContent(
+                    type="text", 
+                    text=f"❌ {e.detail}\n\n💡 Tip: Upgrade to Pro at https://trynia.ai/billing for unlimited API access."
+                )]
+            else:
+                return [TextContent(
+                    type="text",
+                    text=f"❌ {str(e)}\n\n💡 Tip: You've reached the free tier limit. Upgrade to Pro for unlimited access."
+                )]
+        else:
+            return [TextContent(type="text", text=f"❌ {str(e)}")]
     except Exception as e:
         logger.error(f"Error in deep research: {str(e)}")
         return [TextContent(