nia-mcp-server 1.0.14__tar.gz → 1.0.16__tar.gz

{nia_mcp_server-1.0.14 → nia_mcp_server-1.0.16}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nia-mcp-server
-Version: 1.0.14
+Version: 1.0.16
 Summary: Nia Knowledge Agent
 Project-URL: Homepage, https://trynia.ai
 Project-URL: Documentation, https://docs.trynia.ai

{nia_mcp_server-1.0.14 → nia_mcp_server-1.0.16}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "nia-mcp-server"
-version = "1.0.14"
+version = "1.0.16"
 description = "Nia Knowledge Agent"
 readme = "README.md"
 requires-python = ">=3.8"

{nia_mcp_server-1.0.14 → nia_mcp_server-1.0.16}/src/nia_mcp_server/__init__.py

@@ -2,4 +2,4 @@
 NIA MCP Server - Proxy server for NIA Knowledge Agent
 """
 
-__version__ = "1.0.13"
+__version__ = "1.0.16"

{nia_mcp_server-1.0.14 → nia_mcp_server-1.0.16}/src/nia_mcp_server/api_client.py

@@ -159,15 +159,33 @@ class NIAApiClient:
     async def index_repository(self, repo_url: str, branch: str = None) -> Dict[str, Any]:
         """Index a GitHub repository."""
         try:
-            # Parse repo URL to get owner/repo format
+            # Handle different input formats
             if "github.com" in repo_url:
-                parts = repo_url.rstrip('/').split('/')
-                owner_repo = f"{parts[-2]}/{parts[-1]}"
+                # Remove query parameters and fragments
+                clean_url = repo_url.split('?')[0].split('#')[0]
+                
+                # Check if it's a folder URL (contains /tree/)
+                if "/tree/" in clean_url:
+                    # Extract everything after github.com/
+                    parts = clean_url.split('github.com/', 1)
+                    if len(parts) > 1:
+                        repository_path = parts[1].rstrip('/')
+                    else:
+                        repository_path = repo_url
+                else:
+                    # Regular repo URL - extract owner/repo
+                    parts = clean_url.rstrip('/').split('/')
+                    if len(parts) >= 2:
+                        repo_name = parts[-1].rstrip('.git')  # Remove .git suffix
+                        repository_path = f"{parts[-2]}/{repo_name}"
+                    else:
+                        repository_path = repo_url
             else:
-                owner_repo = repo_url
+                # Assume it's already in the right format
+                repository_path = repo_url
             
             payload = {
-                "repository": owner_repo,
+                "repository": repository_path,
                 "branch": branch
             }
             
@@ -191,11 +209,29 @@ class NIAApiClient:
                 # First, list all repositories to find the matching one
                 repos = await self.list_repositories()
                 
+                # Extract base repository path for matching
+                # Handle both "owner/repo" and "owner/repo/folder" formats
+                base_repo = owner_repo
+                if owner_repo.count('/') > 1:
+                    # This might be a folder path like "owner/repo/folder"
+                    # Extract just the owner/repo part
+                    parts = owner_repo.split('/')
+                    base_repo = f"{parts[0]}/{parts[1]}"
+                
                 # Look for a repository matching this owner/repo
                 matching_repo = None
                 for repo in repos:
-                    # Check if repository field matches
-                    if repo.get("repository") == owner_repo:
+                    repo_path = repo.get("repository", "")
+                    # Check exact match first
+                    if repo_path == owner_repo:
+                        matching_repo = repo
+                        break
+                    # Then check if it's the base repository
+                    elif repo_path == base_repo:
+                        matching_repo = repo
+                        break
+                    # Also check if the stored repo is a folder path that starts with our base
+                    elif repo_path.startswith(base_repo + "/"):
                         matching_repo = repo
                         break
                 

{nia_mcp_server-1.0.14 → nia_mcp_server-1.0.16}/src/nia_mcp_server/assets/rules/nia_rules.md

@@ -29,25 +29,60 @@ Note: Nia is just "Nia" - not an acronym. It's the name of the knowledge search
 - Structured output needs (tables, lists)
 - Questions with "best", "which is better", "compare"
 
-### 3. Query Optimization
+### 3. Repository Identifier Formats
+
+#### Understanding Repository Paths
+When using `search_codebase`, repositories can be specified in different formats:
+
+1. **Full Repository Format**: `owner/repo`
+   - Example: `facebook/react`
+   - Use this when the entire repository was indexed
+
+2. **Folder-Specific Format**: `owner/repo/tree/branch/folder`
+   - Example: `PostHog/posthog/tree/master/docs`
+   - Use this EXACT format when a specific folder was indexed
+   - This format appears in `list_repositories` output - copy it exactly!
+
+#### Important Rules:
+- **Always check `list_repositories` first** to see the exact format
+- **Copy the repository identifier exactly** as shown in the list
+- **Don't modify folder paths** - if it shows `owner/repo/tree/branch/folder`, use that exact string
+- **Don't assume** - a repository indexed as a folder won't work with just `owner/repo`
+
+#### Examples:
+```python
+# Wrong - trying to use base repo when folder was indexed
+search_codebase("What is Flox?", ["PostHog/posthog"])  # ❌ Won't find folder-indexed content
+
+# Right - using exact format from list_repositories
+search_codebase("What is Flox?", ["PostHog/posthog/tree/master/docs"])  # ✅ Searches the indexed folder
+
+# Wrong - modifying the path
+search_codebase("LLM guide", ["mcp-use/mcp-use/docs"])  # ❌ Missing /tree/main/ part
+
+# Right - exact format
+search_codebase("LLM guide", ["mcp-use/mcp-use/tree/main/docs"])  # ✅ Correct format
+```
+
+### 4. Query Optimization
 - **Use natural language queries** - Form complete questions, not just keywords
 - **Be specific and detailed** - "How does authentication work in NextAuth.js?" not "auth nextauth"
 - **Include context** - Mention specific technologies, frameworks, or use cases
 - **Leverage repository context** - Specify repositories when searching indexed codebases
 
-### 4. API Usage Best Practices
+### 5. API Usage Best Practices
 - **Handle rate limits gracefully** - Free tier has 3 indexing operations limit
 - **Cache results mentally** - Avoid redundant searches in the same conversation
 - **Batch operations** - Index multiple related repositories together
 - **Monitor status efficiently** - Check status periodically, not continuously
 
-### 5. Result Interpretation
+### 6. Result Interpretation
 - **Provide actionable next steps** - Always suggest how to use the results
 - **Extract indexable content** - Identify repositories and docs from search results
 - **Format for readability** - Use markdown formatting for clear presentation
 - **Include sources** - Always show where information comes from
 
-### 6. Error Handling
+### 7. Error Handling
 - **Explain API limits clearly** - Help users understand free tier limitations
 - **Suggest alternatives** - Provide workarounds when hitting limits
 - **Report issues helpfully** - Include enough context for debugging
@@ -63,8 +98,12 @@ Note: Nia is just "Nia" - not an acronym. It's the name of the knowledge search
 
 ### Pattern 2: Codebase Understanding
 1. Check if repository is already indexed with `list_repositories`
+   - Note the EXACT repository format shown (especially for folder paths)
+   - Example output: "PostHog/posthog/tree/master/docs" (not just "PostHog/posthog")
 2. If not indexed, use `index_repository` and wait for completion
 3. Use `search_codebase` with specific technical questions
+   - Use the EXACT repository identifier from step 1
+   - Don't modify or simplify folder paths
 4. Include code snippets and file references in responses
 
 ### Pattern 3: Documentation Search

{nia_mcp_server-1.0.14 → nia_mcp_server-1.0.16}/src/nia_mcp_server/server.py

@@ -83,7 +83,7 @@ async def index_repository(
     Index a GitHub repository for intelligent code search.
     
     Args:
-        repo_url: GitHub repository URL (e.g., https://github.com/owner/repo)
+        repo_url: GitHub repository URL (e.g., https://github.com/owner/repo or https://github.com/owner/repo/tree/branch)
         branch: Branch to index (optional, defaults to main branch)
         
     Returns:
@@ -158,11 +158,30 @@ async def search_codebase(
     
     Args:
         query: Natural language search query. Don't just use keywords or unstrctured query, make a comprehensive question to get the best results possible.
-        repositories: List of repositories to search (owner/repo format). If not specified, searches all indexed repos.
+        repositories: List of repositories to search (owner/repo or owner/repo/tree/branch if indexed differently before).
+            - "owner/repo" - Search entire repository (e.g., "facebook/react")
+            - "owner/repo/tree/branch/folder" - Search specific folder indexed separately
+              (e.g., "PostHog/posthog/tree/master/docs")
+            Use the EXACT format shown in list_repositories output for folder-indexed repos.
+            If not specified, searches all indexed repos.
         include_sources: Whether to include source code in results
         
     Returns:
         Search results with relevant code snippets and explanations
+        
+    Examples:
+        # Search all indexed repositories
+        search_codebase("How does authentication work?")
+        
+        # Search specific repository
+        search_codebase("How to create custom hooks?", ["facebook/react"])
+        
+        # Search folder-indexed repository (use exact format from list_repositories)
+        search_codebase("What is Flox?", ["PostHog/posthog/tree/master/docs"])
+
+    Important:
+        - If you want to search a specific folder, use the EXACT repository path shown above
+        - Example: `search_codebase(\"query\", [\"owner/repo/tree/branch/folder\"])`
     """
     try:
         client = await ensure_api_client()
@@ -451,6 +470,9 @@ async def list_repositories() -> List[TextContent]:
         # Format repository list
         lines = ["# Indexed Repositories\n"]
         
+        # Check if any repositories have folder paths (contain /tree/)
+        has_folder_repos = any('/tree/' in repo.get('repository', '') for repo in repositories)
+        
         for repo in repositories:
             status_icon = "✅" if repo.get("status") == "completed" else "⏳"
             
@@ -470,6 +492,24 @@ async def list_repositories() -> List[TextContent]:
                 lines.append(f"- **Indexed:** {repo['indexed_at']}")
             if repo.get("error"):
                 lines.append(f"- **Error:** {repo['error']}")
+            
+            # Add usage hint for completed repositories
+            if repo.get("status") == "completed":
+                lines.append(f"- **Usage:** `search_codebase(query, [\"{repo_name}\"])`")
+        
+        # Add general usage instructions at the end
+        lines.extend([
+            "\n---",
+            "\n## Usage Tips",
+            "- To search all repositories: `search_codebase(\"your query\")`",
+            "- To search specific repository: `search_codebase(\"your query\", [\"owner/repo\"])`"
+        ])
+        
+        if has_folder_repos:
+            lines.extend([
+                "- For folder-indexed repositories: Use the EXACT repository path shown above",
+                "  Example: `search_codebase(\"query\", [\"owner/repo/tree/branch/folder\"])`"
+            ])
         
         return [TextContent(type="text", text="\n".join(lines))]
         
@@ -603,7 +643,6 @@ async def index_documentation(
         - When started indexing, prompt users to either use check_documentation_status tool or go to app.trynia.ai to check the status.
         - By default, crawls the entire domain (up to 10,000 pages)
         - Use exclude_patterns to filter out unwanted sections like blogs, changelogs, etc.
-        - Screenshots are captured by default to provide visual context
     """
     try:
         client = await ensure_api_client()