fix(core): skip empty/invalid chunks before embedding; guard OpenAI embeddings

Avoid 400 errors from OpenAI when chunker yields empty strings by filtering
invalid texts in LeannBuilder.build_index. Add validation fail-fast in
OpenAI embedding path to surface upstream issues earlier. Keeps passages and
embeddings aligned during build.

Refs #54

README.md

@@ -31,7 +31,7 @@ LEANN achieves this through *graph-based selective recomputation* with *high-deg
   <img src="assets/effects.png" alt="LEANN vs Traditional Vector DB Storage Comparison" width="70%">
 </p>
 
-> **The numbers speak for themselves:** Index 60 million text chunks in just 6GB instead of 201GB. From emails to browser history, everything fits on your laptop. [See detailed benchmarks for different applications below ↓](#storage-comparison)
+> **The numbers speak for themselves:** Index 60 million text chunks in just 6GB instead of 201GB. From emails to browser history, everything fits on your laptop. [See detailed benchmarks for different applications below ↓](#-storage-comparison)
 
 
 🔒 **Privacy:** Your data never leaves your laptop. No OpenAI, no cloud, no "terms of service".
@@ -70,8 +70,8 @@ uv venv
 source .venv/bin/activate
 uv pip install leann
 ```
-
+<!--
-> Low-resource? See “Low-resource setups” in the [Configuration Guide](docs/configuration-guide.md#low-resource-setups).
+> Low-resource? See “Low-resource setups” in the [Configuration Guide](docs/configuration-guide.md#low-resource-setups). -->
 
 <details>
 <summary>
@@ -426,21 +426,21 @@ Once the index is built, you can ask questions like:
 **The future of code assistance is here.** Transform your development workflow with LEANN's native MCP integration for Claude Code. Index your entire codebase and get intelligent code assistance directly in your IDE.
 
 **Key features:**
-- 🔍 **Semantic code search** across your entire project
+- 🔍 **Semantic code search** across your entire project, fully local index and lightweight
 - 📚 **Context-aware assistance** for debugging and development
 - 🚀 **Zero-config setup** with automatic language detection
 
 ```bash
 # Install LEANN globally for MCP integration
-uv tool install leann-core
+uv tool install leann-core --with leann
-
+claude mcp add --scope user leann-server -- leann_mcp
 # Setup is automatic - just start using Claude Code!
 ```
 Try our fully agentic pipeline with auto query rewriting, semantic search planning, and more:
 
 ![LEANN MCP Integration](assets/mcp_leann.png)
 
-**Ready to supercharge your coding?** [Complete Setup Guide →](packages/leann-mcp/README.md)
+**🔥 Ready to supercharge your coding?** [Complete Setup Guide →](packages/leann-mcp/README.md)
 
 ## 🖥️ Command Line Interface
 
@@ -457,7 +457,8 @@ leann --help
 **To make it globally available:**
 ```bash
 # Install the LEANN CLI globally using uv tool
-uv tool install leann-core
+uv tool install leann-core --with leann
+
 
 # Now you can use leann from anywhere without activating venv
 leann --help

packages/leann-core/src/leann/api.py

@@ -306,6 +306,23 @@ class LeannBuilder:
     def build_index(self, index_path: str):
         if not self.chunks:
             raise ValueError("No chunks added.")
+
+        # Filter out invalid/empty text chunks early to keep passage and embedding counts aligned
+        valid_chunks: list[dict[str, Any]] = []
+        skipped = 0
+        for chunk in self.chunks:
+            text = chunk.get("text", "")
+            if isinstance(text, str) and text.strip():
+                valid_chunks.append(chunk)
+            else:
+                skipped += 1
+        if skipped > 0:
+            print(
+                f"Warning: Skipping {skipped} empty/invalid text chunk(s). Processing {len(valid_chunks)} valid chunks"
+            )
+            self.chunks = valid_chunks
+            if not self.chunks:
+                raise ValueError("All provided chunks are empty or invalid. Nothing to index.")
         if self.dimensions is None:
             self.dimensions = len(
                 compute_embeddings(

packages/leann-core/src/leann/cli.py

@@ -148,6 +148,30 @@ Examples:
             type=str,
             help="Comma-separated list of file extensions to include (e.g., '.txt,.pdf,.pptx'). If not specified, uses default supported types.",
         )
+        build_parser.add_argument(
+            "--doc-chunk-size",
+            type=int,
+            default=256,
+            help="Document chunk size in tokens/characters (default: 256)",
+        )
+        build_parser.add_argument(
+            "--doc-chunk-overlap",
+            type=int,
+            default=128,
+            help="Document chunk overlap (default: 128)",
+        )
+        build_parser.add_argument(
+            "--code-chunk-size",
+            type=int,
+            default=512,
+            help="Code chunk size in tokens/lines (default: 512)",
+        )
+        build_parser.add_argument(
+            "--code-chunk-overlap",
+            type=int,
+            default=50,
+            help="Code chunk overlap (default: 50)",
+        )
 
         # Search command
         search_parser = subparsers.add_parser("search", help="Search documents")
@@ -726,6 +750,37 @@ Examples:
             print(f"Index '{index_name}' already exists. Use --force to rebuild.")
             return
 
+        # Configure chunking based on CLI args before loading documents
+        # Guard against invalid configurations
+        doc_chunk_size = max(1, int(args.doc_chunk_size))
+        doc_chunk_overlap = max(0, int(args.doc_chunk_overlap))
+        if doc_chunk_overlap >= doc_chunk_size:
+            print(
+                f"⚠️  Adjusting doc chunk overlap from {doc_chunk_overlap} to {doc_chunk_size - 1} (must be < chunk size)"
+            )
+            doc_chunk_overlap = doc_chunk_size - 1
+
+        code_chunk_size = max(1, int(args.code_chunk_size))
+        code_chunk_overlap = max(0, int(args.code_chunk_overlap))
+        if code_chunk_overlap >= code_chunk_size:
+            print(
+                f"⚠️  Adjusting code chunk overlap from {code_chunk_overlap} to {code_chunk_size - 1} (must be < chunk size)"
+            )
+            code_chunk_overlap = code_chunk_size - 1
+
+        self.node_parser = SentenceSplitter(
+            chunk_size=doc_chunk_size,
+            chunk_overlap=doc_chunk_overlap,
+            separator=" ",
+            paragraph_separator="\n\n",
+        )
+        self.code_parser = SentenceSplitter(
+            chunk_size=code_chunk_size,
+            chunk_overlap=code_chunk_overlap,
+            separator="\n",
+            paragraph_separator="\n\n",
+        )
+
         all_texts = self.load_documents(docs_paths, args.file_types)
         if not all_texts:
             print("No documents found")

packages/leann-core/src/leann/embedding_compute.py

@@ -244,6 +244,16 @@ def compute_embeddings_openai(texts: list[str], model_name: str) -> np.ndarray:
     except ImportError as e:
         raise ImportError(f"OpenAI package not installed: {e}")
 
+    # Validate input list
+    if not texts:
+        raise ValueError("Cannot compute embeddings for empty text list")
+    # Extra validation: abort early if any item is empty/whitespace
+    invalid_count = sum(1 for t in texts if not isinstance(t, str) or not t.strip())
+    if invalid_count > 0:
+        raise ValueError(
+            f"Found {invalid_count} empty/invalid text(s) in input. Upstream should filter before calling OpenAI."
+        )
+
     api_key = os.getenv("OPENAI_API_KEY")
     if not api_key:
         raise RuntimeError("OPENAI_API_KEY environment variable not set")
@@ -263,8 +273,16 @@ def compute_embeddings_openai(texts: list[str], model_name: str) -> np.ndarray:
     print(f"len of texts: {len(texts)}")
 
     # OpenAI has limits on batch size and input length
-    max_batch_size = 1000  # Conservative batch size
+    max_batch_size = 800  # Conservative batch size because the token limit is 300K
     all_embeddings = []
+    # get the avg len of texts
+    avg_len = sum(len(text) for text in texts) / len(texts)
+    print(f"avg len of texts: {avg_len}")
+    # if avg len is less than 1000, use the max batch size
+    if avg_len > 300:
+        max_batch_size = 500
+
+    # if avg len is less than 1000, use the max batch size
 
     try:
         from tqdm import tqdm

packages/leann-core/src/leann/mcp.py

@@ -64,19 +64,6 @@ def handle_request(request):
                             "required": ["index_name", "query"],
                         },
                     },
-                    {
-                        "name": "leann_status",
-                        "description": "📊 Check the health and stats of your code indexes - like a medical checkup for your codebase knowledge!",
-                        "inputSchema": {
-                            "type": "object",
-                            "properties": {
-                                "index_name": {
-                                    "type": "string",
-                                    "description": "Optional: Name of specific index to check. If not provided, shows status of all indexes.",
-                                }
-                            },
-                        },
-                    },
                     {
                         "name": "leann_list",
                         "description": "📋 Show all your indexed codebases - your personal code library! Use this to see what's available for search.",
@@ -118,15 +105,6 @@ def handle_request(request):
                 ]
                 result = subprocess.run(cmd, capture_output=True, text=True)
 
-            elif tool_name == "leann_status":
-                if args.get("index_name"):
-                    # Check specific index status - for now, we'll use leann list and filter
-                    result = subprocess.run(["leann", "list"], capture_output=True, text=True)
-                    # We could enhance this to show more detailed status per index
-                else:
-                    # Show all indexes status
-                    result = subprocess.run(["leann", "list"], capture_output=True, text=True)
-
             elif tool_name == "leann_list":
                 result = subprocess.run(["leann", "list"], capture_output=True, text=True)
 

packages/leann-mcp/README.md

@@ -13,10 +13,20 @@ This installs the `leann` CLI into an isolated tool environment and includes bot
 
 ## 🚀 Quick Setup
 
-Add the LEANN MCP server to Claude Code:
+Add the LEANN MCP server to Claude Code. Choose the scope based on how widely you want it available. Below is the command to install it globally; if you prefer a local install, skip this step:
 
 ```bash
-claude mcp add leann-server -- leann_mcp
+# Global (recommended): available in all projects for your user
+claude mcp add --scope user leann-server -- leann_mcp
+```
+
+- `leann-server`: the display name of the MCP server in Claude Code (you can change it).
+- `leann_mcp`: the Python entry point installed with LEANN that starts the MCP server.
+
+Verify it is registered globally:
+
+```bash
+claude mcp list | cat
 ```
 
 ## 🛠️ Available Tools
@@ -25,27 +35,36 @@ Once connected, you'll have access to these powerful semantic search tools in Cl
 
 - **`leann_list`** - List all available indexes across your projects
 - **`leann_search`** - Perform semantic searches across code and documents
-- **`leann_ask`** - Ask natural language questions and get AI-powered answers from your codebase
+
 
 ## 🎯 Quick Start Example
 
 ```bash
+# Add locally if you did not add it globally (current folder only; default if --scope is omitted)
+claude mcp add leann-server -- leann_mcp
+
 # Build an index for your project (change to your actual path)
-leann build my-project --docs ./
+# See the advanced examples below for more ways to configure indexing
+# Set the index name (replace 'my-project' with your own)
+leann build my-project --docs $(git ls-files)
 
 # Start Claude Code
 claude
 ```
 
-## 🚀 Advanced Usage Examples
+## 🚀 Advanced Usage Examples to build the index
 
 ### Index Entire Git Repository
 ```bash
-# Index all tracked files in your git repository, note right now we will skip submodules, but we can add it back easily if you want
+# Index all tracked files in your Git repository.
+# Note: submodules are currently skipped; we can add them back if needed.
 leann build my-repo --docs $(git ls-files) --embedding-mode sentence-transformers --embedding-model all-MiniLM-L6-v2 --backend hnsw
 
-# Index only specific file types from git
+# Index only tracked Python files from Git.
 leann build my-python-code --docs $(git ls-files "*.py") --embedding-mode sentence-transformers --embedding-model all-MiniLM-L6-v2 --backend hnsw
+
+# If you encounter empty requests caused by empty files (e.g., __init__.py), exclude zero-byte files. Thanks @ww2283 for pointing [that](https://github.com/yichuan-w/LEANN/issues/48) out
+leann build leann-prospec-lig --docs $(find ./src -name "*.py" -not -empty) --embedding-mode openai --embedding-model text-embedding-3-small
 ```
 
 ### Multiple Directories and Files
@@ -73,7 +92,7 @@ leann build docs-and-configs --docs $(git ls-files "*.md" "*.yml" "*.yaml" "*.js
 ```
 
 
-**Try this in Claude Code:**
+## **Try this in Claude Code:**
 ```
 Help me understand this codebase. List available indexes and search for authentication patterns.
 ```
@@ -82,6 +101,7 @@ Help me understand this codebase. List available indexes and search for authenti
   <img src="../../assets/claude_code_leann.png" alt="LEANN in Claude Code" width="80%">
 </p>
 
+If you see a prompt asking whether to proceed with LEANN, you can now use it in your chat!
 
 ## 🧠 How It Works
 
@@ -117,3 +137,11 @@ To remove LEANN
 ```
 uv pip uninstall leann leann-backend-hnsw leann-core
 ```
+
+To globally remove LEANN (for version update)
+```
+uv tool list | cat
+uv tool uninstall leann-core
+command -v leann || echo "leann gone"
+command -v leann_mcp || echo "leann_mcp gone"
+```

packages/wechat-exporter/__init__.py

@@ -0,0 +1 @@
+__all__ = []

packages/wechat-exporter/main.py

@@ -136,5 +136,9 @@ def export_sqlite(
     connection.commit()
 
 
-if __name__ == "__main__":
+def main():
     app()
+
+
+if __name__ == "__main__":
+    main()

pyproject.toml

@@ -10,6 +10,7 @@ requires-python = ">=3.9"
 dependencies = [
     "leann-core",
     "leann-backend-hnsw",
+    "typer>=0.12.3",
     "numpy>=1.26.0",
     "torch",
     "tqdm",
@@ -84,6 +85,11 @@ documents = [
 
 [tool.setuptools]
 py-modules = []
+packages = ["wechat_exporter"]
+package-dir = { "wechat_exporter" = "packages/wechat-exporter" }
+
+[project.scripts]
+wechat-exporter = "wechat_exporter.main:main"
 
 
 [tool.uv.sources]