agent-coderag 1.1.0__tar.gz → 1.2.1__tar.gz

agent_coderag-1.2.1/PKG-INFO

@@ -0,0 +1,207 @@
+Metadata-Version: 2.4
+Name: agent-coderag
+Version: 1.2.1
+Summary: Lightweight semantic code search and distillation utility for AI coding agents. It solves the API knowledge gap via real-time local signature extraction and intent analysis without PyTorch. Optimized for token efficiency, it compresses codebase context into compact semantic summaries stored in a local DuckDB vector similarity index.
+Author-email: Igor Boloban <naranor@gmail.com>
+License: MIT
+Project-URL: Homepage, https://github.com/naranor/agent-coderag
+Project-URL: Repository, https://github.com/naranor/agent-coderag
+Project-URL: Issues, https://github.com/naranor/agent-coderag/issues
+Project-URL: Changelog, https://github.com/naranor/agent-coderag/blob/main/CHANGELOG.md
+Keywords: rag,ai-agents,semantic-search,code-analysis,context-compression,onnx,local-embeddings
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Classifier: Topic :: Text Processing :: Indexing
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+License-File: NOTICE
+Requires-Dist: duckdb
+Requires-Dist: numpy
+Requires-Dist: litellm
+Requires-Dist: onnxruntime
+Requires-Dist: tokenizers
+Requires-Dist: pydantic
+Requires-Dist: httpx
+Requires-Dist: aiofiles
+Requires-Dist: javalang
+Requires-Dist: pathspec
+Dynamic: license-file
+
+# Agent-CodeRAG
+
+<p align="center">
+  <img src="https://raw.githubusercontent.com/naranor/agent-coderag/main/assets/banner.svg" alt="Agent-CodeRAG Banner" width="400">
+</p>
+
+> **Fast. Local. Agent-First. Token-Efficient.** Bridge the gap between AI coding agents and your local environment.
+
+[![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
+[![PyPI Version](https://img.shields.io/pypi/v/agent-coderag.svg)](https://pypi.org/project/agent-coderag/)
+[![Downloads](https://img.shields.io/pypi/dm/agent-coderag.svg)](https://pypi.org/project/agent-coderag/)
+[![Tests](https://github.com/naranor/agent-coderag/actions/workflows/ci.yml/badge.svg)](https://github.com/naranor/agent-coderag/actions/workflows/ci.yml)
+[![Coverage](https://img.shields.io/badge/coverage-81%25-green)](https://github.com/naranor/agent-coderag/actions)
+
+---
+
+## Table of Contents
+
+- [Why Agent-CodeRAG?](#why-agent-coderag)
+- [Quick Start](#quick-start)
+- [Features](#-features)
+- [How It Works](#-how-it-works)
+- [API Discovery](#-api-discovery)
+- [For AI Agents](#-for-ai-agents)
+- [Development](#-development)
+- [Contributing](#-contributing)
+- [License](#-license)
+
+---
+
+## Why Agent-CodeRAG?
+
+AI coding agents often **hallucinate** when calling library APIs because their training data is static. This leads to a **"Fail-Fix-Fail" cycle** — broken code, token waste, and frustration.
+
+**The Problem:** Your agent knows Pydantic v1 (`model.dict()`), but your environment uses v2 (`model.model_dump()`). Result: 5000+ wasted tokens trying to "fix" something it doesn't understand.
+
+**The Solution:** Agent-CodeRAG extracts *actual* API signatures from your installed libraries and provides the LLM with real-time, environment-specific context — saving up to **80% of context window tokens**.
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## Quick Start
+
+```bash
+# 1. Install
+pip install agent-coderag
+
+# 2. Setup (download ONNX model)
+agent-coderag setup
+
+# 3. Configure LLM (optional, for AI distillation)
+agent-coderag config --url "http://localhost:11434" --provider "ollama" --model "qwen2.5-coder:7b"
+
+# 4. Index your project
+agent-coderag sync --all
+
+# 5. Search
+agent-coderag search "how to handle errors"
+```
+
+**Docker:**
+```bash
+docker build -t agent-coderag .
+docker run -v ~/.cache/agent-coderag:/root/.cache/agent-coderag agent-coderag setup
+```
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## ✨ Features
+
+- **⚡ No PyTorch** — Uses `onnxruntime` and `tokenizers` (Rust) for instant startup
+- **💾 DuckDB VSS** — High-performance vector search in a single local file
+- **🏗️ Multi-Language** — Native indexing for Python (AST) and Java (javalang)
+- **🔄 Delta-Sync** — SHA-256 hashing re-distills only changed code
+- **🔌 Hybrid Intelligence** — Works offline; adds AI-distilled reasoning when LLM is connected
+- **📡 API Discovery** — Extract live API signatures from Python modules or Java JARs
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🛠 How It Works
+
+```mermaid
+graph TD
+    A[Local Python Code] --> B[AST Parser]
+    B --> C{Delta-Sync}
+    C -- Changed/New --> D[LLM Distiller]
+    C -- Unchanged --> E[Local Cache]
+    D --> F[Semantic Summary]
+    E --> F
+    F --> G[ONNX Embedder]
+    G --> H[(DuckDB VSS)]
+    H --> I[Semantic Search / JSON API]
+```
+
+1. **AST Parser** — Parses your Python code
+2. **Delta-Sync** — Uses SHA-256 to detect changes
+3. **LLM Distiller** — Generates semantic summaries (optional)
+4. **ONNX Embedder** — Creates embeddings locally
+5. **DuckDB VSS** — Stores vectors for fast similarity search
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 📡 API Discovery
+
+```bash
+agent-coderag api pydantic
+```
+
+Returns the *live* public API, methods, and signatures for any installed library.
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🤖 For AI Agents
+
+Agent-CodeRAG is built for programmatic consumption:
+
+1. **Search First**: `agent-coderag --json search "topic" --limit 1`
+2. **Use Intent**: The `summary` field provides technical intent — skip reading unnecessary files
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🔧 Development
+
+```bash
+# Run tests
+pytest tests/
+pytest e2e_tests/
+
+# Setup pre-commit hooks
+pip install pre-commit
+pre-commit install
+```
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🤝 Contributing
+
+Contributions are welcome! Here's how to get started:
+
+1. **Fork** the repository
+2. **Clone** your fork: `git clone https://github.com/YOUR_USERNAME/agent-coderag.git`
+3. **Create a branch**: `git checkout -b feature/your-feature`
+4. **Make changes** and commit with [Conventional Commits](https://www.conventionalcommits.org/)
+5. **Run tests**: `pytest tests/`
+6. **Push** to your fork and create a **Pull Request**
+
+See [CONTRIBUTING.md](CONTRIBUTING.md) for detailed guidelines.
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 📄 License
+
+MIT © 2026 [Igor Boloban](https://github.com/naranor)
+
+[🔝 Back to top](#table-of-contents)

agent_coderag-1.2.1/README.md

@@ -0,0 +1,171 @@
+# Agent-CodeRAG
+
+<p align="center">
+  <img src="https://raw.githubusercontent.com/naranor/agent-coderag/main/assets/banner.svg" alt="Agent-CodeRAG Banner" width="400">
+</p>
+
+> **Fast. Local. Agent-First. Token-Efficient.** Bridge the gap between AI coding agents and your local environment.
+
+[![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
+[![PyPI Version](https://img.shields.io/pypi/v/agent-coderag.svg)](https://pypi.org/project/agent-coderag/)
+[![Downloads](https://img.shields.io/pypi/dm/agent-coderag.svg)](https://pypi.org/project/agent-coderag/)
+[![Tests](https://github.com/naranor/agent-coderag/actions/workflows/ci.yml/badge.svg)](https://github.com/naranor/agent-coderag/actions/workflows/ci.yml)
+[![Coverage](https://img.shields.io/badge/coverage-81%25-green)](https://github.com/naranor/agent-coderag/actions)
+
+---
+
+## Table of Contents
+
+- [Why Agent-CodeRAG?](#why-agent-coderag)
+- [Quick Start](#quick-start)
+- [Features](#-features)
+- [How It Works](#-how-it-works)
+- [API Discovery](#-api-discovery)
+- [For AI Agents](#-for-ai-agents)
+- [Development](#-development)
+- [Contributing](#-contributing)
+- [License](#-license)
+
+---
+
+## Why Agent-CodeRAG?
+
+AI coding agents often **hallucinate** when calling library APIs because their training data is static. This leads to a **"Fail-Fix-Fail" cycle** — broken code, token waste, and frustration.
+
+**The Problem:** Your agent knows Pydantic v1 (`model.dict()`), but your environment uses v2 (`model.model_dump()`). Result: 5000+ wasted tokens trying to "fix" something it doesn't understand.
+
+**The Solution:** Agent-CodeRAG extracts *actual* API signatures from your installed libraries and provides the LLM with real-time, environment-specific context — saving up to **80% of context window tokens**.
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## Quick Start
+
+```bash
+# 1. Install
+pip install agent-coderag
+
+# 2. Setup (download ONNX model)
+agent-coderag setup
+
+# 3. Configure LLM (optional, for AI distillation)
+agent-coderag config --url "http://localhost:11434" --provider "ollama" --model "qwen2.5-coder:7b"
+
+# 4. Index your project
+agent-coderag sync --all
+
+# 5. Search
+agent-coderag search "how to handle errors"
+```
+
+**Docker:**
+```bash
+docker build -t agent-coderag .
+docker run -v ~/.cache/agent-coderag:/root/.cache/agent-coderag agent-coderag setup
+```
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## ✨ Features
+
+- **⚡ No PyTorch** — Uses `onnxruntime` and `tokenizers` (Rust) for instant startup
+- **💾 DuckDB VSS** — High-performance vector search in a single local file
+- **🏗️ Multi-Language** — Native indexing for Python (AST) and Java (javalang)
+- **🔄 Delta-Sync** — SHA-256 hashing re-distills only changed code
+- **🔌 Hybrid Intelligence** — Works offline; adds AI-distilled reasoning when LLM is connected
+- **📡 API Discovery** — Extract live API signatures from Python modules or Java JARs
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🛠 How It Works
+
+```mermaid
+graph TD
+    A[Local Python Code] --> B[AST Parser]
+    B --> C{Delta-Sync}
+    C -- Changed/New --> D[LLM Distiller]
+    C -- Unchanged --> E[Local Cache]
+    D --> F[Semantic Summary]
+    E --> F
+    F --> G[ONNX Embedder]
+    G --> H[(DuckDB VSS)]
+    H --> I[Semantic Search / JSON API]
+```
+
+1. **AST Parser** — Parses your Python code
+2. **Delta-Sync** — Uses SHA-256 to detect changes
+3. **LLM Distiller** — Generates semantic summaries (optional)
+4. **ONNX Embedder** — Creates embeddings locally
+5. **DuckDB VSS** — Stores vectors for fast similarity search
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 📡 API Discovery
+
+```bash
+agent-coderag api pydantic
+```
+
+Returns the *live* public API, methods, and signatures for any installed library.
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🤖 For AI Agents
+
+Agent-CodeRAG is built for programmatic consumption:
+
+1. **Search First**: `agent-coderag --json search "topic" --limit 1`
+2. **Use Intent**: The `summary` field provides technical intent — skip reading unnecessary files
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🔧 Development
+
+```bash
+# Run tests
+pytest tests/
+pytest e2e_tests/
+
+# Setup pre-commit hooks
+pip install pre-commit
+pre-commit install
+```
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🤝 Contributing
+
+Contributions are welcome! Here's how to get started:
+
+1. **Fork** the repository
+2. **Clone** your fork: `git clone https://github.com/YOUR_USERNAME/agent-coderag.git`
+3. **Create a branch**: `git checkout -b feature/your-feature`
+4. **Make changes** and commit with [Conventional Commits](https://www.conventionalcommits.org/)
+5. **Run tests**: `pytest tests/`
+6. **Push** to your fork and create a **Pull Request**
+
+See [CONTRIBUTING.md](CONTRIBUTING.md) for detailed guidelines.
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 📄 License
+
+MIT © 2026 [Igor Boloban](https://github.com/naranor)
+
+[🔝 Back to top](#table-of-contents)

agent_coderag-1.2.1/agent_coderag.egg-info/PKG-INFO

@@ -0,0 +1,207 @@
+Metadata-Version: 2.4
+Name: agent-coderag
+Version: 1.2.1
+Summary: Lightweight semantic code search and distillation utility for AI coding agents. It solves the API knowledge gap via real-time local signature extraction and intent analysis without PyTorch. Optimized for token efficiency, it compresses codebase context into compact semantic summaries stored in a local DuckDB vector similarity index.
+Author-email: Igor Boloban <naranor@gmail.com>
+License: MIT
+Project-URL: Homepage, https://github.com/naranor/agent-coderag
+Project-URL: Repository, https://github.com/naranor/agent-coderag
+Project-URL: Issues, https://github.com/naranor/agent-coderag/issues
+Project-URL: Changelog, https://github.com/naranor/agent-coderag/blob/main/CHANGELOG.md
+Keywords: rag,ai-agents,semantic-search,code-analysis,context-compression,onnx,local-embeddings
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Classifier: Topic :: Text Processing :: Indexing
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+License-File: NOTICE
+Requires-Dist: duckdb
+Requires-Dist: numpy
+Requires-Dist: litellm
+Requires-Dist: onnxruntime
+Requires-Dist: tokenizers
+Requires-Dist: pydantic
+Requires-Dist: httpx
+Requires-Dist: aiofiles
+Requires-Dist: javalang
+Requires-Dist: pathspec
+Dynamic: license-file
+
+# Agent-CodeRAG
+
+<p align="center">
+  <img src="https://raw.githubusercontent.com/naranor/agent-coderag/main/assets/banner.svg" alt="Agent-CodeRAG Banner" width="400">
+</p>
+
+> **Fast. Local. Agent-First. Token-Efficient.** Bridge the gap between AI coding agents and your local environment.
+
+[![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
+[![PyPI Version](https://img.shields.io/pypi/v/agent-coderag.svg)](https://pypi.org/project/agent-coderag/)
+[![Downloads](https://img.shields.io/pypi/dm/agent-coderag.svg)](https://pypi.org/project/agent-coderag/)
+[![Tests](https://github.com/naranor/agent-coderag/actions/workflows/ci.yml/badge.svg)](https://github.com/naranor/agent-coderag/actions/workflows/ci.yml)
+[![Coverage](https://img.shields.io/badge/coverage-81%25-green)](https://github.com/naranor/agent-coderag/actions)
+
+---
+
+## Table of Contents
+
+- [Why Agent-CodeRAG?](#why-agent-coderag)
+- [Quick Start](#quick-start)
+- [Features](#-features)
+- [How It Works](#-how-it-works)
+- [API Discovery](#-api-discovery)
+- [For AI Agents](#-for-ai-agents)
+- [Development](#-development)
+- [Contributing](#-contributing)
+- [License](#-license)
+
+---
+
+## Why Agent-CodeRAG?
+
+AI coding agents often **hallucinate** when calling library APIs because their training data is static. This leads to a **"Fail-Fix-Fail" cycle** — broken code, token waste, and frustration.
+
+**The Problem:** Your agent knows Pydantic v1 (`model.dict()`), but your environment uses v2 (`model.model_dump()`). Result: 5000+ wasted tokens trying to "fix" something it doesn't understand.
+
+**The Solution:** Agent-CodeRAG extracts *actual* API signatures from your installed libraries and provides the LLM with real-time, environment-specific context — saving up to **80% of context window tokens**.
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## Quick Start
+
+```bash
+# 1. Install
+pip install agent-coderag
+
+# 2. Setup (download ONNX model)
+agent-coderag setup
+
+# 3. Configure LLM (optional, for AI distillation)
+agent-coderag config --url "http://localhost:11434" --provider "ollama" --model "qwen2.5-coder:7b"
+
+# 4. Index your project
+agent-coderag sync --all
+
+# 5. Search
+agent-coderag search "how to handle errors"
+```
+
+**Docker:**
+```bash
+docker build -t agent-coderag .
+docker run -v ~/.cache/agent-coderag:/root/.cache/agent-coderag agent-coderag setup
+```
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## ✨ Features
+
+- **⚡ No PyTorch** — Uses `onnxruntime` and `tokenizers` (Rust) for instant startup
+- **💾 DuckDB VSS** — High-performance vector search in a single local file
+- **🏗️ Multi-Language** — Native indexing for Python (AST) and Java (javalang)
+- **🔄 Delta-Sync** — SHA-256 hashing re-distills only changed code
+- **🔌 Hybrid Intelligence** — Works offline; adds AI-distilled reasoning when LLM is connected
+- **📡 API Discovery** — Extract live API signatures from Python modules or Java JARs
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🛠 How It Works
+
+```mermaid
+graph TD
+    A[Local Python Code] --> B[AST Parser]
+    B --> C{Delta-Sync}
+    C -- Changed/New --> D[LLM Distiller]
+    C -- Unchanged --> E[Local Cache]
+    D --> F[Semantic Summary]
+    E --> F
+    F --> G[ONNX Embedder]
+    G --> H[(DuckDB VSS)]
+    H --> I[Semantic Search / JSON API]
+```
+
+1. **AST Parser** — Parses your Python code
+2. **Delta-Sync** — Uses SHA-256 to detect changes
+3. **LLM Distiller** — Generates semantic summaries (optional)
+4. **ONNX Embedder** — Creates embeddings locally
+5. **DuckDB VSS** — Stores vectors for fast similarity search
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 📡 API Discovery
+
+```bash
+agent-coderag api pydantic
+```
+
+Returns the *live* public API, methods, and signatures for any installed library.
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🤖 For AI Agents
+
+Agent-CodeRAG is built for programmatic consumption:
+
+1. **Search First**: `agent-coderag --json search "topic" --limit 1`
+2. **Use Intent**: The `summary` field provides technical intent — skip reading unnecessary files
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🔧 Development
+
+```bash
+# Run tests
+pytest tests/
+pytest e2e_tests/
+
+# Setup pre-commit hooks
+pip install pre-commit
+pre-commit install
+```
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 🤝 Contributing
+
+Contributions are welcome! Here's how to get started:
+
+1. **Fork** the repository
+2. **Clone** your fork: `git clone https://github.com/YOUR_USERNAME/agent-coderag.git`
+3. **Create a branch**: `git checkout -b feature/your-feature`
+4. **Make changes** and commit with [Conventional Commits](https://www.conventionalcommits.org/)
+5. **Run tests**: `pytest tests/`
+6. **Push** to your fork and create a **Pull Request**
+
+See [CONTRIBUTING.md](CONTRIBUTING.md) for detailed guidelines.
+
+[🔝 Back to top](#table-of-contents)
+
+---
+
+## 📄 License
+
+MIT © 2026 [Igor Boloban](https://github.com/naranor)
+
+[🔝 Back to top](#table-of-contents)

{agent_coderag-1.1.0 → agent_coderag-1.2.1}/agent_coderag.egg-info/SOURCES.txt

@@ -15,6 +15,7 @@ code_rag/core/manager.py
 code_rag/core/models.py
 code_rag/discovery/__init__.py
 code_rag/discovery/dependency.py
+code_rag/discovery/java_discovery.py
 code_rag/entry/__init__.py
 code_rag/entry/cli.py
 code_rag/intelligence/__init__.py
@@ -22,8 +23,17 @@ code_rag/intelligence/distiller.py
 code_rag/intelligence/embedder.py
 code_rag/parsers/__init__.py
 code_rag/parsers/ast_index.py
+code_rag/parsers/java_parser.py
+code_rag/parsers/multi_parser.py
 code_rag/storage/__init__.py
 code_rag/storage/duckdb_impl.py
 tests/test_ast_parser.py
+tests/test_cli.py
 tests/test_code_rag_simple.py
-tests/test_dependency_discovery.py
+tests/test_dependency_discovery.py
+tests/test_distiller.py
+tests/test_embedder.py
+tests/test_interfaces.py
+tests/test_java_parser.py
+tests/test_multi_parser.py
+tests/test_parsers.py

{agent_coderag-1.1.0 → agent_coderag-1.2.1}/agent_coderag.egg-info/requires.txt

@@ -6,3 +6,5 @@ tokenizers
 pydantic
 httpx
 aiofiles
+javalang
+pathspec

{agent_coderag-1.1.0 → agent_coderag-1.2.1}/code_rag/core/interfaces.py

@@ -1,6 +1,6 @@
 from abc import ABC, abstractmethod
 from typing import List, Optional
-from .models import KnowledgeUnit
+from .models import KnowledgeUnit, Relation
 
 class IParser(ABC):
     """Interface for extracting structure from code."""
@@ -22,6 +22,14 @@ class IStorage(ABC):
     async def search_units(self, query: str, limit: int = 5) -> List[KnowledgeUnit]:
         pass
 
+    @abstractmethod
+    async def upsert_relation(self, relation: Relation):
+        pass
+
+    @abstractmethod
+    async def get_relations(self, unit_id: str, direction: str = "out") -> List[Relation]:
+        pass
+
 class IIntelligence(ABC):
     """Interface for LLM-based analysis (distillation, embeddings)."""
     @abstractmethod

{agent_coderag-1.1.0 → agent_coderag-1.2.1}/code_rag/core/manager.py

@@ -11,10 +11,11 @@ class CodeRAGManager:
     Orchestrates the RAG workflow: parsing, distillation, and storage.
     """
     
-    def __init__(self, storage: IStorage, parser: IParser, intelligence: IIntelligence):
+    def __init__(self, storage: IStorage, parser: IParser, intelligence: IIntelligence, max_concurrency: int = 10):
         self.storage = storage
         self.parser = parser
         self.intelligence = intelligence
+        self.semaphore = asyncio.Semaphore(max_concurrency)
 
     async def sync_file(self, file_path: str, force_distill: bool = False):
         """
@@ -42,13 +43,14 @@ class CodeRAGManager:
                 logger.info("Summary missing for %s", unit.name)
             
             if should_distill:
-                logger.info("Distilling summary for %s...", unit.id)
-                try:
-                    unit.summary = await self.intelligence.summarize(raw_code, unit.name)
-                except Exception as e:
-                    logger.error("Failed to distill %s: %s", unit.name, e)
-                    # Keep old summary if available, otherwise stay None
-                    unit.summary = existing_unit.summary if existing_unit else None
+                async with self.semaphore:
+                    logger.info("Distilling summary for %s...", unit.id)
+                    try:
+                        unit.summary = await self.intelligence.summarize(raw_code, unit.name)
+                    except Exception as e:
+                        logger.error("Failed to distill %s: %s", unit.name, e)
+                        # Keep old summary if available, otherwise stay None
+                        unit.summary = existing_unit.summary if existing_unit else None
             else:
                 # Reuse existing summary if code hasn't changed
                 unit.summary = existing_unit.summary if existing_unit else None