Skip to main content

Development Guide

This guide provides comprehensive instructions for developers who want to contribute to PowerMem, including how to set up the development environment, build the project, and add new integrations.

Table of Contents​

Development Environment Setup​

Prerequisites​

  • Python 3.11 or higher
  • pip (Python package manager)
  • Git

Installation​

  1. Clone the repository:
git clone https://github.com/oceanbase/powermem.git
cd powermem
  1. Install development dependencies:
# Install the project in development mode with all dependencies
make install-dev

# Or install with test dependencies
make install-test

Alternatively, you can use pip directly:

pip install -e ".[dev,test]"
  1. Set up environment variables:

Copy the example environment file and configure it:

cp .env.example .env
# Edit .env with your configuration

Development Tools​

The project uses several development tools that are installed with dev dependencies:

  • pytest: Testing framework
  • black: Code formatter
  • isort: Import sorter
  • flake8: Linter
  • mypy: Type checker

Build dashboard​

To build the dashboard, you need to have Node.js and pnpm installed. Then run:

cd dashboard
pnpm install
pnpm build
cd ..
cp -r dashboard/dist/* src/server/dashboard/

Building the Project​

Build Distribution Packages​

To build wheel and source distribution packages:

make build-package

This will:

  • Clean previous build artifacts
  • Build both wheel (.whl) and source distribution (.tar.gz) packages
  • Output files to the dist/ directory

Check Built Package​

Before publishing, verify the package:

make build-check

This runs twine check to validate the package.

Install Build Tools​

If you need to install build tools manually:

make install-build-tools

This installs build and twine packages.

Install Locally​

To install the built package locally:

make install-local

This installs the wheel package from dist/ directory.

Contributing Code​

Workflow​

  1. Fork the repository on GitHub
  2. Create a feature branch:
git checkout -b feature/your-feature-name
  1. Make your changes following the code style guidelines
  2. Write tests for your changes
  3. Run tests to ensure everything passes:
make test
  1. Format and lint your code:
make format
make lint
  1. Commit your changes with clear commit messages
  2. Push to your fork and create a Pull Request

Commit Message Guidelines​

  • Use clear, descriptive commit messages
  • Start with a verb in imperative mood (e.g., "Add", "Fix", "Update")
  • Reference issue numbers when applicable (e.g., "Fix #123: Add error handling")

Pull Request Guidelines​

  • Provide a clear description of changes
  • Include tests for new features
  • Update documentation if needed
  • Ensure all tests pass
  • Follow the code style guidelines

Adding a Vector Store​

To add a new vector store provider, you need to:

  1. Create the vector store implementation
  2. Create the configuration class
  3. Register it in the factory

Step 1: Create Vector Store Implementation​

Create a new file in src/powermem/storage/your_provider/your_provider.py:

"""
YourProvider vector store implementation
"""
import logging
from typing import Any, Dict, List, Optional

from powermem.storage.base import VectorStoreBase, OutputData
from powermem.utils.utils import generate_snowflake_id

logger = logging.getLogger(__name__)


class YourProviderVectorStore(VectorStoreBase):
    """YourProvider vector store implementation"""
    
    def __init__(
        self,
        collection_name: str = "memories",
        # Add your provider-specific parameters
        host: Optional[str] = None,
        port: Optional[int] = None,
        **kwargs
    ):
        """
        Initialize the vector store.
        
        Args:
            collection_name: Name of the collection
            host: Server host
            port: Server port
        """
        self.collection_name = collection_name
        # Initialize your provider client here
        logger.info(f"YourProviderVectorStore initialized")
    
    def create_col(self, name=None, vector_size=None, distance="cosine"):
        """Create a new collection."""
        collection_name = name or self.collection_name
        # Implement collection creation logic
        pass
    
    def insert(self, vectors: List[List[float]], payloads=None, ids=None) -> List[int]:
        """
        Insert vectors into the collection.
        
        Args:
            vectors: List of vectors to insert
            payloads: List of payload dictionaries
            ids: Optional list of IDs (if None, will generate Snowflake IDs)
            
        Returns:
            List[int]: List of generated or provided IDs
        """
        if not vectors:
            return []
        
        if payloads is None:
            payloads = [{} for _ in vectors]
        
        # Generate IDs if not provided
        if ids is None:
            ids = [generate_snowflake_id() for _ in range(len(vectors))]
        
        # Implement insertion logic
        # ...
        
        return ids
    
    def search(
        self, 
        query: List[float], 
        vectors: Optional[List[List[float]]] = None, 
        limit: int = 5, 
        filters: Optional[Dict[str, Any]] = None
    ) -> List[OutputData]:
        """
        Search for similar vectors.
        
        Args:
            query: Query vector
            vectors: Optional list of vectors to search in
            limit: Maximum number of results
            filters: Optional metadata filters
            
        Returns:
            List[OutputData]: List of search results
        """
        # Implement search logic
        results = []
        # ...
        return results
    
    def delete(self, vector_id: int) -> bool:
        """Delete a vector by ID."""
        # Implement deletion logic
        return True
    
    def update(self, vector_id: int, vector=None, payload=None) -> bool:
        """Update a vector and its payload."""
        # Implement update logic
        return True
    
    def get(self, vector_id: int) -> Optional[OutputData]:
        """Retrieve a vector by ID."""
        # Implement retrieval logic
        return None
    
    def list_cols(self) -> List[str]:
        """List all collections."""
        # Implement list logic
        return []
    
    def delete_col(self, name=None) -> bool:
        """Delete a collection."""
        # Implement deletion logic
        return True
    
    def col_info(self) -> Dict[str, Any]:
        """Get information about a collection."""
        # Implement info retrieval
        return {}
    
    def list(self, filters=None, limit=None) -> List[OutputData]:
        """List all memories."""
        # Implement list logic
        return []
    
    def reset(self) -> bool:
        """Reset by deleting the collection and recreating it."""
        self.delete_col()
        self.create_col()
        return True

Step 2: Create Configuration Class​

Create a configuration file in src/powermem/storage/config/your_provider.py:

"""
Configuration for YourProvider vector store
"""
from typing import Optional
from pydantic import Field

from powermem.storage.config.base import BaseVectorStoreConfig


class YourProviderConfig(BaseVectorStoreConfig):
    """Configuration for YourProvider vector store"""
    
    host: str = Field(description="Server host")
    port: int = Field(default=5432, description="Server port")
    # Add other provider-specific fields
    
    class Config:
        extra = "allow"

Update src/powermem/storage/configs.py to include your config:

from powermem.storage.config.your_provider import YourProviderConfig

class VectorStoreConfig(BaseModel):
    # ...
    _provider_configs: Dict[str, str] = {
        "oceanbase": "OceanBaseConfig",
        "pgvector": "PGVectorConfig",
        "sqlite": "SQLiteConfig",
        "your_provider": "YourProviderConfig",  # Add this
    }

Step 3: Register in Factory​

Update src/powermem/storage/factory.py:

class VectorStoreFactory:
    provider_to_class = {
        "oceanbase": "powermem.storage.oceanbase.oceanbase.OceanBaseVectorStore",
        "sqlite": "powermem.storage.sqlite.sqlite_vector_store.SQLiteVectorStore",
        "pgvector": "powermem.storage.pgvector.pgvector.PGVectorStore",
        "postgres": "powermem.storage.pgvector.pgvector.PGVectorStore",
        "your_provider": "powermem.storage.your_provider.your_provider.YourProviderVectorStore",  # Add this
    }

Step 4: Add Tests​

Create test file tests/integration/test_your_provider_vector_store.py:

import pytest
from powermem.storage.factory import VectorStoreFactory


def test_your_provider_create_col():
    """Test collection creation"""
    config = {
        "collection_name": "test_collection",
        "host": "127.0.0.1",
        "port": 5432,
    }
    store = VectorStoreFactory.create("your_provider", config)
    store.create_col("test_col", vector_size=128, distance="cosine")
    assert store.col_info() is not None


def test_your_provider_insert_and_search():
    """Test insertion and search"""
    config = {
        "collection_name": "test_collection",
        "host": "127.0.0.1",
        "port": 5432,
    }
    store = VectorStoreFactory.create("your_provider", config)
    store.create_col("test_col", vector_size=128, distance="cosine")
    
    vectors = [[0.1] * 128, [0.2] * 128]
    payloads = [{"text": "test1"}, {"text": "test2"}]
    ids = store.insert(vectors, payloads)
    
    assert len(ids) == 2
    
    results = store.search([0.1] * 128, limit=1)
    assert len(results) > 0

Adding an LLM Provider​

To add a new LLM provider:

  1. Create the LLM implementation
  2. Create the configuration class
  3. Register it in the factory

Step 1: Create LLM Implementation​

Create src/powermem/integrations/llm/your_provider.py:

"""
YourProvider LLM implementation
"""
from typing import Dict, List, Optional

from powermem.integrations.llm import LLMBase
from powermem.integrations.llm.config.base import BaseLLMConfig
from powermem.integrations.llm.config.your_provider import YourProviderConfig

# Import your provider's SDK
try:
    from your_provider_sdk import YourProviderClient
except ImportError:
    raise ImportError(
        "The 'your_provider_sdk' library is required. "
        "Please install it using 'pip install your_provider_sdk'."
    )


class YourProviderLLM(LLMBase):
    """YourProvider LLM implementation"""
    
    def __init__(self, config: Optional[YourProviderConfig] = None):
        if config is None:
            config = YourProviderConfig()
        elif isinstance(config, dict):
            config = YourProviderConfig(**config)
        
        super().__init__(config)
        
        # Initialize your provider client
        self.client = YourProviderClient(
            api_key=self.config.api_key,
            base_url=getattr(self.config, 'base_url', None),
        )
    
    def generate_response(
        self,
        messages: List[Dict[str, str]],
        response_format=None,
        tools: Optional[List[Dict]] = None,
        tool_choice: str = "auto",
        **kwargs,
    ) -> str:
        """
        Generate a response based on the given messages.
        
        Args:
            messages: List of message dicts with 'role' and 'content'
            response_format: Optional response format specification
            tools: Optional list of tools for function calling
            tool_choice: Tool choice strategy
            **kwargs: Additional parameters
            
        Returns:
            str: Generated response text
        """
        # Prepare parameters
        params = self._get_supported_params(
            messages=messages,
            tools=tools,
            tool_choice=tool_choice,
            **kwargs
        )
        
        # Add provider-specific parameters
        params.update({
            "model": self.config.model,
            "temperature": self.config.temperature,
            "max_tokens": self.config.max_tokens,
        })
        
        # Call your provider's API
        response = self.client.chat.completions.create(**params)
        
        # Parse and return response
        return self._parse_response(response, tools)
    
    def _parse_response(self, response, tools=None):
        """Parse the response from the provider"""
        if tools:
            # Handle tool calls if supported
            return {
                "content": response.choices[0].message.content,
                "tool_calls": getattr(response.choices[0].message, 'tool_calls', []),
            }
        else:
            return response.choices[0].message.content

Step 2: Create Configuration Class​

Create src/powermem/integrations/llm/config/your_provider.py:

"""
Configuration for YourProvider LLM
"""
from typing import Optional
from pydantic import Field

from powermem.integrations.llm.config.base import BaseLLMConfig


class YourProviderConfig(BaseLLMConfig):
    """Configuration for YourProvider LLM"""
    
    base_url: Optional[str] = Field(
        default=None,
        description="Base URL for the API"
    )
    # Add provider-specific fields
    
    class Config:
        extra = "allow"

Step 3: Register in Factory​

Update src/powermem/integrations/llm/factory.py:

from powermem.integrations.llm.config.your_provider import YourProviderConfig

class LLMFactory:
    provider_to_class = {
        # ... existing providers ...
        "your_provider": ("powermem.integrations.llm.your_provider.YourProviderLLM", YourProviderConfig),
    }

Step 4: Add to Dependencies​

If your provider requires a new dependency, add it to pyproject.toml:

[project]
dependencies = [
    # ... existing dependencies ...
    "your_provider_sdk>=1.0.0",
]

Adding an Embedding Provider​

To add a new embedding provider:

  1. Create the embedding implementation
  2. Register it in the factory

Step 1: Create Embedding Implementation​

Create src/powermem/integrations/embeddings/your_provider.py:

"""
YourProvider embedding implementation
"""
from typing import Literal, Optional, List

from powermem.integrations.embeddings import EmbeddingBase
from powermem.integrations.embeddings.config.base import BaseEmbedderConfig

# Import your provider's SDK
try:
    from your_provider_sdk import YourProviderEmbeddingClient
except ImportError:
    raise ImportError(
        "The 'your_provider_sdk' library is required. "
        "Please install it using 'pip install your_provider_sdk'."
    )


class YourProviderEmbedding(EmbeddingBase):
    """YourProvider embedding implementation"""
    
    def __init__(self, config: Optional[BaseEmbedderConfig] = None):
        super().__init__(config)
        
        # Initialize your provider client
        self.client = YourProviderEmbeddingClient(
            api_key=getattr(self.config, 'api_key', None),
            model=self.config.model,
        )
    
    def embed(
        self, 
        text: str, 
        memory_action: Optional[Literal["add", "search", "update"]] = None
    ) -> List[float]:
        """
        Get the embedding for the given text.
        
        Args:
            text: The text to embed
            memory_action: The type of embedding to use (optional)
            
        Returns:
            List[float]: The embedding vector
        """
        # Some providers support different embedding models for different actions
        model = self.config.model
        if memory_action == "search" and hasattr(self.config, 'search_model'):
            model = self.config.search_model
        
        # Call your provider's API
        response = self.client.embed(text, model=model)
        
        return response.embedding

Step 2: Register in Factory​

Update src/powermem/integrations/embeddings/factory.py:

class EmbedderFactory:
    provider_to_class = {
        # ... existing providers ...
        "your_provider": "powermem.integrations.embeddings.your_provider.YourProviderEmbedding",
    }

Adding a Reranker Provider​

Rerankers improve search results by reordering documents based on relevance to the query. To add a new reranker provider:

  1. Create the reranker implementation
  2. Register it in the factory

Step 1: Create Reranker Implementation​

Create src/powermem/integrations/rerank/your_provider.py:

"""
YourProvider reranker implementation
"""
import os
from typing import List, Optional, Tuple

from powermem.integrations.rerank.base import RerankBase
from powermem.integrations.rerank.config.base import BaseRerankConfig

# Import your provider's SDK
try:
    from your_provider_sdk import YourProviderRerankClient
except ImportError:
    raise ImportError(
        "The 'your_provider_sdk' library is required. "
        "Please install it using 'pip install your_provider_sdk'."
    )


class YourProviderRerank(RerankBase):
    """YourProvider reranker implementation"""
    
    def __init__(self, config: Optional[BaseRerankConfig] = None):
        super().__init__(config)
        
        # Set default model
        self.config.model = self.config.model or "your-rerank-model"
        
        # Initialize your provider client
        api_key = self.config.api_key or os.getenv("YOUR_PROVIDER_API_KEY")
        if not api_key:
            raise ValueError(
                "API key is required. Set YOUR_PROVIDER_API_KEY environment variable "
                "or pass api_key in config."
            )
        
        self.client = YourProviderRerankClient(api_key=api_key)
    
    def rerank(
        self, 
        query: str, 
        documents: List[str], 
        top_n: Optional[int] = None
    ) -> List[Tuple[int, float]]:
        """
        Rerank documents based on relevance to the query.
        
        Args:
            query: The search query
            documents: List of document texts to rerank
            top_n: Number of top results to return
            
        Returns:
            List[Tuple[int, float]]: List of (document_index, relevance_score) tuples,
                                     sorted by relevance score in descending order
        """
        if not query or not query.strip():
            raise ValueError("Query cannot be empty")
        
        if not documents or len(documents) == 0:
            raise ValueError("Documents list cannot be empty")
        
        # Use provided top_n or return all results
        effective_top_n = top_n if top_n is not None else len(documents)
        
        try:
            # Call your provider's rerank API
            response = self.client.rerank(
                query=query,
                documents=documents,
                model=self.config.model,
                top_n=effective_top_n,
            )
            
            # Parse results - format: [(index, score), ...]
            results = []
            for item in response.results:
                index = item.index
                score = item.relevance_score
                results.append((index, float(score)))
            
            # Sort by score descending (highest first)
            results.sort(key=lambda x: x[1], reverse=True)
            
            return results
            
        except Exception as e:
            raise Exception(f"Failed to rerank documents: {e}")

Step 2: Register in Factory​

Update src/powermem/integrations/rerank/factory.py:

class RerankFactory:
    provider_to_class = {
        "qwen": "powermem.integrations.rerank.qwen.QwenRerank",
        "your_provider": "powermem.integrations.rerank.your_provider.YourProviderRerank",  # Add this
    }

Step 3: Configure Reranker​

Enable reranker in your configuration:

from powermem import Memory

config = {
    "reranker": {
        "enabled": True,
        "provider": "your_provider",
        "config": {
            "model": "your-rerank-model",
            "api_key": "your-api-key",
        },
    },
    # ... other config
}

memory = Memory(config=config)

Or via environment variables:

export RERANKER_ENABLED=true
export RERANKER_PROVIDER=your_provider
export RERANKER_MODEL=your-rerank-model
export RERANKER_API_KEY=your-api-key

Testing​

Running Tests​

# Run all tests
make test

# Run unit tests only
make test-unit

# Run integration tests only
make test-integration

# Run end-to-end tests
make test-e2e

# Run tests with coverage
make test-coverage

Writing Tests​

  • Unit tests: Test individual functions and classes in isolation
  • Integration tests: Test interactions between components
  • E2E tests: Test complete workflows

Place tests in:

  • tests/unit/ for unit tests
  • tests/integration/ for integration tests
  • tests/e2e/ for end-to-end tests

Test Structure​

import pytest
from powermem import Memory


def test_feature_name():
    """Test description"""
    # Arrange
    memory = Memory()
    
    # Act
    result = memory.some_method()
    
    # Assert
    assert result is not None

Code Style and Standards​

Formatting​

The project uses black for code formatting and isort for import sorting:

# Format code
make format

# Check formatting without making changes
make format-check

Linting​

# Run linter
make lint

Type Checking​

# Run type checker
make type-check

Code Style Guidelines​

  1. Follow PEP 8 style guide
  2. Use type hints for function parameters and return values
  3. Write docstrings for all classes and public methods
  4. Keep functions focused - one responsibility per function
  5. Use meaningful variable names
  6. Handle errors appropriately - use try/except where needed
  7. Add logging for important operations

Example Code Style​

"""
Module docstring describing the purpose of this module.
"""
import logging
from typing import Dict, List, Optional

logger = logging.getLogger(__name__)


class ExampleClass:
    """
    Class docstring describing what this class does.
    
    Args:
        param1: Description of param1
        param2: Description of param2
    """
    
    def __init__(self, param1: str, param2: Optional[int] = None):
        """Initialize the class."""
        self.param1 = param1
        self.param2 = param2
        logger.info(f"Initialized ExampleClass with param1={param1}")
    
    def example_method(self, input_data: Dict[str, str]) -> List[str]:
        """
        Method docstring describing what this method does.
        
        Args:
            input_data: Dictionary containing input data
            
        Returns:
            List of processed strings
            
        Raises:
            ValueError: If input_data is empty
        """
        if not input_data:
            raise ValueError("input_data cannot be empty")
        
        # Implementation here
        return []

Publishing to PyPI​

Prerequisites​

Before publishing to PyPI, ensure you have:

  1. PyPI account: Create an account at pypi.org
  2. API tokens: Generate API tokens from your PyPI account settings
  3. Configure credentials: Set up ~/.pypirc or use environment variables

Publishing Process​

  1. Update version in pyproject.toml:
[project]
version = "0.1.1"  # Increment version number

  1. Update changelog (if you maintain one)

  2. Build and check the package:

make build-check
  1. Test locally:
make install-local
# Test the installed package
python -c "import powermem; print(powermem.__version__)"
  1. Publish to TestPyPI first (recommended):
make publish-testpypi
  1. Test installation from TestPyPI:
pip install --index-url https://test.pypi.org/simple/ powermem
  1. Publish to PyPI:
make publish-pypi

Version Management​

Follow Semantic Versioning:

  • MAJOR version (1.0.0): Incompatible API changes
  • MINOR version (0.1.0): New functionality in a backward compatible manner
  • PATCH version (0.0.1): Backward compatible bug fixes

Git Tagging​

After publishing, create a git tag:

git tag -a v0.1.1 -m "Release version 0.1.1"
git push origin v0.1.1

Debugging​

Enable Debug Logging​

Set the logging level to DEBUG:

import logging
logging.basicConfig(level=logging.DEBUG)

Or set environment variable:

export LOG_LEVEL=DEBUG

Common Debugging Scenarios​

  1. Vector Store Connection Issues:
from powermem.storage.factory import VectorStoreFactory

# Test connection
config = {
    "collection_name": "test",
    "host": "127.0.0.1",
    "port": 2881,
    # ... other config
}
store = VectorStoreFactory.create("oceanbase", config)
print(store.col_info())  # Check if connection works
  1. LLM API Issues:
from powermem.integrations.llm.factory import LLMFactory

# Test LLM connection
config = {
    "model": "gpt-4",
    "api_key": "your-key",
}
llm = LLMFactory.create("openai", config)
response = llm.generate_response([{"role": "user", "content": "test"}])
print(response)
  1. Embedding Issues:
from powermem.integrations.embeddings.factory import EmbedderFactory

# Test embedding
config = {
    "model": "text-embedding-3-small",
    "api_key": "your-key",
}
embedder = EmbedderFactory.create("openai", config)
vector = embedder.embed("test text")
print(f"Vector dimension: {len(vector)}")

Performance Optimization​

Vector Store Optimization​

  1. Index Configuration:
# For OceanBase
config = {
    "index_type": "HNSW",  # Use HNSW for better performance
    "vidx_metric_type": "l2",  # Choose appropriate metric
    # ... other config
}
  1. Batch Operations:
# Insert multiple vectors at once
vectors = [[0.1] * 1536 for _ in range(100)]
payloads = [{"text": f"text_{i}"} for i in range(100)]
ids = store.insert(vectors, payloads)  # Batch insert
  1. Connection Pooling:
# Use connection pooling for better performance
config = {
    "pool_size": 10,
    "max_overflow": 20,
    # ... other config
}

Memory Optimization​

  1. Use Async Operations:
from powermem import AsyncMemory

memory = AsyncMemory()
# Async operations are more efficient for I/O-bound tasks
await memory.add("text", user_id="user1")
  1. Batch Memory Operations:
# Add multiple memories efficiently
memories = [
    "User likes Python",
    "User works at tech company",
    "User prefers coffee over tea",
]
for mem in memories:
    memory.add(mem, user_id="user1")

Common Issues and Solutions​

Issue: Import Errors​

Problem: ModuleNotFoundError when importing provider-specific modules

Solution: Install the required dependencies:

# For specific providers
pip install pyobvector  # For OceanBase
pip install pgvector    # For PostgreSQL
pip install dashscope    # For Qwen

Issue: Configuration Not Loading​

Problem: Configuration from .env file not being loaded

Solution: Ensure the file path is correct:

from powermem import create_memory

# Explicitly specify config file
memory = create_memory(config_file=".env")

Issue: Vector Dimension Mismatch​

Problem: ValueError: Vector dimension mismatch

Solution: Ensure embedding model dimensions match vector store configuration:

# Check embedding dimensions
embedder = EmbedderFactory.create("openai", {"model": "text-embedding-3-small"})
vector = embedder.embed("test")
print(f"Dimension: {len(vector)}")  # Should be 1536 for text-embedding-3-small

# Configure vector store with matching dimensions
store_config = {
    "embedding_model_dims": 1536,  # Must match embedding dimension
    # ... other config
}

Issue: Connection Timeout​

Problem: Timeout errors when connecting to vector store

Solution: Increase timeout and check network:

config = {
    "host": "127.0.0.1",
    "port": 2881,
    "connect_timeout": 30,  # Increase timeout
    # ... other config
}

Advanced Topics​

Custom Fact Extraction​

You can customize how facts are extracted from messages:

from powermem import Memory

custom_prompt = """
Extract key facts from the following conversation.
Focus on: user preferences, important events, relationships.
"""

memory = Memory(
    config={
        "custom_fact_extraction_prompt": custom_prompt,
        # ... other config
    }
)

Custom Memory Update Logic​

Customize how memories are updated:

custom_update_prompt = """
Update existing memories based on new information.
Merge similar memories and remove outdated ones.
"""

memory = Memory(
    config={
        "custom_update_memory_prompt": custom_update_prompt,
        # ... other config
    }
)

Sub-Stores Configuration​

Configure sub-stores for better data organization:

config = {
    "storage_type": "oceanbase",
    "sub_stores": [
        {
            "name": "user_preferences",
            "filters": {"type": "preference"},
        },
        {
            "name": "events",
            "filters": {"type": "event"},
        },
    ],
    # ... other config
}
memory = Memory(config=config)

Graph Store Integration​

Enable graph store for relationship-based retrieval:

config = {
    "graph_store": {
        "provider": "oceanbase",
        "config": {
            # Graph store configuration
        },
    },
    # ... other config
}
memory = Memory(config=config)

Additional Resources​

  • API Documentation: See docs/api/ for detailed API reference
  • Examples: Check examples/ directory for usage examples
  • Architecture: See docs/architecture/overview.md for system architecture
  • Issues: Report bugs or request features on GitHub Issues

Getting Help​

Happy coding! 🚀