Smriti Memory

An intelligent memory layer for AI applications with RAG (Retrieval-Augmented Generation) capabilities. Smriti Memory provides sophisticated memory management that can store, retrieve, and update contextual information using vector databases and LLM-powered decision making.

🚀 Prerequisites

Before using Smriti Memory, you'll need to obtain API keys from the following services:

Required API Keys

1. Pinecone API Key - For vector database storage
2. Groq API Key - For LLM operations (memory decisions and chat)
3. Gemini API Key - For additional LLM capabilities

Quick Setup

1. Set Environment Variables (recommended):

export PINECONE_API_KEY="your-pinecone-api-key"
export GROQ_API_KEY="your-groq-api-key"
export GEMINI_KEY="your-gemini-api-key"

2. Or Pass Keys Directly:

from smriti import MemoryConfig, MemoryManager

config = MemoryConfig(
    pinecone_api_key="your-pinecone-key",
    groq_api_key="your-groq-key", 
    gemini_api_key="your-gemini-key"
)
memory_manager = MemoryManager(config)

📖 Need help getting your API keys? Check out our detailed Setup Guide with step-by-step instructions!

Features

• 🧠 Intelligent Memory Management: Uses LLM to decide what information to store, update, or ignore
• 🔍 Semantic Search: Find relevant memories using vector similarity search
• 🔄 Memory Updates: Automatically detect and update existing memories with new information
• 📊 Memory Statistics: Track and analyze memory usage patterns
• 🚀 Easy Integration: Simple API for adding to any AI application
• 🛠️ CLI Interface: Command-line tools for memory operations
• ⚙️ Configurable: Flexible configuration for different use cases

Installation

pip install smriti-memory

Development Installation

git clone https://github.com/amanyadav721/smriti.git
cd smriti-memory
pip install -e .

Quick Start

Basic Usage

from smriti import MemoryManager, MemoryConfig

# Initialize with default configuration
memory_manager = MemoryManager()

# Add memory from a chat interaction
chat_thread = [
    {"user": "I like pizza and reading sci-fi books", "ai": "That's great! What's your favorite sci-fi book?"}
]

result = memory_manager.add_memory("user123", chat_thread)
print(result)
# Output: {'success': True, 'memory': [...], 'action': 'added', 'count': 1}

# Search for relevant memories
search_result = memory_manager.search_memories("user123", "pizza")
print(search_result)
# Output: {'success': True, 'results': [...], 'count': 1}

# Search with custom parameters (top_k and namespace)
search_result = memory_manager.search_memories("user123", "pizza", top_k=5, namespace="user_understanding")
print(f"Found {len(search_result['results'])} results")

# Chat with memory context
chat_result = memory_manager.chat_with_memory("user123", "What do I like?")
print(chat_result["response"])
# Output: "Based on our previous conversation, you like pizza and reading sci-fi books..."

CLI Usage

# Add memory
smriti add-memory user123 --chat-thread '[{"user": "I like pizza"}]'

# Search memories
smriti search user123 --query "pizza" --verbose

# Chat with memory
smriti chat user123 --query "What do I like?"

# Get user statistics
smriti stats user123

# Delete all memories
smriti delete user123

Configuration

Environment Variables

Set these environment variables for API access:

export PINECONE_API_KEY="your-pinecone-api-key"
export GROQ_API_KEY="your-groq-api-key"
export GEMINI_KEY="your-gemini-api-key"

Custom Configuration

from smriti import MemoryConfig, MemoryManager

config = MemoryConfig(
    pinecone_api_key="your-key",
    groq_api_key="your-key",
    gemini_api_key="your-key",
    llm_model="llama-3.1-8b-instant",
    llm_temperature=0.3,
    default_namespace="user_understanding",
    max_memory_length=1000,
    similarity_threshold=0.7,
    max_search_results=10
)

memory_manager = MemoryManager(config)

API Reference

MemoryManager

The main class for managing memories.

add_memory(user_id: str, chat_thread: List[Dict[str, Any]]) -> Dict[str, Any]

Add memory based on a chat thread.

Parameters:

• user_id: Unique identifier for the user
• chat_thread: List of dictionaries with "user" and/or "ai" keys

Returns:

{
    "success": bool,
    "memory": List[Dict],
    "action": str,  # "added" or "ignored"
    "namespace": str,
    "count": int,
    "storage_result": Dict
}

search_memories(user_id: str, query: str, namespace: Optional[str] = None, top_k: Optional[int] = None) -> Dict[str, Any]

Search for relevant memories.

Parameters:

• user_id: User identifier
• query: Search query
• namespace: Optional namespace to search in
• top_k: Number of results to return

Returns:

{
    "success": bool,
    "results": List[Dict],
    "query": str,
    "namespace": str,
    "count": int
}

chat_with_memory(user_id: str, query: str, add_to_memory: bool = True) -> Dict[str, Any]

Generate a chat response using memory context.

Parameters:

• user_id: User identifier
• query: User's query
• add_to_memory: Whether to add the interaction to memory

Returns:

{
    "success": bool,
    "response": str,
    "memory_context": Dict,
    "memory_result": Dict
}

delete_user_memories(user_id: str) -> Dict[str, Any]

Delete all memories for a user.

get_user_stats(user_id: str) -> Dict[str, Any]

Get statistics about a user's memories.

MemoryConfig

Configuration class for customizing behavior.

Attributes:

• pinecone_api_key: Pinecone API key
• groq_api_key: Groq API key
• gemini_api_key: Gemini API key
• llm_model: LLM model name
• llm_temperature: LLM temperature
• default_namespace: Default namespace for memories
• max_memory_length: Maximum length of memory text
• similarity_threshold: Similarity threshold for search
• max_search_results: Maximum number of search results

Advanced Usage

Custom Memory Types

# Add memory with custom namespace
chat_thread = [{"user": "I have a meeting on Friday"}]
result = memory_manager.add_memory("user123", chat_thread)

# Search in specific namespace
search_result = memory_manager.search_memories(
    "user123", 
    "meeting", 
    namespace="facts"
)

# Search with custom top_k parameter
search_result = memory_manager.search_memories(
    "user123", 
    "pizza", 
    top_k=3  # Get only top 3 results
)

# Search with both namespace and top_k
search_result = memory_manager.search_memories(
    "user123", 
    "work", 
    namespace="professional",
    top_k=10  # Get top 10 results from professional namespace
)

Batch Operations

# Add multiple memories
chat_threads = [
    [{"user": "I like pizza"}],
    [{"user": "I work at Google"}],
    [{"user": "I have a dog named Max"}]
]

for thread in chat_threads:
    memory_manager.add_memory("user123", thread)

Error Handling

from smriti import SmritiError, MemoryError

try:
    result = memory_manager.add_memory("user123", chat_thread)
    if not result["success"]:
        print(f"Error: {result['error']}")
except SmritiError as e:
    print(f"Smriti error: {e.message}")
    print(f"Details: {e.details}")

CLI Commands

smriti add-memory

Add memory from a chat thread.

smriti add-memory user123 --chat-thread '[{"user": "I like pizza"}]'

smriti search

Search for memories.

smriti search user123 --query "pizza" --namespace "user_understanding" --top-k 5

smriti chat

Chat with memory context.

smriti chat user123 --query "What do I like?" --no-memory

smriti delete

Delete all memories for a user.

smriti delete user123

smriti stats

Get user memory statistics.

smriti stats user123 --verbose

Examples

Chatbot Integration

from smriti import MemoryManager

class Chatbot:
    def __init__(self):
        self.memory_manager = MemoryManager()
    
    def respond(self, user_id: str, message: str) -> str:
        # Get response with memory context
        result = self.memory_manager.chat_with_memory(user_id, message)
        
        if result["success"]:
            return result["response"]
        else:
            return "I'm sorry, I encountered an error. Please try again."
    
    def add_memory(self, user_id: str, user_message: str, ai_response: str):
        chat_thread = [{"user": user_message, "ai": ai_response}]
        self.memory_manager.add_memory(user_id, chat_thread)

# Usage
chatbot = Chatbot()
response = chatbot.respond("user123", "What do I like?")
print(response)

Web Application Integration

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from smriti import MemoryManager

app = FastAPI()
memory_manager = MemoryManager()

class ChatRequest(BaseModel):
    user_id: str
    message: str

@app.post("/chat")
async def chat(request: ChatRequest):
    result = memory_manager.chat_with_memory(request.user_id, request.message)
    
    if result["success"]:
        return {"response": result["response"]}
    else:
        raise HTTPException(status_code=500, detail=result["error"])

@app.get("/memories/{user_id}")
async def get_memories(user_id: str, query: str):
    result = memory_manager.search_memories(user_id, query)
    return result

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Development

Setup Development Environment

git clone https://github.com/amanyadav721/smriti.git
cd smriti-memory
pip install -e ".[dev]"

Running Tests

pytest

Code Formatting

black smriti/
flake8 smriti/
mypy smriti/

Building Documentation

pip install -e ".[docs]"
cd docs
make html

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Built with Pinecone for vector storage
• Powered by Groq for fast LLM inference
• Uses Google Gemini for embeddings
• Inspired by research in memory-augmented neural networks

Support

• 📧 Email: ad721603@gmail.com