Model Adapters

Reeflect uses adapters to integrate with different LLM providers, making it easy to switch between models or use multiple models in the same application.

Available Adapters

OpenAI Adapter

Integrate with GPT models and embeddings.

OpenAIAdapter

Anthropic Adapter

Integrate with Claude models and embeddings.

AnthropicAdapter

Mistral Adapter

Integrate with Mistral AI models and embeddings.

MistralAdapter

Llama Adapter

Integrate with local Llama models.

LlamaAdapter

Adapter Configuration

# OpenAI Adapter
from reeflect.adapters.openai import OpenAIAdapter

openai_adapter = OpenAIAdapter(
    api_key="your_openai_api_key",
    embedding_model="text-embedding-3-small",
    completion_model="gpt-4-turbo-preview",
    max_tokens_per_memory=250
)

# Anthropic Adapter
from reeflect.adapters.anthropic import AnthropicAdapter

anthropic_adapter = AnthropicAdapter(
    api_key="your_anthropic_api_key",
    embedding_model="claude-3-sonnet-20240229",
    completion_model="claude-3-opus-20240229",
    max_tokens_per_memory=300
)

Custom Adapter Implementation

You can implement custom adapters for other LLM providers by extending the BaseMemoryAdapter class:

from reeflect.adapters.base import BaseMemoryAdapter
from reeflect.core.memory import Memory
from typing import Any, Dict, List, Optional

class CustomAdapter(BaseMemoryAdapter):
    """Custom adapter for your LLM provider."""
    
    def __init__(self, api_key, **kwargs):
        self.api_key = api_key
        self.client = initialize_your_client(api_key)
        
    def generate_embedding(self, text: str) -> List[float]:
        # Implement embedding generation
        pass
        
    def batch_generate_embeddings(self, texts: List[str]) -> List[List[float]]:
        # Implement batch embedding generation
        pass
        
    def inject_memories_to_prompt(
        self,
        prompt: str,
        memories: List[Memory],
        max_tokens: int = 1000
    ) -> str:
        # Implement memory injection
        pass
        
    # Implement other required methods...
        
# Use your custom adapter
memory = Reeflect(
    adapter=CustomAdapter(api_key="your_api_key"),
    storage_config={...}
)

Next Steps

Now that you understand how to connect Reeflect to different LLM providers, explore Memory Hierarchies to learn how the system organizes different types of memories.