Configuration

Remina uses a unified configuration system powered by Pydantic.

Configuration Structure

from remina import Memory
 
config = {
    # L1 Cache (Redis - fixed)
    "cache": { ... },
    
    # L2 Storage (pluggable)
    "storage": { ... },
    
    # Vector Store (pluggable)
    "vector_store": { ... },
    
    # Embeddings (pluggable)
    "embedder": { ... },
    
    # LLM (pluggable)
    "llm": { ... },
    
    # Scoring weights
    "weight_recency": 0.4,
    "weight_frequency": 0.3,
    "weight_importance": 0.3,
    
    # Thresholds
    "consolidation_threshold": 0.85,
    "dedup_threshold": 0.9,
    "max_facts_per_conversation": 10,
}
 
memory = Memory(config)

Cache Configuration

Redis L1 cache is fixed (not pluggable) for performance consistency.

"cache": {
    "redis_url": "redis://localhost:6379",
    "ttl_seconds": 3600,
    "max_memories_per_user": 100,
    "enabled": True,
    "key_prefix": "remina",
}

Option	Type	Default	Description
`redis_url`	str	`redis://localhost:6379`	Redis connection URL
`ttl_seconds`	int	3600	Cache TTL
`max_memories_per_user`	int	100	Max cached per user
`enabled`	bool	True	Enable/disable caching
`key_prefix`	str	`remina`	Redis key prefix

Storage Configuration

L2 persistent storage. Select one provider.

SQLite (Default)

"storage": {
    "provider": "sqlite",
    "config": {
        "path": "~/.remina/memories.db",
        "collection_name": "memories",
    }
}

PostgreSQL

"storage": {
    "provider": "postgres",
    "config": {
        "host": "localhost",
        "port": 5432,
        "database": "remina",
        "user": "postgres",
        "password": "password",
        "collection_name": "memories",
        "min_connections": 1,
        "max_connections": 10,
    }
}

MongoDB

"storage": {
    "provider": "mongodb",
    "config": {
        "uri": "mongodb://localhost:27017",
        "database": "remina",
        "collection_name": "memories",
    }
}

Vector Store Configuration

Chroma (Default)

"vector_store": {
    "provider": "chroma",
    "config": {
        "path": "~/.remina/chroma",
        "collection_name": "remina",
    }
}

Qdrant

"vector_store": {
    "provider": "qdrant",
    "config": {
        "url": "http://localhost:6333",
        "api_key": None,
        "collection_name": "remina",
        "embedding_dims": 1536,
    }
}

Pinecone

"vector_store": {
    "provider": "pinecone",
    "config": {
        "api_key": "your-api-key",
        "index_name": "remina",
        "namespace": "default",
        "embedding_dims": 768,
        "cloud": "aws",
        "region": "us-east-1",
    }
}

Embedder Configuration

OpenAI (Default)

"embedder": {
    "provider": "openai",
    "config": {
        "api_key": None,  # Uses OPENAI_API_KEY env
        "model": "text-embedding-3-small",
        "dimensions": 1536,
        "base_url": None,
    }
}

Google Gemini

"embedder": {
    "provider": "gemini",
    "config": {
        "api_key": None,  # Uses GOOGLE_API_KEY env
        "model": "models/text-embedding-004",
        "dimensions": 768,
    }
}

Cohere

"embedder": {
    "provider": "cohere",
    "config": {
        "api_key": None,  # Uses COHERE_API_KEY env
        "model": "embed-english-v3.0",
    }
}

Ollama (Local)

"embedder": {
    "provider": "ollama",
    "config": {
        "base_url": "http://localhost:11434",
        "model": "nomic-embed-text",
    }
}

HuggingFace (Local)

"embedder": {
    "provider": "huggingface",
    "config": {
        "model": "sentence-transformers/all-MiniLM-L6-v2",
        "device": "cpu",
    }
}

LLM Configuration

OpenAI (Default)

"llm": {
    "provider": "openai",
    "config": {
        "api_key": None,  # Uses OPENAI_API_KEY env
        "model": "gpt-4o-mini",
        "temperature": 0.1,
        "max_tokens": 2000,
        "base_url": None,
    }
}

Google Gemini

"llm": {
    "provider": "gemini",
    "config": {
        "api_key": None,  # Uses GOOGLE_API_KEY env
        "model": "gemini-2.0-flash",
        "temperature": 0.1,
        "max_tokens": 2000,
    }
}

Anthropic Claude

"llm": {
    "provider": "anthropic",
    "config": {
        "api_key": None,  # Uses ANTHROPIC_API_KEY env
        "model": "claude-3-5-sonnet-20240620",
        "temperature": 0.1,
        "max_tokens": 2000,
    }
}

Ollama (Local)

"llm": {
    "provider": "ollama",
    "config": {
        "base_url": "http://localhost:11434",
        "model": "llama3.2",
        "temperature": 0.1,
    }
}

Scoring Weights

Control memory ranking during search:

# Must sum to 1.0
"weight_recency": 0.4,      # Temporal decay factor
"weight_frequency": 0.3,    # Access count factor
"weight_importance": 0.3,   # Base importance factor

Thresholds

# Consolidation: similarity threshold for merging
"consolidation_threshold": 0.85,
 
# Deduplication: similarity threshold for duplicate detection
"dedup_threshold": 0.9,
 
# Max facts extracted per conversation
"max_facts_per_conversation": 10,

Environment Variables

Variable	Provider
`OPENAI_API_KEY`	OpenAI embeddings/LLM
`OPENAI_BASE_URL`	Custom OpenAI endpoint
`GOOGLE_API_KEY`	Gemini embeddings/LLM
`ANTHROPIC_API_KEY`	Anthropic Claude
`COHERE_API_KEY`	Cohere embeddings
`PINECONE_API_KEY`	Pinecone vector store

Complete Example

from remina import Memory
 
config = {
    "cache": {
        "redis_url": "redis://localhost:6379",
        "ttl_seconds": 3600,
        "enabled": True,
    },
    "storage": {
        "provider": "postgres",
        "config": {
            "host": "localhost",
            "port": 5432,
            "database": "remina",
            "user": "postgres",
            "password": "password",
        }
    },
    "vector_store": {
        "provider": "qdrant",
        "config": {
            "url": "http://localhost:6333",
            "collection_name": "memories",
            "embedding_dims": 768,
        }
    },
    "embedder": {
        "provider": "gemini",
        "config": {
            "model": "models/text-embedding-004",
            "dimensions": 768,
        }
    },
    "llm": {
        "provider": "gemini",
        "config": {
            "model": "gemini-2.0-flash",
            "temperature": 0.1,
        }
    },
    "weight_recency": 0.4,
    "weight_frequency": 0.3,
    "weight_importance": 0.3,
    "dedup_threshold": 0.9,
}
 
memory = Memory(config)

Pydantic Config Classes

For type safety:

from remina.configs.base import MemoryConfig
from remina.configs.storage import StorageConfig
from remina.configs.vectors import VectorStoreConfig
from remina.configs.embeddings import EmbedderConfig
from remina.configs.llms import LLMConfig
from remina.configs.cache import CacheConfig
 
config = MemoryConfig(
    cache=CacheConfig(redis_url="redis://localhost:6379"),
    storage=StorageConfig(provider="postgres", config={...}),
    vector_store=VectorStoreConfig(provider="qdrant", config={...}),
    embedder=EmbedderConfig(provider="gemini", config={...}),
    llm=LLMConfig(provider="gemini", config={...}),
)
 
memory = Memory(config)

Next Steps

Providers — Detailed provider documentation
Examples — Configuration examples
Deployment — Production deployment

Core Concepts Overview