mnemefusion 0.1.2

Unified memory engine for AI applications - SQLite for AI memory (CPU build)

MnemeFusion Python Bindings

Python bindings for MnemeFusion - a unified memory engine for AI applications.

Installation

From Source (Development)

cd mnemefusion-python
pip install maturin
maturin develop

From PyPI (when published)

pip install mnemefusion

Quick Start

import mnemefusion

# Open or create a database
memory = mnemefusion.Memory("brain.mfdb", config={
    "embedding_dim": 384,
    "entity_extraction_enabled": True,
})

# Add a memory
memory_id = memory.add(
    "Meeting scheduled for March 15th",
    embedding=[0.1] * 384,  # Your embedding vector
    metadata={"type": "event"}
)

# Semantic search
results = memory.search(query_embedding, top_k=10)
for mem, score in results:
    print(f"{score:.3f}: {mem['content']}")

# Intelligent query with intent classification
intent, results = memory.query(
    "Why was the meeting cancelled?",
    query_embedding,
    limit=10
)
print(f"Intent: {intent['intent']} ({intent['confidence']:.2f})")
for mem, scores in results:
    print(f"Fused score: {scores['fused_score']:.3f}")
    print(f"Content: {mem['content']}")

# Close the database
memory.close()

Using Context Manager

with mnemefusion.Memory("brain.mfdb") as memory:
    memory.add("Some content", embedding)
    # Database automatically closed when exiting

API Reference

Memory Class

__init__(path, config=None)

Create or open a memory database.

Parameters:

• path (str): Path to the .mfdb file
• config (dict, optional): Configuration options
  • embedding_dim (int): Dimension of embedding vectors (default: 384)
  • entity_extraction_enabled (bool): Enable automatic entity extraction (default: True)

Returns: Memory instance

add(content, embedding, metadata=None, timestamp=None)

Add a new memory to the database.

Parameters:

• content (str): Text content
• embedding (List[float]): Vector embedding
• metadata (Dict[str, str], optional): Key-value metadata
• timestamp (float, optional): Unix timestamp (seconds since epoch)

Returns: Memory ID as string

get(memory_id)

Retrieve a memory by ID.

Parameters:

• memory_id (str): Memory ID

Returns: Dictionary with memory data, or None if not found

Memory dict structure:

{
    "id": str,
    "content": str,
    "embedding": List[float],
    "metadata": Dict[str, str],
    "created_at": float  # Unix timestamp
}

delete(memory_id)

Delete a memory by ID.

Parameters:

• memory_id (str): Memory ID

Returns: bool - True if deleted, False if not found

search(query_embedding, top_k)

Semantic similarity search.

Parameters:

• query_embedding (List[float]): Query vector
• top_k (int): Number of results to return

Returns: List of (memory_dict, similarity_score) tuples

query(query_text, query_embedding, limit)

Intelligent multi-dimensional query with intent classification.

Parameters:

• query_text (str): Natural language query
• query_embedding (List[float]): Query vector
• limit (int): Maximum number of results

Returns: Tuple of (intent_dict, results_list)

• intent_dict: {"intent": str, "confidence": float}
• results_list: List of (memory_dict, scores_dict) tuples

Scores dict structure:

{
    "semantic_score": float,
    "temporal_score": float,
    "causal_score": float,
    "entity_score": float,
    "fused_score": float
}

Intent types:

• "Temporal": Time-based queries ("yesterday", "recent")
• "Causal": Cause-effect queries ("why", "because")
• "Entity": Entity-focused queries ("about X", "mentioning Y")
• "Factual": Generic semantic search

count()

Get the number of memories in the database.

Returns: int - Count of memories

add_causal_link(cause_id, effect_id, confidence, evidence)

Add a causal relationship between two memories.

Parameters:

• cause_id (str): Memory ID of the cause
• effect_id (str): Memory ID of the effect
• confidence (float): Confidence score (0.0 to 1.0)
• evidence (str): Text explaining the relationship

get_causes(memory_id, max_hops)

Get causes of a memory (backward traversal).

Parameters:

• memory_id (str): Memory ID
• max_hops (int): Maximum traversal depth

Returns: List of causal paths (each path is a list of memory ID strings)

get_effects(memory_id, max_hops)

Get effects of a memory (forward traversal).

Parameters:

• memory_id (str): Memory ID
• max_hops (int): Maximum traversal depth

Returns: List of causal paths (each path is a list of memory ID strings)

list_entities()

List all entities in the database.

Returns: List of entity dictionaries

Entity dict structure:

{
    "id": str,
    "name": str,
    "mention_count": int,
    "metadata": Dict[str, str]
}

close()

Close the database and save all indexes.

Examples

See examples/basic_usage.py for a comprehensive example.

Requirements

• Python >= 3.8
• Rust toolchain (for building from source)

License

MIT