gistlattice 0.2.0

GistLattice is a memory-augmented Python library with episodic recall, semantic state, and durable consolidation.

Cognitive Memory for AI Agents.
A beautiful, three-method Python library for giving your LLMs durable, long-term memory.

---

🧠 Why GistLattice?

Most "Agent Memory" systems simply dump raw chat transcripts into a vector database. GistLattice is different. It uses an LLM to actively reflect on a conversation before saving it.

When you ask GistLattice to remember an interaction, it extracts:

1. Gist: A concise summary of the factual information.
2. Valence: The emotional tone (-1.0 to 1.0). Did the user get angry? Were they excited? Your agent will remember their mood!
3. Importance: A score (0.0 to 1.0) dictating how crucial this memory is. Passing comments decay quickly; major life events are cemented permanently.

📦 Installation

Install the base library (defaults to fast, in-memory databases):

pip install gistlattice

Install production backends (optional):

pip install gistlattice[postgres,neo4j,redis]

Install specific LLM providers:

pip install gistlattice[openai]
pip install gistlattice[gemini]
pip install gistlattice[anthropic]
pip install gistlattice[ollama]

🚀 Quick Start

The entire surface area of the library is encapsulated in a single, elegant class: GistLattice.

from gistlattice import GistLattice

def main() -> None:
    # 1. Initialize the client (defaults to in-memory storage)
    memory = GistLattice(provider="openai", tenant_id="tenant-a", user_id="user-123")

    # 2. Store an interaction synchronously
    # This automatically buffers the interaction to eliminate per-turn LLM database mutations.
    memory.remember(
        prompt="I'm feeling really stressed about the product launch tomorrow.",
        response="I understand. Let's review the final checklist to make sure we are ready."
    )
    print("Memory buffered!")

    # 3. Retrieve formatted context to inject into your next LLM prompt
    context = memory.hydrate_context("What should I do next?")
    print("\n--- Hydrated Context ---")
    print(context)

if __name__ == "__main__":
    main()

[!TIP] For Best Performance: GistLattice fully supports native async/await! If you are building a high-throughput API (like FastAPI), use memory.aremember(), memory.aretrieve(), and memory.ahydrate_context() to bypass background thread overhead.

📐 Architecture Flow

GistLattice intercepts conversations and routes them through a robust processing pipeline:

graph LR
    App[Your AI App] --> Remember(memory.remember)
    App --> Hydrate(memory.hydrate_context)
    
    Remember --> Buffer[MemoryBufferController]
    Buffer --> LLM[LLM Reflection Analysis]
    
    LLM --> Storage[(StorageProvider: Postgres / Neo4j)]
    
    Hydrate --> Storage

📚 Documentation

We have completely stripped out the architectural jargon. Our documentation is heavily focused on getting you building instantly:

1. How it Works (A Human Example): A plain-English story demonstrating the difference between short-term buffers and GistLattice.
2. User Guide & API Reference: Everything you need to know about remember(), hydrate_context(), and retrieve().
3. Supported Providers: How to switch between OpenAI, Gemini, Anthropic, and Ollama.
4. Production Backends: How to connect to Redis, PostgreSQL, and Neo4j for durable, at-scale memory.

🛠 Examples

Want to see runnable code? Check out the examples/ directory:

• 01_basic_openai_sync.py
• 02_hybrid_gemini_async.py
• production/01_programmatic.py
• production/02_environment_variables.py

🤝 Contributing

We would love your help making GistLattice the standard for agent memory! To contribute:

1. Fork the repository.
2. Install dependencies: pip install -e ".[dev,openai,postgres,neo4j,redis]"
3. Run the test suite:

   python3 -m unittest discover -s tests -v
   

4. Submit a Pull Request!

📜 License

This project is licensed under the Apache License 2.0.