mindlayer

Open-source, model-agnostic memory layer for LLMs.

Drop mindlayer into any LLM application to give it persistent, structured memory — no API key required, no infrastructure to run, no vendor lock-in.

import mindlayer

with mindlayer.MemCore() as mem:
    mem.add("My name is Alice. I prefer dark mode and I work in Python.")
    results = mem.search("programming preferences")
    for r in results:
        print(r.content)

---

Why mindlayer?

Most LLM apps lose context between sessions. Vector databases are heavy to set up. Existing memory libraries tie you to a specific LLM or cloud service.

mindlayer is:

• Zero config — SQLite storage, works out of the box
• Model agnostic — plug in any LLM or use the built-in Gemma extractor
• Pure library — no daemon, no background process, no ports
• Open source — MIT licensed, runs fully offline

---

Installation

pip install mindlayer-ai

With semantic vector search (downloads ~130MB embedding model on first use):

pip install "mindlayer-ai[vector]"

With LLM-powered extraction (downloads Gemma ~800MB on first use):

pip install "mindlayer-ai[llm]"

---

Architecture

mindlayer uses a 3-layer memory model inspired by human cognition:

Layer	Description	Promoted when
Working	Short-term, recent facts	accessed 3+ times
Episodic	Mid-term, frequently used facts	accessed 10+ times
Semantic	Long-term, core knowledge	stays permanently

5 core primitives

Primitive	What it does
Ingestion	Extracts discrete facts from raw text
Consolidation	Promotes memories across layers
Decay	Reduces scores on idle memories, prunes stale ones
Retrieval	Fetches relevant memories, boosts access count
Conflict resolution	Deduplicates near-duplicate memories

---

Usage

Default (rule-based extractor, no LLM needed)

import mindlayer

mem = mindlayer.MemCore()
mem.add("I am a Python developer. I love open source.")
results = mem.search("developer")

With semantic vector search (best recall)

# pip install "mindlayer[vector]"
mem = mindlayer.MemCore(use_vector=True)
mem.add("I prefer concise explanations and dislike verbose output.")
results = mem.search("communication style")  # matches semantically, not just by keyword

With Gemma LLM extractor (best quality)

mem = mindlayer.MemCore(use_llm=True)
# Downloads gemma-3-1b-it-Q4_K_M.gguf (~800MB) on first run
mem.add("Long conversation text with lots of context...")

Bring your own LLM extractor

from mindlayer.extractors.base import BaseExtractor

class MyExtractor(BaseExtractor):
    def extract(self, text: str) -> list[str]:
        # call OpenAI, Anthropic, Ollama, anything
        return ["fact 1", "fact 2"]

mem = mindlayer.MemCore(extractor=MyExtractor())

Custom storage backend

from mindlayer.storage.base import BaseStorage

class PostgresStorage(BaseStorage):
    # implement the interface
    ...

mem = mindlayer.MemCore(storage=PostgresStorage())

Memory maintenance

mem.consolidate()   # promote memories based on access patterns
mem.decay()         # decay and prune stale memories

---

Roadmap

• Vector similarity search (FAISS / sqlite-vec)
• Async support
• PostgreSQL storage backend
• LLM-based conflict resolution
• REST API server mode
• JavaScript / TypeScript port

---

Contributing

Contributions are welcome. Please open an issue before submitting large PRs.

git clone https://github.com/Genious07/mindlayer
cd mindlayer
pip install -e ".[dev,vector]"
pytest

---

License

MIT