ReCALL Lite — A Lightweight Memory Engine for LLMs

A simple, powerful, human-like memory layer for everyday LLM applications.

---

🚀 What Is ReCALL?

ReCALL is a human-inspired memory mechanism created by Project Genesis to address one of the biggest limitations of modern language models:

👉 LLMs forget everything once the context window ends.

ReCALL introduces a persistent, structured, cortex-like memory layer that allows AI systems to:

• store information across sessions
• retrieve memories based on meaning
• summarize and compress older memories
• merge similar information
• maintain stable long-term knowledge

It acts like a mini cognitive memory organ, letting LLMs recall facts the way humans do — not by scrolling back in history, but by semantic association.

ReCALL is one of Project Genesis’s core innnovations.

---

⚡ What Is ReCALL Lite?

ReCALL Lite is the public, simplified version of the original ReCALL memory system — fully rebuilt to:

• be extremely lightweight
• run on CPU
• work with any LLM (OpenAI, Gemini, HuggingFace, Ollama)
• require no complex setup
• give developers plug-and-play memory for their models

ReCALL Lite includes:

• long-term structured memory nodes
• semantic retrieval
• automatic summarization
• graph-based linking of related memory
• node merging to reduce redundancy
• multi-session persistence (save/load)

It’s simple enough for anyone to use, but powerful enough to transform how LLMs handle memory in real apps.

---

📌 Benchmark: ReCALL Lite vs Normal Context Window

Using 100 contexts and 100 questions on Gemma 3B, with 10-second intervals:

System	Accuracy
Normal Context Window	~20%
ReCALL Lite	~79%

This shows how even a lightweight memory layer can significantly boost factual retention.

Note: Due to limited compute resources, the Project Genesis team was not able to run large-scale benchmarks at this stage. This is an initial testimonial benchmark, and full standardized benchmarking will begin soon as Genesis scaling continues.

---

🎯 Why Project Genesis Built ReCALL Lite

Project Genesis created ReCALL Lite for one important mission:

👉 To open the door for researchers, engineers, and AI enthusiasts to experience what an AI-generated memory framework looks like — and build on it.

ReCALL Lite serves as:

• a teaching tool
• a research starting point
• a minimal reproduction of a much larger memory architecture
• a collaboration bridge between the open-source community and Project Genesis

ReCALL (the original) is a major AGI subsystem. ReCALL Lite allows the world to:

• experiment with the concepts
• understand the design philosophy
• explore long-term memory for LLMs
• contribute to future development

It’s not the full system — but it’s the first accessible step into cognitive memory engineering.

---

🧩 How ReCALL Lite Helps Everyday LLM Developers

Modern LLMs:

• lose older messages
• can’t persist information across sessions
• rely only on the current context window

ReCALL Lite upgrades any LLM into a memory-capable agent that can:

• remember user info
• recall facts days later
• summarize conversations into memory chunks
• retrieve relevant knowledge automatically
• persist knowledge between runs

Perfect for:

• AI companions
• chatbots
• productivity assistants
• long-term agents
• research tools
• custom RAG alternatives
• autonomous workflows

If you’ve ever wanted your model to “remember stuff like a human,” ReCALL Lite is built exactly for that.

---

👨‍💻 How Developers Can Use ReCALL Lite

You can install ReCALL Lite directly from PyPI:

pip install lite-recall

Once installed, here’s a simple example showing how ReCALL Lite stores and retrieves memory:

from lite_recall import recall_lite, LiteAgent, GeminiAPIConnector

# Initialize your LLM model (Gemini, OpenAI, HF, etc.)
model = GeminiAPIConnector(
    api_key="YOUR_KEY",                # Replace with your Gemini API key
    model_version="gemma-3n-e4b-it"    # Any supported Gemini/Gemma model
)

# Initialize the memory system
memory = recall_lite(
    memory_prefix="recall_lite",
    summarizer_model=model             # Optional: enables summary-based compression
)

# Create an agent that combines memory + LLM
agent = LiteAgent(memory, model)

# Store a memory
print(agent.process("My name is Krishna."))

# Ask something later → ReCALL Lite retrieves stored memory automatically
print(agent.process("What is my name?"))

mem.save()
print("\nMemory saved.")

ReCALL Lite handles:

• memory storage
• summaries
• retrieval
• merging
• forgetting
• context injection

Automatically.

---

🤝 ReCALL Lite & Project Genesis

ReCALL Lite is compatible with — and inspired by — the original ReCALL system built by Project Genesis.

Project Genesis aims to solve AGI’s memory pillar.

ReCALL is part of that pursuit, designed to bring:

• long-term retention
• structured memory
• evolving knowledge
• cortex-like organization
• multi-session continuity

ReCALL Lite is the open-source bridge toward that vision — giving the world a glimpse into the architectures Genesis is creating.

---

🌐 Why the Full ReCALL System Is Far More Powerful

ReCALL Lite is only the accessible public layer — a simplified version designed for everyday developers. The original ReCALL, created inside Project Genesis, is an advanced AGI-inspired memory architecture that goes far beyond what this lightweight version provides.

While Lite handles basic long-term storage, summarization, and semantic retrieval…

🔥 The full ReCALL system implements a much deeper cognitive model:

• multi-layered memory organization inspired by human cortex structures
• adaptive retention and forgetting tuned for long-term stability
• dynamic reinforcement signals that strengthen important memories
• protection layers for critical knowledge
• context-aware memory routing
• hierarchical linking that evolves with use

These capabilities position ReCALL as one of the first practical frameworks aimed at solving the memory pillar of AGI.

⭐ Why this matters

AGI needs four foundations:

1. Memory
2. Reasoning
3. Perception
4. Consciousness

ReCALL directly targets the first pillar — enabling an AI system to develop the kind of long-term, structured, continuously-evolving memory humans rely on.

It’s not just storage. It’s not just RAG. It’s cognitive memory engineering.

---

🤝 Open Call for Collaborations & Partnerships

ReCALL Lite is intentionally open-sourced so universities, engineers, startups, and independent researchers can:

• understand the direction of cognitive memory frameworks
• experiment with a working minimal version
• contribute ideas and enhancements
• propose research partnerships
• help shape the next generation of memory systems

The full ReCALL system, however, remains internal to Project Genesis and is reserved for:

• strategic partnerships
• research collaborations
• AGI labs
• institutions working on cognitive architectures
• organizations building autonomous agents

ReCALL Lite is the gateway. By working with it, developers can demonstrate interest, provide feedback, and potentially become part of the teams testing or extending the full ReCALL framework.

---

🌌 What Is Project Genesis?

Project Genesis is the world’s first autonomous AI research engine designed to:

• invent new frameworks
• build working systems
• evaluate its own designs
• evolve them automatically
• explore AI Frameworks, quantum models, and drug formulation discoveries

Genesis autonomously generated:

• ARF-OP
• MSGL
• CasualLite
• ReCALL
• NoToxic
• CogniMesh
• EvoSoul
• AIDRA

And more.

Project Genesis is not a chatbot. It is an autonomous scientific partner — built to explore the foundations of Scientific and Technical Discoveries.

ReCALL Lite is your invitation to join this mission.

---