Upload README.md

README.md

@@ -7,175 +7,6 @@ tags:
   - local-llm
 license: apache-2.0
 ---
-
-# Lambda Mindlink Memotron
-
-A three-hemisphere cognitive AI architecture running entirely locally via llama-cpp-python.
-
-## Architecture
-
-*   **Logic AI** — Left hemisphere: analytical, structured reasoning
-*   **Muse AI** — Right hemisphere: creative, intuitive synthesis  
-*   **Lambda Mind** — Stem brain: vector synthesis of both hemispheres
-*   **Condensatron** — Post-level memory compression (surprise extraction)
-*   **Fractaltron** — Second-order fractal compression of memory capsules
-
-## Models Used
-
-Download the GGUFs from Hugging Face and place in `ai-models/`:
-
-*   [google/gemma-4-E2B-it](https://huggingface.co/...) — development/debug
-*   [Qwen/Qwen3-...](https://huggingface.co/...) — optional swap-in
-
-## Install
-
-```bash
-pip install llama-cpp-python --extra-index-url \
-  https://abetlen.github.io/llama-cpp-python/whl/cu121  # adjust for your CUDA version
-pip install -r requirements.txt
-```
-
-## Run
-
-```bash
-python main.py
-```
-
-## Memory Architecture of the KV cache
-
-1. If at startup `N_MEMORY_CAPSULES_TO_LOAD > 0`, then the Memory Capsule and the response are appended to garden["M"], already being Memory Capsules.  
-2. The user chat is appended to the garden["Z"] being uncompressed.  
-3. condensatron: If the garden["n_tok_tot"]["Z"] > CONDENSATRON_THRESHOLD_LEVEL`: start condensatron_cycle.  
-4. fractaltron: If the `HEMISPHERES[brain_type]["loader"]["n_ctx"] - (CONDENSATRON_THRESHOLD_LEVEL + RESERVED_FOR_PROMPT)`:  
-    1. Check and process garden["Z"] posts with condesatron and append to garden["M"]  
-    2. Now process part of garden["M"] with the fractaltron_cycle  
-
-```
-[Instructions] → [garden["M"] memory capsules] → [garden["Z"] post-level cache]
-                        ↑ Fractaltron                    ↑ Condensatron
-```
-
----
-
-## Setup
-
-```bash
-# --- Linux RedHat Fedora ------ #
-# First you must install the c++ compiler
-# to compile the llama-cpp-python package
-sudo dnf install -y cmake gcc-c++ python3-devel
-
-#  Navigate to the project folder
-cd lambda-mindlink-memotron
-# Create the virtual environment
-python -m venv .venv
-# Activate the virtual environment
-source .venv/bin/activate
-
-# ------ llama-cpp-python ------ #
-# ROCm AMD Ryzen iGPU support version
-CMAKE_ARGS="-DGGML_HIPBLAS=on" pip install llama-cpp-python
-# CPU-only version
-pip install llama-cpp-python
-
-# Nvidia CUDA version (not tested)
-pip install llama-cpp-python --extra-index-url \
-  https://abetlen.github.io/llama-cpp-python/whl/cu121
-
-# Uninstall any previous failed attempts
-pip uninstall llama-cpp-python -y
-# ------------------------------ #
-
-# Install the project requirements
-pip install -r requirements.txt
-```
-
-## Run Lambda
-
-```bash
-# Open your Terminal
-# Navigate to the project folder
-cd lambda-mindlink-memotron
-
-# Linux: Activate the virtual environment
-source .venv/bin/activate
-
-# Start Lambda
-python main.py
-```
-
----
-
-## Memory: Startup
-
-> Load the basic vector synthesis memory into the garden["M"]
-
-Note: this requires to compute each Memory Capsule one-by-one for the KV cache.
-
-```bash
-# Find and set the variable in config.py
-# Default settings (~10k tokens)
-N_MEMORY_CAPSULES_TO_LOAD: int = 4
-```
-
----
-
-## RAM estimate (three 4B Q6_K instances)
-
-| Component | ~Size |
-|-----------|-------|
-| 3× model weights | ~10.5 GB |
-| 3× KV caches (6144 ctx) | ~1.0 GB |
-| OS + overhead | ~2.0 GB |
-| **Total** | **~13.5 GB** |
-
----
-
-## Folder structure
-
-```
-lambda-mindlink-memotron/
-├── ai-models/
-|   └── CoT_/
-├── db/
-├── model-readme/
-├── prompt/
-├── main.py
-├── config.py
-├── requirements.txt
-└── README.md
-```
-
----
-
-## Future milestones (from the flowchart)
-
-| Milestone | Description |
-|-----------|-------------|
-| **X-factor (Awareness)** | News / web feed input routed to Garden['X'] |
-| **Y-factor (Consciousness)** | Self-reflection loop — the "I think therefore I am" |
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
 
 # Lambda Mindlink Memotron