🌉 Technical Overview: ChiasmBridge & Isomorphic Projection

Release 1.0.0 (Binary Distribution) • Hardware Platform: NVIDIA RTX CUDA GPUs • Core Library: libchiasm.so

🌐 1. High-Level Overview

Multi-modal Large Language Models (LLMs) often use vision encoders (e.g. 7B Vision models) with hidden output dimensions of 3,584, while larger text LLMs (e.g. 24B LLMs) require input embedding dimensions of 5,120.

When attempting to pair a 7B Vision model with a 24B Text LLM, standard GGUF loaders fail due to dimension mismatch (3,584 ≠ 5,120).

ChiasmBridge (libchiasm.so) solves this by performing Isomorphic Orthogonal Subspace Projection directly in CUDA GPU memory. It maps the 3,584 visual channels losslessly into the 24B model's 5,120-dim space with zero feature distortion and zero training required.

🏛️ 2. Visual Architecture Diagram (Interactive Mermaid Render)

flowchart TD subgraph Input ["1. Visual Input"] Image["👁️ Image / Screen Pixels"] end subgraph VisionEncoder ["2. 7B Vision Encoder"] Encoder["📷 Vision Encoder\n(Outputs 3,584-dim Vision Tokens)"] end subgraph ChiasmBridge ["3. ChiasmBridge (libchiasm.so)"] Bridge["🌉 Isomorphic Subspace Projection\n(CUDA VRAM Translation 3,584 -> 5,120)"] end subgraph TargetLLM ["4. Target Cognitive LLM"] LLM["🐺 24B LLM (kalos:24b)\n(Receives 5,120-dim Visual Tokens)"] end Image --> Encoder Encoder -->|3,584-dim Tokens| Bridge Bridge -->|5,120-dim Tokens| LLM LLM --> Response["💬 Multimodal Visual Perception & Response"]

🔬 3. Component Breakdown

⚡ 1. 7B Vision Token Extraction

Extracts high-level visual features (colors, shapes, textures, objects) from raw image pixels, outputting 3,584-dimensional feature vectors per visual patch.

🌉 2. Isomorphic Subspace Projection (libchiasm.so)

Translates 3,584-dim vision tokens into the 24B model's 5,120-dim input space in CUDA VRAM. Preserves all 3,584 original visual channels 100% untouched and un-distorted.

🛠️ 4. Python Integration Example

from chiasm_bridge import SAMBridge, SAMBridgeConfig

config = SAMBridgeConfig(source_dim=3584, target_dim=5120)
bridge = SAMBridge(config)

translated_tokens = bridge(raw_vision_tokens)