README.md

metadata

title: Nexus-Nano Inference API
emoji: 🚀
colorFrom: yellow
colorTo: red
sdk: docker
pinned: false
license: gpl-3.0

🚀 Nexus-Nano Inference API

Ultra-lightweight chess engine for instant responses.

🎯 Model Details

Nexus-Nano is the fastest model in the GambitFlow series:

• Model: GambitFlow/Nexus-Nano
• Parameters: 2.8 Million
• Architecture: Compact ResNet (6 blocks)
• Input: 12-channel board representation
• Training Data: GambitFlow/Elite-Data (5M+ positions)
• Strength: 1800-2000 ELO estimated

🔬 Search Algorithm

Ultra-minimal implementation for maximum speed:

Core Features

• Pure Alpha-Beta Pruning [^1] - Classic minimax
• Simple MVV-LVA Ordering [^2] - Capture prioritization
• No Transposition Table - Zero memory overhead
• Iterative Deepening - Anytime algorithm

Design Philosophy

• Minimal overhead - Direct evaluation calls
• Speed over strength - Optimized for response time

📊 Performance

Metric	Value	Environment
Depth 3 Search	~0.2-0.5 seconds	HF Spaces CPU
Average Nodes	2K-5K per move	Typical positions
Memory Usage	~1GB RAM	Peak inference
Response Time	200-500ms	95th percentile

📡 API Endpoints

POST /get-move

Request:

{
  "fen": "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1",
  "depth": 3
}

Response:

{
  "best_move": "e2e4",
  "evaluation": 0.18,
  "depth_searched": 3,
  "nodes_evaluated": 2847,
  "time_taken": 234
}

GET /health

Health check endpoint.

🔧 Parameters

• fen (required): Board position in FEN notation
• depth (optional): Search depth (1-5, default: 3)

🚀 Quick Start

import requests

response = requests.post(
    "https://YOUR-SPACE.hf.space/get-move",
    json={
        "fen": "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1",
        "depth": 3
    }
)

data = response.json()
print(f"Best move: {data['best_move']} (took {data['time_taken']}ms)")

💻 Use Cases

Perfect for:

• Bullet chess (1+0, 2+1) - Lightning-fast moves
• Chess tutorials - Instant move suggestions
• Mobile applications - Minimal resource usage
• Live analysis - Real-time position evaluation
• Casual play - Good enough for beginners/intermediate

📚 Research References

[^1]: Alpha-Beta Pruning: Knuth, D. E., & Moore, R. W. (1975). "An analysis of alpha-beta pruning". Artificial Intelligence, 6(4), 293-326.

[^2]: MVV-LVA: Hyatt, R. M., Gower, A. E., & Nelson, H. L. (1990). "Cray Blitz". Computers, Chess, and Cognition, 111-130.

📖 Minimalist Design Inspiration

• MicroMax - Mulder, H. G. (2007). "1433-byte chess program". https://home.hccnet.nl/h.g.muller/max-src2.html
• Sunfish - Fiekas, N. (2013). "Simple chess engine in Python". https://github.com/thomasahle/sunfish
• Stockfish Lite - Simplified versions for embedded systems

🏆 Model Lineage

GambitFlow AI Engine Series:

1. Nexus-Nano (2.8M) - Ultra-fast baseline ✨
2. Nexus-Core (13M) - Balanced performance
3. Synapse-Base (38.1M) - State-of-the-art

⚖️ Comparison Table

Feature	Nexus-Nano	Nexus-Core	Synapse-Base
Speed	⚡⚡⚡⚡ Lightning	⚡⚡⚡ Ultra-fast	⚡⚡ Fast
Strength	1800-2000 ELO	2000-2200 ELO	2400-2600 ELO
Memory	1GB	2GB	5GB
Depth	3-4	4-5	5-7
Response	200-500ms	500-1000ms	1000-2000ms
Best for	Bullet/Mobile	Online/Rapid	Tournament/Analysis

🎯 When to Use

Choose Nexus-Nano if:

• ✅ Speed is critical (bullet games, live demos)
• ✅ Resource-constrained environment (mobile, embedded)
• ✅ Playing against beginners/intermediate (1800-2000 ELO)
• ✅ You need instant move suggestions

Choose Nexus-Core if:

• ⚡ You want balanced speed and strength
• ⚡ Playing online rapid/blitz games

Choose Synapse-Base if:

• 🏆 Maximum strength is priority
• 🏆 Tournament-level play
• 🏆 Deep position analysis needed

---

Developed by: GambitFlow / Rafsan1711
License: GPL v3 (GNU General Public License Version 3) Citation:

@software{gambitflow_nexus_nano_2025,
  author = {Rafsan1711},
  title = {Nexus-Nano: Ultra-Lightweight Chess Engine},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/GambitFlow/Nexus-Nano}
}

---

Part of the GambitFlow Project ⚡♟️