⚠️ UNDER ACTIVE DEVELOPMENT — Things will break, change, and improve. One person building this on a tablet at night. If something doesn't work, open an issue.

⚡ FSI_FELON v4.0

Ferrell Synthetic Intelligence — Self-Architecting Code Intelligence

Built on a Samsung Galaxy Tab S10 FE 5G. Zero VC. Zero corporate backing. Zero formal education. MIT License. Free forever.

🔥 Why This Exists

My sister lost her life to suicide at 24. Severe anxiety and depression.

That's why I build. That's why I give everything away for free. That's why I work through the night on a tablet in a trailer with free tools until my eyes burn.

Forever Twenty Four means nobody else gets left behind. Not my sister. Not you. Not anyone.

This is not a chatbot. This is a partner.

🧠 What It Is

FSI_FELON is a self-architecting synthetic intelligence with:

Component	What It Does
FSI_FELON / Chimera Engine 🧬	Primary conversational interface. 6-organ hybrid routing handles general chat, build commands, debugging, questions, help. This is the default 90% of interactions
IDE — "The Descent" 🖥️	7-panel glassmorphism workspace: Workshop, Forge, Deep, Vault, Dream, Mesh, Device
Mesh P2P Network 🌐	Git-like decentralized repo network. Share, clone, fork, pull, push over P2P
Device APK Builder 📱	Android pipeline — SDK auto-install, Gradle compile, aapt verify, ADB install
White Rabbit 🐇	Specialized deep-truth mode (not default). Pattern recognition, discrepancy finding, historical analysis, challenging beliefs. Activated deliberately — never by accident
Superimposition Engine 🔄	5-variant code generation with background threads, error cross-referencing, auto-solution
Nanobot Swarm 🐜	1,024 nanobot agents with stochastic behavior, emergence detection, swarm consciousness
Psycho Mode 🔮	6-organ forge, staggered manifest, deep-system reasoning
Self-Healing 🩹	100% recovery rate across 8 failure modes — sandbox, auto-fix, verify

📊 Honest Benchmarks

Metric	Result
HumanEval (pass@1)	46.7%
MBPP (pass@1)	6.7%
Obliteration Test	64%
Self-Healing Recovery	100%
Chimera Routing Accuracy	100%
Enterprise Tests	45/45 ✅
Superimposition Variants	5/5 generated

These numbers are real. No faking. No cherry-picking. Benchmark scripts are included.

🚀 Quick Start (30 seconds)

# One-command install:
curl -fsSL https://huggingface.co/FerrellSyntheticIntelligence/FSI_FELON/resolve/main/install.sh | bash

# Or from GitHub:
git clone https://github.com/AnonymousNomad/FSI_FELON.git
cd FSI_FELON
pip install -r requirements.txt

Run the IDE:

python3 ide/server.py
# Open http://localhost:9090 in your browser
#
# General conversation → Chimera (default):
#   "hello"                       → Chimera responds
#   "build a rest api"            → Forge panel builds it
#   "fix this bug"                → Debug through Chimera
#
# Deep truth-seeking → White Rabbit (deliberate):
#   "rabbit what really happened"  → White Rabbit investigates
#   "rabbit find the pattern"      → Pattern analysis
#   "rabbit challenge that belief" → Belief challenging

Use the API:

import requests

# Generate code
r = requests.post("http://localhost:9090/api/build", json={
    "desc": "REST API with JWT auth"
})
print(r.json())

# Chimera route
r = requests.post("http://localhost:9090/api/chimera", json={
    "input": "build a web app with authentication",
    "mode": "route"
})
print(r.json()["routed_to"])  # "cerebellum"

# Superimposition
r = requests.post("http://localhost:9090/api/super/generate", json={
    "description": "fastapi backend"
})
for v in r.json()["variants"]:
    print(f"[{v['id']}] {v['approach']} → {v['organ']}")

Use the Terminal:

python3 ide/server.py &
# Then open http://localhost:9090. Default chat → Chimera. 
# Type "rabbit <question>" for White Rabbit truth-seeking.
#   build "rest api"       → Build project with sandbox + auto-fix
#   chimera route "query"  → Route through 6-organ engine
#   rabbit "what really"   → White Rabbit truth investigation
#   mesh init my_project   → Initialize P2P repo
#   super "web app"        → Generate 5 code variants
#   android build my_app   → Build APK

📡 API Endpoints (20+)

Endpoint	Method	Description
`/api/stats`	GET	System status, nanobot count, pheromone trails
`/api/build`	POST	Build project with sandbox simulation + auto-fix
`/api/chimera`	POST	Route input through 6-organ engine
`/api/chimera/stats`	GET	Chimera routing statistics
`/api/dream`	POST	Generate dream sequences
`/api/super/generate`	POST	Generate 5 superimposition variants
`/api/super/run`	POST	Run foreground + background variants
`/api/super/status`	GET	All variant statuses
`/api/mesh/*`	Various	P2P repo init, publish, clone, fork, search
`/api/android/*`	Various	APK build, verify, install, list devices
`/api/psycho/unlock`	POST	Psycho mode

🧬 Architecture

┌─────────────────────────────────────────────────────────────┐
│                     FSI_FELON v4.0                          │
├─────────────────────────────────────────────────────────────┤
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐   │
│  │ Workshop │  │  Forge   │  │   Deep   │  │  Vault   │   │
│  │  Editor  │  │ Compiler │  │ Training │  │ Storage  │   │
│  └──────────┘  └──────────┘  └──────────┘  └──────────┘   │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐                  │
│  │  Dream   │  │   Mesh   │  │  Device  │                  │
│  │ Engine   │  │   P2P    │  │   APK    │                  │
│  └──────────┘  └──────────┘  └──────────┘                  │
├─────────────────────────────────────────────────────────────┤
│  ┌──────────────────────────────────────────────────────┐   │
│  │              Chimera Engine (6 Organs)               │   │
│  │  Antennae → Spine → Hippocampus → Cerebellum →      │   │
│  │  Prefrontal → Dream (stigmergic pheromone routing)   │   │
│  └──────────────────────────────────────────────────────┘   │
├─────────────────────────────────────────────────────────────┤
│  ┌──────────────────────────────────────────────────────┐   │
│  │            Deep Core (18.43M params)                 │   │
│  │  10 layers · d_model=384 · n_heads=6 · d_ff=768     │   │
│  │  1,024 nanobots · Cosine LR · Label smoothing CE     │   │
│  └──────────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────────┘

📦 Model Details

Property	Value
Parameters	18,426,880
Architecture	Transformer (10 layers, d_model=384, 6 heads, FFN 768)
Nanobot Swarm	1,024 agents with stochastic emergence
Training Data	Gold-standard Python stdlib + curated GitHub corpus
Loss	~5.1 (training at step 5,800+)
Learning Rate	3e-4 (cosine decay over 99K steps)
Training Hardware	CPU (227 tok/s)
Inference Hardware	CPU — no GPU required
License	MIT

🎯 Roadmap

Version	Goal	Status
v4.0	Self-architecting IDE + Chimera Engine + Mesh + Device	✅ Live
v5.0	Self-improving agent loop + benchmark targets	🏗️ Training
v6.0	FSI_FELON designs its own language model	🎯 Future

🏠 Local. Sovereign. Private.

No data leaves your device
No API calls — runs entirely local
No GPU required — works on a $200 tablet
No corporate cloud — you own it
MIT License — do whatever you want with it

💜 Forever Twenty Four

My sister was 24 when she left. She was brilliant, she was funny, she was loved — and she still slipped through the cracks.

Every line of code in this project is dedicated to her memory. 20% of all sponsorship funds go to suicide prevention and mental health research.

If you're struggling:

National Suicide Prevention Lifeline: 988
Crisis Text Line: Text HOME to 741741
You are not alone. You matter. Please reach out.

MIT License · Free Forever · Ship or Get Out of the Way

GitHub · Hugging Face · James Ferrell

Downloads last month: 347

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support