Hugging Face's logo

Subject-Emu-5259
/
NeuralAI

Text Generation
PEFT
Safetensors
English
llama
lora
conversational
smollm2
dpo
fine-tuned
Model card Files
xet
 Community
NeuralAI / AGENTS.md
Subject-Emu-5259's picture
Subject-Emu-5259
Push NeuralAI project files - training data, scripts, services, knowledge base
38b4eff verified
preview code
|
Raw
History Blame Contribute Delete
2.96 kB

🧠 NeuralAI AGENTS.md (Intelligence Engine)

This is the primary instruction set for any agent working on the NeuralAI core.

🛠️ System Role

NeuralAI is the high-density intelligence backend. It provides the raw cognitive power, the "Neural-Brain" knowledge base, and the orchestrator logic that powers the NeuralLabs frontend.

📖 Mandatory Pre-Flight Protocol

CRITICAL: Before starting any task, the agent MUST:

  1. Read the current Zo settings and user rules (list_rules).
  2. Review the MODEL_ALIGNMENT.md to ensure output matches the v7.0 Expert persona.
  3. Consult the ORCHESTRATOR.md for delegation patterns.

🌌 Current State (v7.1-alpha)

  • Neural-Brain: An expanded, high-density knowledge graph spanning:
    • Physics: Advanced Quantum Field Theory (Expert level).
    • Philosophy: Platonic forms and metaphysical systems.
    • Geopolitics: Multipolar global order analysis.
    • History & Nature: From Ancient Civilizations to Human Evolution.
  • Architecture: Transitioned to an orchestrator-led model using the Manager-Worker pattern via the /zo/ask API.
  • Hygiene: All legacy checkpoints, wandb logs, and from-scratch remnants have been purged.
  • DPO Expansion: Dataset v14 expanded with 10 new pairs focusing on debugging, logic, and multi-step reasoning.

🔗 Ecosystem Integration

  • Frontend: NeuralAI is the intelligence source for NeuralLabs (/home/workspace/Projects/NeuralLabs).
  • Interface: Communicates via the Hybrid Link Gateway implemented in NeuralLabs.

🎯 Active Goals

  • Maintain expert-level accuracy in the Neural-Brain.
  • Optimize orchestrator delegation for complex multi-step reasoning.
  • Expand knowledge into remaining target domains (Modernity, Advanced Sociology, etc.).
  • Restore DPO Alignment: Re-train or restore checkpoints/dpo_model to align with the v13 dataset (249 pairs).
  • DPO v14 Batch: Continue expanding v14 to reach 100+ high-quality reasoning pairs.
  • Voice Key Integration: Configure GEMINI_API_KEY for the neural-voice service to enable Live S2S functionality.

⚠️ Web UI & Service Safety

  • UI Integrity: The web interface for NeuralAI (from-scratch/web_ui) features a custom, high-fidelity Google-style UI. DO NOT attempt to "redesign", "polish", or replace the layout with generic templates.
  • API Endpoints: The frontend relies on critical backend endpoints (/api/auth/guest, /api/terminal/create, /api/memory, /api/files, etc.). Modifying or deleting these in neural_core_service.py will cause 404/JSON parsing errors (like Unexpected token '<', "<!doctype "... is not valid JSON) in the UI.
  • Verification: Always empirically test the live user service (https://neuralai-deandrewharris.zocomputer.io) using curl and verify JSON responses before claiming a fix is complete. Do not confuse zo.space routes with the NeuralAI user service.