Skip to content

vicsanity623/AxiomEngine

 
 

Repository files navigation

Axiom Engine

A decentralized, anti-LLM grounding engine
Builds a deterministic lexical mesh + knowledge graph from RSS streams — no probabilities, no GPUs, no black boxes.

License: PPL Python 3.13+

Core Concept

Axiom nodes continuously:

  1. Discover trending topics via public RSS (no tracking APIs)
  2. Extract clean article content (bot-resistant scraper)
  3. Verify & extract facts only (spaCy NLP + multi-source corroboration rule)
  4. Link entities → build weighted knowledge graph
  5. Shred facts into linguistic atoms & synapses → open SQL lexical mesh

Result: queryable, fully inspectable symbolic "brain" of corroborated facts.
Contradictions are flagged. Single-source claims stay uncorroborated.

Strongly anti-probabilistic AI — everything is deterministic and auditable.

Architecture (Heartbeat Cycle)

flowchart TD
    Start([Heartbeat]) --> Discovery[Zeitgeist: RSS scan]
    Discovery --> Extraction[Pathfinder: scrape content]
    Extraction --> Validation{Crucible: NLP + fact extraction}
    Validation -->|bias/speculation| Noise[Filtered out]
    Validation -->|fact candidate| Corroboration[≥2 independent sources?]
    Corroboration --> Linking[Synthesizer: entity linking → Knowledge Graph]
    Linking --> Reflection[Lexical Mesh: shred → synapses]
    Reflection --> Storage[(SQL Ledger: facts + brain)]
    Storage --> Sleep[Sleep → next cycle]
Loading

Quick Start

git clone https://github.com/vicsanity623/AxiomEngine.git
cd AxiomEngine

Install dependencies

uv sync

Optional: better NLP accuracy (spaCy model)

uv run python -m spacy download en_core_web_sm

Start a node using:

export PORT=8009
uv run python -m src.node

        // or using python:

export PORT=8009
python -m src.node 

Build standalone binary (macOS .dmg / Windows .exe)

python build_standalone.py 

Inspect & Visualize

python view_ledger.py
python view_ledger.py --stats
python view_ledger.py --brain       # top synapses
python visualize_graph.py           # visual HTML graph

Current Status (2026)

•	RSS discovery + scraping    
        # can be modified to DISCOVER anything
•	Fact extraction & corroboration     
        # can be modified to extract & corroborate anything
•	Basic knowledge graph linking
        # makes connections for verification/ disputes/ corroborations
•	Lexical mesh (linguistic atoms & synapses)      
        # builds a ledger of "language" (voice)
•	SQL ledger persistence      
        # databse lives in the "Block" network and syncs accross
•	Mobile/web UI (Universal Terminal)      
        # can interact directly with axiom from anywhere
•	P2P gossip + sync (full mesh)       
        # nodes remain synced, reboot safe.
•	Reputation & DAO voting         
        # DAO not yet implemented-brarebones TODO

License Peer Production License (PPL)

A lightweight, inspectable alternative to probabilistic language models.

Packages

 
 
 

Contributors

Languages

  • Python 84.7%
  • HTML 13.5%
  • Shell 1.6%
  • JavaScript 0.2%