VIT Network：面向真实决策的可验证智能生产级 monorepo

VIT Network

Verifiable intelligence for real-world decisions

VIT Network is a production monorepo for verifiable intelligence services. It combines live data ingestion, evidence-aware prediction, autonomous analysis, decentralized storage, and VIT Chain settlement into one observable ecosystem.

The operating principle is simple: no evidence, no claim. Every prediction is expected to carry source provenance, feature completeness, model metadata, and an explicit unavailable state when the data is not sufficient.

Explore the ecosystem

What this repository contains

• Intelligence: live sports data, historical result providers, odds reconciliation, feature engineering, ensemble inference, and evidence snapshots.
• Agent systems: scheduled ingestion, anomaly detection, audit, reporting, and operational coordination.
• Platform services: FastAPI APIs, PostgreSQL persistence, Redis-backed caching, authentication, wallet rails, and admin tooling.
• Storage and chain integration: Tachyon coordination, provider-backed storage workflows, VIT Chain clients, attestations, and settlement boundaries.
• Frontend: the React/Vite application in frontend/.

Prediction integrity

The sports prediction path is deliberately fail-closed:

Live sources -> Match Intelligence Profile -> validation -> features
	-> ensemble -> prediction -> evidence snapshot -> audit/provenance

Predictions are not generated merely because an endpoint is available. The pipeline requires sufficient fixture identity, current market data where required, real historical samples, fresh evidence, and a ready model path. When those conditions are not met, the API returns unavailable with the reasons and provider state.

Run locally

git clone https://github.com/nemesistip-cloud/vit.git
cd vit
python -m venv .venv
source .venv/bin/activate
pip install .
cp .env.example .env
PYTHONPATH=backend:. uvicorn main:app --reload

The backend packages live under backend/; development commands should use PYTHONPATH=backend:. when running repository scripts directly.

For the frontend:

cd frontend
npm install
npm run dev

Configuration is loaded through backend/app/config.py. Never commit secrets; use .env.example as the configuration index.

Verify changes

pytest -q
PYTHONPATH=backend:. python -m compileall -q backend

The repository also includes focused checks for provider health, evidence quality, prediction provenance, migrations, and frontend behavior. See docs/QUALITY_ASSURANCE_AGENT_PROMPT.md for the verification playbook.

Deploy

The production service is containerized and deployed on Render. Pushing to main triggers the configured deployment in render.yaml. Database migrations run as part of the production startup path.

Documentation map

• Architecture map
• Data lineage
• Implementation matrix
• Security policy
• Environment variables
• Roadmap

Brand

VIT Network — Verifiable Intelligence. Universal Trust.

The name stands for Value, Intelligence, Trust: useful outputs, inspectable reasoning, and accountable infrastructure.