pranavpanchal1326
ARGUS-PRISM
Python

Pre-crime intelligence system for mule account detection. Catches the warming phase 72 hours before illicit funds arrive.| iDEA 2.0 | PS3

Last updated Jun 30, 2026
10
Stars
0
Forks
0
Issues
0
Stars/day
Attention Score
39
Language breakdown
Python 100.0%
โ–ธ Files click to expand
README

ARGUS-PRISM โ€” V3

Pre-crime Intelligence System for Mule Detection โ€” Union Bank of India, FIU.
"The hundred eyes see what others cannot. This time, everything they show is real."

V3 is the production-grade rebuild of the iDEA 2.0 hackathon system. It is built under three laws (see PRD-V3.md on the docs branch):

  • Contract first โ€” a feature does not exist until its endpoint is in
contracts/openapi.yaml. There is no USE_MOCK flag anywhere.
  • No fake anything โ€” every number on every screen is computed by the backend
from real (simulated) transaction data flowing through the real pipeline.
  • No secrets on glass โ€” no keys, tokens, or security material is ever rendered
in the UI. Integrity is communicated through status, never raw key material.

Repository layout (monorepo)

contracts/        OpenAPI 3.1 contract โ€” the single source of truth (shared, dual-approval)
backend/          FastAPI services: auth, alerts, accounts, cases, graph, autostr, โ€ฆ
  app/core/       config, response envelope, problem+json, security, logging
  app/engines/    the proven IP: WarmthScore, recruiter, taint, autostr
  app/simulator/  seeded, YAML-campaign scenario engine (the live transaction faucet)
ml-models/        pre-trained model artifacts + training/data-generation code
infra/            docker-compose, Dockerfiles, CI-adjacent ops
.github/          CI: ruff+mypy, contract tests, secret-grep tripwire

Branch ownership

  • Aditya โ€” backend-api, backend-pipeline, ml-models, auth, infra, CI, simulator, chatbot backend.
  • Pranav โ€” ui-shell, ui-live-feed, ui-chatbot, docs, design system, all screens.
  • Shared โ€” contracts/openapi.yaml (dual approval). main is the protected trunk.

Quick start (backend, local โ€” no Docker required)

cd backend
python -m venv .venv && . .venv/Scripts/activate   # Windows
pip install -e ".[dev]"
cp ../.env.example ../.env
uvicorn app.main:app --reload --port 8000

โ†’ http://localhost:8000/health and /docs

Without Docker the backend falls back to a local SQLite database and an in-memory event bus, so the API can be developed before standing up Postgres/Neo4j/Redis.

Full stack (Docker)

cp .env.example .env
docker compose -f infra/docker-compose.yml up --build

API โ†’ :8000 Postgres โ†’ :5432 Neo4j โ†’ :7474/:7687 Redis โ†’ :6379 Ollama โ†’ :11434

The contract workflow

  • Add the endpoint to contracts/openapi.yaml (PR to main).
  • Backend implements it; frontend generates its TS client from the same contract.
  • CI runs schemathesis against the running API to machine-check the implementation.
main only receives merges that pass CI. Conventional commits. No direct pushes to main.
๐Ÿ”— More in this category

ยฉ 2026 GitRepoTrend ยท pranavpanchal1326/ARGUS-PRISM ยท Updated daily from GitHub