How abhi70802/agentic-trading-platform works
Northstar is a safe-first trading research platform: market data and supplied news are normalized into stored market and intelligence records, analyzed into hypotheses and calibrated risk decisions, then replayed or paper-executed under compliance and approval controls. Its connected runtime pipeline is optional and unset by default; live broker orders are disabled. The graph also preserves documented research capabilities and the local dashboard. Relationships for unsampled modules are included only where the README describes them; source excerpts provide additional implementation evidence.
Main entry points: runtime_pipeline.py Runtime pipeline · api.py API · store.py SQLite store · ai_agents.py Strategy agents
Components 30
Market data and research
- Provider adapters
angelone_adapter.py - Market normalization
marketdata.py - Feature snapshots
features.py - India reference data
india_market.py - Market breadth
breadth.py - India regime
india_regime.py - Historical similarity
historical_similarity.py - Backtesting and replay
backtesting.py
Intelligence and hypotheses
- News intelligence
intelligence.py - Strategy agents
ai_agents.py - LLM gateway
llm_gateway.py
Risk and decisions
- Probability calibration
probability_calibration.py - Decision engine
decision_engine.py - India cost model
india_economics.py - Portfolio sizing
portfolio.py - Compliance policies
compliance.py
Execution and governance
- Paper trading session
paper_trading.py - Paper execution
execution.py - Order management
oms.py - Controlled-live gate
controlled_live.py - Broker adapter
broker_adapter.py - Trade trace
trade_governance.py
API and persistence
- Runtime pipeline
runtime_pipeline.py - SQLite store
store.py - Trading memory
rag.py - API
api.py - Local dashboard
static/app.js
Other components
- Research user
- Data providers
- Broker
Connections 47
- Data providers → Provider adapters: supply data
README.md - Provider adapters → Market normalization: normalize ticks
README.md - Market normalization → Feature snapshots: derive snapshots
README.md - Feature snapshots → SQLite store: persist audit events
README.md - Market normalization → SQLite store: store candles
README.md - Feature snapshots → Market breadth: measure participation
README.md - Feature snapshots → India regime: classify market
README.md - Feature snapshots → Historical similarity: find analogues
README.md - Market normalization → Backtesting and replay: replay history
README.md - Data providers → News intelligence: supply articles
README.md - News intelligence → SQLite store: persist intelligence
README.md - News intelligence → Strategy agents: provide evidence
README.md - Strategy agents → LLM gateway: request hypotheses
README.md - LLM gateway → Strategy agents: return structured output
README.md - Strategy agents → SQLite store: record decisions
README.md - Probability calibration → Decision engine: supply probability
README.md - Decision engine → India cost model: calculate costs
decision_engine.py - Decision engine → Portfolio sizing: size and screen
decision_engine.py - Decision engine → Compliance policies: evaluate policy
decision_engine.py - Runtime pipeline → Strategy agents: generate hypotheses
README.md - Runtime pipeline → Probability calibration: calibrate history
README.md - Runtime pipeline → Decision engine: evaluate candidate
README.md - Runtime pipeline → Paper trading session: submit paper decision
README.md - Paper trading session → Paper execution: submit paper order
paper_trading.py - Paper execution → Compliance policies: check controls
execution.py - Paper execution → Portfolio sizing: check risk
execution.py - Paper execution → Order management: manage orders
- Runtime pipeline → SQLite store: write trade journal
README.md - Runtime pipeline → Trading memory: write trading memory
README.md - Trading memory → SQLite store: persist vectors
README.md - Runtime pipeline → Historical similarity: consult analogues
README.md - Runtime pipeline → India regime: consult regime
README.md - Runtime pipeline → Market breadth: consult breadth
README.md - Runtime pipeline → Trade trace: validate trade trace
README.md - Trade trace → SQLite store: store trace
README.md - Controlled-live gate → Broker adapter: submit gated order
controlled_live.py - Broker adapter → Broker: request broker action
- API → Market normalization: serve market data
- API → News intelligence: serve news endpoints
README.md - API → Strategy agents: serve hypotheses
README.md - API → Trading memory: serve memory search
README.md - API → Market breadth: serve breadth
README.md - API → India regime: serve regime
README.md - API → Historical similarity: serve similarity
README.md - API → SQLite store: read and write records
- Research user → Local dashboard: use dashboard
README.md - Local dashboard → API: call API