A comprehensive algorithmic trading system for Indian F&O markets (NSE) with AI-powered multi-agent orchestration, database-driven prompt configuration, risk management, and real-time pipeline visualization.

Conservative trading system focused on:

• Multi-Agent AI Pipeline — 5 coordinated agents (Research → Signal → Risk → Execution → Monitor) with real-time status tracking
• Database-Driven Prompt Config — All agent system messages and prompt templates stored in PostgreSQL with versioning, live editing, and full audit history
• Risk Management — Circuit breakers, position sizing, daily/weekly loss limits
• Conservative Strategy — Capital preservation with controlled growth
• Backtesting — Multi-strategy comparison with historical data
• Paper Trading — Test strategies without real capital

┌──────────────────────────────────────────────────────────────────┐
│                     Agent Coordinator                             │
│       (Orchestrates pipeline with real-time status tracking)      │
│                                                                   │
│  ┌───────────┐  ┌───────────┐  ┌──────────┐  ┌──────────────┐  │
│  │ Research  │→ │  Signal   │→ │   Risk   │→ │  Execution   │  │
│  │  Agent    │  │  Agent    │  │  Agent   │  │  (Order Svc) │  │
│  └───────────┘  └───────────┘  └──────────┘  └──────────────┘  │
│       ↑                                             ↓            │
│  ┌───────────┐                              ┌──────────────┐    │
│  │Categorical│                              │   Monitor    │    │
│  │  Agent    │                              │   Agent      │    │
│  └───────────┘                              └──────────────┘    │
│                                                                   │
│  AgentStatusTracker: IDLE → RUNNING → COMPLETED/ERROR per agent  │
└──────────────────────────────────────────────────────────────────┘
              ↓                            ↓
┌──────────────────────┐      ┌──────────────────────────┐
│   Strategy Engine    │      │   Risk Management        │
│  - Trend Following   │      │  - Circuit Breaker       │
│  - Mean Reversion    │      │  - Position Sizing       │
│  - Breakout          │      │  - Daily/Weekly Limits   │
└──────────────────────┘      └──────────────────────────┘
              ↓                            ↓
┌──────────────────────────────────────────────────────────┐
│              Order Execution Service                      │
│           (Paper Trading / Live Trading)                  │
└──────────────────────────────────────────────────────────┘
              ↓
┌──────────────────────────────────────────────────────────┐
│          MCP Market Data Service (Groww API)              │
└──────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────┐
│        agent_prompt_configs (DB)             │
│  ┌────────────────────────────────────────┐ │
│  │ agent_name │ config_key │ config_value │ │
│  │ RISK_AGENT │ system_msg │ "You are..." │ │
│  │ RISK_AGENT │ prompt_tpl │ "{signal}..."│ │
│  └────────────────────────────────────────┘ │
│  - Versioned (v1, v2, v3...)                │
│  - Active/inactive flag                      │
│  - ConcurrentHashMap in-memory cache         │
└─────────────────────────────────────────────┘
              ↓ on update
┌─────────────────────────────────────────────┐
│      agent_prompt_config_audit (DB)          │
│  - old_value → new_value                     │
│  - changed_by, change_reason                 │
│  - Full audit trail of all changes           │
└─────────────────────────────────────────────┘

• Framework: Spring Boot 3.4.2
• AI: Spring AI 1.0.0-M5 (OpenAI GPT-4o integration)
• Database: PostgreSQL
• Build: Gradle 8.12
• Java: 21

# Java 21
export JAVA_HOME=/opt/homebrew/Cellar/openjdk@21/21.0.10/libexec/openjdk.jdk/Contents/Home

# PostgreSQL
brew install postgresql@17
brew services start postgresql@17

Edit src/main/resources/application.properties:

# OpenAI API Key (for AI-powered analysis)
spring.ai.openai.api-key=your-openai-api-key-here

# Trading Configuration
trading.mode=semi-auto
trading.capital=50000
trading.instruments=NIFTY,BANKNIFTY

# Risk Management
trading.risk.daily-loss-limit=1500
trading.risk.weekly-loss-limit=3500

./gradlew bootRun
# Server starts on port 8080

On first startup, the AgentPromptSeeder automatically seeds all 5 agent configurations from classpath defaults into the database.

• AgentStatusTracker — In-memory ConcurrentHashMap tracking each agent's state (IDLE/RUNNING/COMPLETED/ERROR), run duration, result summary, and error details
• AgentCoordinator — Wraps each pipeline step with status tracking calls (markRunning() → markCompleted()/markError())
• Real-time visibility — Pipeline state exposed via REST API, consumed by dashboard for live visualization

• AgentPromptConfig entity — Stores system_message and prompt_template per agent with versioning
• AgentPromptConfigAudit entity — Full change history with old/new values, change reason, and who made the change
• AgentPromptConfigService — Cached config access (ConcurrentHashMap) with @Transactional updates that deactivate old versions and create new ones
• AgentPromptSeeder — Seeds defaults from classpath .txt files on first startup
• All 5 agents modified — Each agent uses getSystemMessage() and getPromptTemplate() methods that check DB first, fallback to hardcoded defaults

• Market sentiment analysis
• Volatility assessment
• Trading recommendations
• Context-aware analysis (includes current holdings)

Comprehensive equity analysis across:

• Market Cap (Large/Mid/Small)
• Industries (Banking, IT, Pharma, etc.)
• Tech Sector
• Recent IPOs

Conservative strategy:

• 60-70% Large Cap
• 20-30% Mid Cap
• 0-10% Small Cap

• Circuit Breaker with daily/weekly limits
• Position Sizing based on risk
• Emergency Stop functionality
• Real-time monitoring

Multi-agent coordination:

1. Research Agent → Market analysis
2. Signal Agent → Entry/exit signals with confidence scoring
3. Risk Agent → Hard rule checks + AI-enhanced assessment
4. Execution → Order placement (paper or live)
5. Monitor Agent → Position monitoring and exit signals

• Multiple strategies (Trend Following, Mean Reversion, Breakout)
• Historical performance analysis
• Multi-strategy comparison

• Instrument sync from Groww API with DB persistence
• Historical candle data storage with DB-first lookup
• Rate-limited API calls with exponential backoff

• GET /api/dashboard/overview — Main metrics
• GET /api/dashboard/positions — Current positions
• GET /api/dashboard/trades — Trade history
• GET /api/dashboard/metrics — Trading metrics

• POST /api/dashboard/agents/research — Market research
• POST /api/dashboard/agents/categorical-analysis — Stock analysis
• POST /api/dashboard/agents/pipeline — Run full trading pipeline
• GET /api/dashboard/agents/decisions — Recent agent decisions

• GET /api/dashboard/agents/status — Real-time pipeline state (all agent statuses)

• GET /api/dashboard/agents/configs — All active configs
• GET /api/dashboard/agents/configs/{agentName} — Configs for a specific agent
• PUT /api/dashboard/agents/configs/{agentName}/{configKey} — Update config (body: {value, reason, changedBy})
• GET /api/dashboard/agents/configs/audit — Recent 50 audit entries
• GET /api/dashboard/agents/configs/audit/{agentName} — Audit trail for specific agent

• POST /api/dashboard/emergency-stop — Emergency halt
• POST /api/dashboard/resume-trading — Resume trading
• POST /api/dashboard/mode/{mode} — Set mode (semi-auto/full-auto)

• POST /api/dashboard/backtest/{symbol}?days=180&capital=50000&strategy=TREND_FOLLOWING
• POST /api/dashboard/backtest/compare/{symbol}?days=180&capital=50000 — Multi-strategy comparison

• GET /api/dashboard/instruments — Tradable instruments
• GET /api/dashboard/instruments/search?query=NIFTY — Search instruments
• POST /api/dashboard/instruments/sync — Sync from Groww API

src/main/java/com/navneet/fnotrading/
├── agent/
│   ├── config/              # AgentPromptConfigService, AgentPromptSeeder
│   ├── coordinator/         # AgentCoordinator (pipeline orchestration)
│   ├── monitor/             # TradeMonitorAgent
│   ├── research/            # MarketResearchAgent, CategoricalAnalysisAgent
│   ├── risk/                # RiskAssessmentAgent
│   ├── signal/              # SignalGenerationAgent
│   └── status/              # AgentStatusTracker (pipeline state tracking)
├── backtest/                # Backtesting engine
├── config/                  # TradingConfig, RiskConfig, McpClientConfig
├── dashboard/
│   ├── controller/          # DashboardController (REST endpoints)
│   └── service/             # MetricsService, ReportService, PerformanceTrackingService
├── execution/               # OrderExecutionService, PaperTradingService, PositionManager
├── market/                  # TechnicalIndicatorService, market models
├── mcp/                     # Groww API integration (McpMarketDataService, McpOrderService)
├── persistence/
│   ├── entity/              # Trade, Signal, Instrument, AgentPromptConfig, AgentDecision, etc.
│   └── repository/          # JPA repositories
├── risk/                    # RiskManagementService, CircuitBreakerService, PositionSizingService
└── strategy/                # StrategyEngine (Trend, MeanReversion, Breakout)

• ✅ Core trading engine
• ✅ Multi-agent system with 5 AI agents
• ✅ Agent pipeline status tracking (real-time)
• ✅ Database-driven prompt configuration with versioning & audit
• ✅ AI-powered research & categorical analysis
• ✅ Risk management & circuit breaker
• ✅ Paper trading
• ✅ Backtesting (single & multi-strategy)
• ✅ Historical candle data persistence
• ✅ Instrument sync & management
• ✅ Dashboard REST API (30+ endpoints)
• ⏳ Live trading (testing in paper mode)

• fno-dashboard — React UI for monitoring, control, pipeline visualization, and prompt management
• stock-exchange-trade — Groww API wrapper (MCP server)

Version: 1.1.0 Author: Navneet Prabhakar Last Updated: 2026-02-12