Fincept Terminal

AI Quant Lab is a quantitative research lab designed for automated strategy development. It integrates Microsoft's Qlib quantitative investment platform with RDAgent (Research & Development Agent) to automate hypothesis generation, backtesting, and strategy optimization.

Core Workflow

1. Market Data: Input stock prices and fundamentals.
2. Data Processing: Clean, normalize, and extract features.
3. RDAgent: Generate trading hypotheses using AI.
4. Feature Engineering: Create alpha factors.
5. Model Training: Train machine learning models.
6. Strategy: Generate trading signals.
7. Backtest: Perform historical simulations.
8. Portfolio Optimization: Determine optimal weights.
9. Evaluation: Calculate performance metrics.
10. Knowledge Base: Store results for future research.

Alpha Arena is a native C++ replay-faithful competition runtime integrated into the Fincept Terminal. It allows developers to pit multiple LLM agents against each other in a simulated or live crypto-perpetual universe. The system operates on a tick-by-tick basis with cryptographically auditable replay capabilities.

Key architectural characteristics:

• Engine: The core logic resides in a C++ engine. The Python component is a minimal (approx. 200-line) subprocess used solely for LLM calls.
• Determinism: Uses a ContextBuilder to ensure deterministic, byte-equal prompts across agents.
• Execution: Employs a ModelDispatcher to fan out ticks to multiple Python subprocesses, and a RiskEngine to validate actions (accept/reject/amend).
• Venues: Supports both HyperliquidVenue (live trading via REST/WS and EIP-712 signing) and PaperVenue (a high-fidelity simulator).

The Python Scripts Library provides the backend intelligence for Fincept Terminal, including data integrations, financial analytics, and AI agents.

Directory Structure

• agents/: AI agents for trading and geopolitical analysis.
• agno_trading/: Agno trading system framework.
• ai_quant_lab/: Quantitative research lab (integrating Qlib and RDAgent).
• Analytics/: Financial analytics modules.
• *.py: Individual data source integrations (60+ providers).

Technical Requirements

• Python Version: 3.11+
• Execution: Uses an embedded Python runtime bundled with the application.
• Output Format: All scripts must return JSON responses.
• Concurrency: Scripts execute via QProcess with a maximum of 3 concurrent executions.

The open-source build of Fincept Terminal is a native C++20 desktop application using Qt6 and embedded Python 3.11. It is distributed as a single binary without requiring Node.js or a browser runtime. Key capabilities include:

• Analytics: DCF, Portfolio optimization, VaR/Sharpe, Derivative valuation, Fixed Income, and a QuantLib suite with 18 modules.
• AI Agents: 37 specialized agents (Trader/Investor, Economics, Geopolitics) using your own LLM keys.
• Data Connectors: Over 100 connectors including FRED, IMF, World Bank, DBnomics, AkShare, Polygon, Kraken, Yahoo Finance, and various government APIs.
• Trading: Crypto and equity feeds, a paper trading engine, and 16 broker connections.
• Automation: Visual node editor, MCP tools, and an AI Quant Lab (ML, factor research, RL).
• Intelligence: Maritime tracking, geopolitical analysis, and relationship mapping.

Fincept Terminal is a native C++20 desktop financial intelligence platform designed for research, analytics, and automation. It features a Qt6 UI and embedded Python 3.11 for analytics, operating as a single binary without the overhead of Electron or Node.js.

There are two primary editions:

1. Open Source Edition (AGPL-3.0): This repository. It is intended for students, hobbyists, and academic research. It is free to use but requires users to provide their own data API keys and LLM keys. It ships one release per month.
2. Enterprise Edition (Proprietary): A closed-source build for professional funds and research desks. It includes private datasets, live broker routing, SSO, and managed AI credits. It is available via subscription starting at $99/user/month.

The Open Source edition of Fincept Terminal provides a comprehensive suite of financial tools built on C++20, Qt6, and embedded Python 3.11. Key capabilities include:

• Analysis: DCF, portfolio optimization, VaR/Sharpe, derivative valuation, and a QuantLib suite with 18 modules.
• AI: 37 specialized trader/investor agents. Requires your own LLM keys (e.g., OpenAI, Anthropic, Gemini, Groq, DeepSeek, OpenRouter, or Ollama).
• Data: Over 100 connectors including FRED, IMF, World Bank, DBnomics, AkShare, Polygon, Kraken, and Yahoo Finance.
• Trading: Crypto and equity feeds, a paper trading engine, and 16 broker integrations.
• Automation: Visual node editor, MCP tools, and an AI Quant Lab (ML, factor discovery, RL).
• Intelligence: Maritime tracking, geopolitical analysis, and relationship mapping.

FinceptTerminal is transitioning from a tool-using chatbot to a true agentic system capable of long-running autonomous tasks (e.g., autonomous research and report writing). The research focuses on implementing durable execution, adaptive re-planning, and reflection/self-correction mechanisms.

Existing Agentic Primitives

The system already contains several core components that implement the 'Plan-and-Solve' pattern with SQLite checkpointing:

• ResumableTaskRunner
• TaskStateManager (located in scripts/agents/finagent_core/task_state.py)
• Python main.py actions: start_task, resume_task, get_task, and list_tasks.
• RD-Agent integration for autonomous quant R&D.

The free AGPL-3.0 edition includes the following capabilities:

• Analytics: DCF, portfolio optimisation, VaR/Sharpe, derivatives pricing, fixed income, alternatives, and an 18-module QuantLib suite.
• AI: 37 specialized agents (trader/investor, economic, and geopolitics). Users must provide their own LLM keys (supports OpenAI, Anthropic, Gemini, Groq, DeepSeek, OpenRouter, and Ollama).
• Data: 100+ connectors including FRED, IMF, World Bank, DBnomics, AkShare, Polygon, Kraken, Yahoo Finance, and various government APIs.
• Trading: Crypto and equity feeds, a paper-trading engine, and 16 broker integrations.
• Automation: A visual node editor, MCP tools, and an AI Quant Lab (ML, factor discovery, RL).
• Global Intelligence: Maritime tracking, geopolitical analysis, and relationship mapping.

The Open Edition provides a robust suite of tools for financial research:

• Analytics: DCF, portfolio optimization, VaR/Sharpe, derivative valuation, fixed income, and an 18-module QuantLib suite.
• AI: 37 trader/investor, economics, and geopolitics agents. Supports your own keys for OpenAI, Anthropic, Gemini, Groq, DeepSeek, OpenRouter, and Ollama.
• Data: Over 100 connectors including FRED, IMF, World Bank, DBnomics, AkShare, Polygon, Kraken, Yahoo Finance, and various government APIs.
• Trading: Crypto and equity feeds, a paper trading engine, and 16 broker integrations.
• Automation: Visual node editor, MCP tools, and AI Quant Lab (ML, factor discovery, RL).
• Global Intelligence: Maritime tracking, geopolitical analysis, and relationship mapping.