Skip to content

AI for Investor

AI for Investor connects natural-language research, knowledge retrieval, strategy development, backtest validation, trading workspaces, and portfolio-risk views into an auditable quantitative research workflow. It accelerates research; it does not replace data validation, risk controls, or human trading decisions.

Get started

From a question to a verifiable result

  1. Use AI Chat and a knowledge base to retrieve research rules, post-mortems, and data definitions.
  2. Create, review, or generate a Backtrader strategy draft in Strategies, then add it to a research workspace.
  3. In Data → Market Data, choose an instrument and check coverage and quality. The page reads the local MySQL market-data warehouse first; AkShare is contacted only after an explicit query.
  4. Run a backtest in a research workspace, inspect metrics and robustness checks, and retain its configuration and result snapshot.
  5. Move human-reviewed work to a trading workspace, then monitor accounts, positions, trades, cumulative P&L, drawdown, and allocation on the portfolio page.

Core capabilities

Domain Current capability
AI and knowledge Knowledge bases, document indexing, cited answers, strategy ideation/review, and AI research
Data trust AkShare warehouse, MySQL-first reads, coverage matrix, quality warnings, explicit online refresh, and caching
Research and backtests Backtrader runs, normalized metrics, reports, strategy versions, research workspaces, and robustness checks
Trading and risk Simulation/trading workspaces, gateway state, portfolio aggregation, position valuation, P&L, and drawdown views
Engineering FastAPI, Vue 3, SQLAlchemy, MySQL/PostgreSQL/SQLite, pytest, Vitest, and Playwright

Operating boundaries

  • Review AI output, RAG evidence, and backtest metrics before acting. Historical results are not future returns.
  • Market data, strategy code, and account data can be sensitive. Keep secrets in environment variables or a secrets manager; never commit them.
  • Live gateways are high-risk. Validate in research/simulation first, then follow your organization’s approval and risk controls.

Documentation map

  • Getting started: install, run, and complete a first research loop.
  • Features: knowledge, market data, strategies, backtests, workspaces, and optimization.
  • Development: architecture, API, and data boundaries.
  • Deployment: Docker and production checklist.
  • Reference: configuration and common commands.

Technology

Layer Technology
Frontend Vue 3, TypeScript, Vite, Element Plus, ECharts, Pinia
Backend FastAPI, Pydantic, SQLAlchemy 2, Uvicorn
Research engine Backtrader; fincore adapters where available, with compatible metric calculation otherwise
Data and AI AkShare, MySQL market-data warehouse, OpenAI-compatible generation, optional ChromaDB / sentence-transformers semantic retrieval
Project Description
backtrader Professional Python algorithmic trading framework (backtesting + live trading); the core fork behind this platform's research engine.
backtrader-skills Offline strategy author/review/test product: datasets + StrategySpec v1 → pytest strategies or three-file bundles with static review and child-process validation.
backtrader-mcp Local-first MCP server: frozen CSV datasets, typed strategy drafts, bounded subprocess runs with status and reports (offline, backtest-only).
backtrader_web This platform: web-based full-cycle Backtrader strategy management (backtesting analysis, paper trading, live execution, data management).
backtrader-agent Offline-first strategy-authoring agent runtime: content-addressed storage, spec validation, scaffolds, static review, and hash-bound execution.
fincore Unified Python toolkit: financial metrics, performance analysis, backtesting, AI insights, and multi-database/data source support.

For repository entry points, internal engineering docs, and the archive policy, see docs/INDEX.md.