Keel: Like a ship's keel that provides stability in turbulent waters, KeelTrader helps you maintain psychological balance in volatile markets.
KeelTrader is an AI-powered performance coach for trading psychology (Web: Next.js, API: FastAPI). It's built around chat, knowledge base (RAG), and a "roundtable" multi-coach discussion mode.
Disclaimer: for educational/research purposes only. This project is not investment advice.
In the TV series Billions, Wendy Rhodes is the in-house performance coach at hedge fund Axe Capital. Her role isn't investment analysis—it's psychological performance optimization:
| What She Does | Real Scenarios |
|---|---|
| Emotional First Aid | When a trader melts down mid-session, she intervenes immediately to restore composure |
| Behavioral Pattern Recognition | Identifies that a trader's losing streak stems from divorce affecting judgment |
| Cognitive Reframing | Helps overconfident star traders recognize their blind spots |
| Psychological Resilience Training | Teaches traders to bounce back quickly after massive losses |
| Performance Enhancement | Improves team-wide trading stability through psychological coaching |
The Real-World Problem:
- Only elite hedge funds can afford full-time performance coaches (salaries $300K+)
- Retail traders and small institutions have zero access to this service
- 80% of trading losses stem from psychological factors, not technical analysis
KeelTrader's Vision:
Democratize Wendy Rhodes' capabilities through AI, giving every trader their own performance coach.
KeelTrader represents the evolution of behavioral finance from theory to practice:
| Generation | Focus | Limitation |
|---|---|---|
| 1.0 Academic Research | Kahneman & Tversky's cognitive biases | Theory stays in textbooks, traders can't apply it |
| 2.0 Robo-Advisors | Automated portfolio allocation | One-size-fits-all, ignores individual psychology |
| 3.0 AI Performance Coach | Real-time psychological intervention | Personalized, actionable, adaptive |
Traditional Approach (Behavioral Finance 1.0 & 2.0):
- Read about loss aversion → Still panic sell during crashes
- Know about confirmation bias → Still cherry-pick bullish news
- Understand overconfidence → Still overtrade after wins
KeelTrader's Approach (Behavioral Finance 3.0):
- Conversational Intervention: AI coaches discuss your trades and help you reflect on emotional patterns in your journal
- Personalized Coaching: Build a library of coaches tailored to different aspects of your trading psychology
- Actionable Guidance: Not just "you have loss aversion" but "here's how to handle this specific trade decision right now"
- Knowledge Base (RAG): Import your own trading books, notes, and strategies for contextual advice
- (Planned) Real-time pattern detection and proactive alerts
Academic Knowledge → [MISSING LINK] → Trading Performance
(BF 1.0) (Your P&L)
↓
KeelTrader
(BF 3.0 Bridge)
Most traders know the theory but fail in execution. KeelTrader bridges this gap by providing:
- Reflective coaching: Review your trades with AI coaches to identify emotional patterns
- Journal analysis: Track your trading decisions and discuss them with specialized coaches
- Accountability partner: Multi-coach roundtable discussions help you examine decisions from different perspectives
- (Planned) Moment-of-truth alerts and proactive pattern recognition
- Chat + projects: organize conversations per project, keep history, stream responses (SSE)
- Roundtable discussion: multiple AI coaches discuss one question with configurable session/message settings
- Knowledge base (RAG): import docs, semantic search (pgvector), inject context by timing
- Attachments: upload images/docs/audio (extract/transcribe where supported)
- Journaling + reports: trading journal, analytics, scheduled reports (Celery)
- Journal import (CSV/XLSX): upload a file and map columns in the UI (works with different broker/export formats)
- Notifications: in-app notifications and alert delivery hooks for pattern alerts, risk warnings, and daily summaries
- Trading intervention system: pre-trade checklists, real-time trade blocking based on risk limits, and proactive alerts
- Behavior pattern tracking: detect and review revenge trading, FOMO, overtrading, and other behavioral patterns
- Exchange integration: connect to crypto exchanges (OKX, Bybit, Coinbase, Kraken) and brokers (IBKR) with trading mode awareness (spot / perpetual futures / stocks / options per connection)
- Agent Matrix (optional profile): multi-agent trading system — Orchestrator, Technical Analyst, Executor, Psychology Coach, Guardian — with 8-layer safety barrier, ghost trading, and Telegram bot interface
- Self-hosted by default: Docker Compose; optional cloud/SaaS mode via env flags
cd keeltrader
Copy-Item .env.example .env # PowerShell (or: cp .env.example .env)
docker compose up -d --build- Web:
http://localhost:3000 - API health:
http://localhost:8000/api/health - API docs:
http://localhost:8000/docs
Full guide: keeltrader/docs/SELF_HOSTING.md
The default Docker Compose setup builds the web with login enabled (NEXT_PUBLIC_AUTH_REQUIRED=1). Set KEELTRADER_AUTO_INIT_TEST_USERS=1 to create test accounts on first startup.
To disable login and use guest mode, set KEELTRADER_AUTH_REQUIRED=0 in .env and rebuild the web image with NEXT_PUBLIC_AUTH_REQUIRED=0.
- Add more “1-click demo” options (cloud deploy templates)
- Improve preset library and import/export
- More evaluators/benchmarks for coaching quality
- Start here:
keeltrader/docs/INDEX.md - Architecture:
keeltrader/docs/ARCHITECTURE.md - Deployment:
keeltrader/docs/DEPLOYMENT.md - Self-hosting:
keeltrader/docs/SELF_HOSTING.md - Deployment modes:
keeltrader/docs/DEPLOYMENT_MODES.md - Custom API setup:
keeltrader/docs/CUSTOM_API_SETUP.md - Internationalization:
keeltrader/docs/I18N_GUIDE.md - Version Management:
docs/VERSION_MANAGEMENT.md⭐ NEW - Changelog:
CHANGELOG.md
- Branch strategy: We use Git Flow. Create feature branches from
developand submit PRs todevelop(notmain). SeeCONTRIBUTING.mdfor details. - Contributing:
CONTRIBUTING.md - Code of Conduct:
CODE_OF_CONDUCT.md - Security policy:
SECURITY.md
KeelTrader supports two modes:
- Self-hosted (default): open-source community edition
- Cloud/SaaS: multi-tenancy, billing, enterprise SSO, analytics (activated only when
DEPLOYMENT_MODE=cloud)
See keeltrader/docs/DEPLOYMENT_MODES.md for details.
KeelTrader 是一套面向交易心理与行为表现的 AI 教练系统（Web: Next.js，API: FastAPI）。核心能力围绕：对话、知识库（RAG）、以及"圆桌讨论"（多教练协作）。
免责声明：仅用于教育/研究目的，本项目 不构成 投资建议。
- 对话 + 项目：按项目组织会话，历史记录，SSE 流式输出
- 圆桌讨论：多位 AI 教练围绕同一问题讨论；支持会话级/消息级设置
- 知识库（RAG）：文档导入、pgvector 语义检索，按时机注入上下文
- 附件：图片/文档/音频上传（按能力抽取/转写）
- 交易日志 + 报告：交易日志、统计、定时报表（Celery）
- 交易日志导入（CSV/XLSX）：支持上传文件并在页面中手动映射列（适配不同券商/平台格式）
- 通知能力：站内通知与告警分发钩子，覆盖模式提醒、风险警告和每日总结
- 交易干预系统：交易前检查清单、基于风险限制的实时交易阻止和主动警报
- 行为模式跟踪：识别并复盘报复性交易、FOMO、过度交易等行为模式
- 交易所集成：连接加密交易所（OKX、Bybit、Coinbase、Kraken）和券商（IBKR），支持按连接配置交易模式（现货 / 永续合约 / 股票 / 期权）
- Agent Matrix（可选 profile）：多 Agent 交易系统 — 协调器、技术分析师、执行器、心理教练、风控守卫 — 含 8 层安全屏障、模拟交易、Telegram 机器人
- 默认可自托管：Docker Compose 一键启动；也支持通过环境变量启用云端模式
cd keeltrader
Copy-Item .env.example .env # PowerShell（或：cp .env.example .env）
docker compose up -d --build- Web：
http://localhost:3000 - API 健康检查：
http://localhost:8000/api/health - API 文档：
http://localhost:8000/docs
完整说明：keeltrader/docs/SELF_HOSTING.md
默认需要登录（KEELTRADER_AUTH_REQUIRED=1）。设置 KEELTRADER_AUTO_INIT_TEST_USERS=1 可在首次启动时创建测试账号。
如需免登录访客模式，将 .env 中 KEELTRADER_AUTH_REQUIRED=0，并以 NEXT_PUBLIC_AUTH_REQUIRED=0 重新构建 web 镜像。
- 增加更多“一键 Demo”部署模板
- 预设教练库的导入/导出与共享
- 教练效果的评测与对比工具
- 从这里开始：
keeltrader/docs/INDEX.md - 架构：
keeltrader/docs/ARCHITECTURE.md - 部署：
keeltrader/docs/DEPLOYMENT.md - 自托管：
keeltrader/docs/SELF_HOSTING.md - 部署模式：
keeltrader/docs/DEPLOYMENT_MODES.md - 自定义 API：
keeltrader/docs/CUSTOM_API_SETUP.md - 国际化：
keeltrader/docs/I18N_GUIDE.md - 版本管理：
docs/VERSION_MANAGEMENT.md - 更新日志：
CHANGELOG.md
- 贡献指南：
CONTRIBUTING.md - 行为准则：
CODE_OF_CONDUCT.md - 安全策略：
SECURITY.md
KeelTrader 支持两种模式：
- Self-Hosted（默认）：开源社区版
- Cloud/SaaS：多租户、计费、企业 SSO、分析（仅在
DEPLOYMENT_MODE=cloud时启用）
详见：keeltrader/docs/DEPLOYMENT_MODES.md