StockPilot

StockPilot is an AI-powered investment research workbench — not a real-time quote terminal, and not a one-off chat session. It's built for long-term, conviction-driven investors who track a small, focused set of stocks and want their AI-generated analysis to persist, accumulate, and be revisited — not disappear at the end of a chat.

Powered by LLM agents and a FastAPI + Vue 3 + SQLite stack, StockPilot turns ad-hoc "ask an AI about a stock" into a structured, repeatable research workflow: daily reports, position-aware specialist analysis, and a searchable archive of everything the AI has told you, all stored locally.

Why StockPilot

• Not a quote/monitoring tool — no tick-by-tick feeds or minute-level alerts. StockPilot is for deep, periodic research on stocks you actually hold or are seriously considering, not for watching the market all day.
• Not a disposable chat — every AI analysis is saved as structured Markdown/HTML and indexed in SQLite, so you can trace how your thesis on a stock evolved over weeks or months.
• Built for a focused watchlist, not the whole market — designed around a handful of stocks you actually care about, with agents that understand your current positions, not generic market-wide scanning.

Screenshots

Main Interface

Overview

It provides:

• Watchlist management with per-stock and bulk analysis
• Structured daily report storage in SQLite and on disk
• Portfolio position tracking and profit history
• Scheduled daily summaries via email or WeCom webhook
• Interactive analysis sessions with streaming updates
• Specialist agent analysis for currently held positions

Highlights

• Report artifacts saved as Markdown and HTML files under backend/reports/
• Built-in report rescanning and analysis-history cleanup endpoints
• Settings page for daily-report LLM config, specialist-agent LLM config, and TickFlow API key storage

Quick Start

1. Install Python dependencies

Install from the repository root because the backend imports packages listed in the top-level requirements.txt. python==3.11

pip install -r requirements.txt

2. Install frontend dependencies

cd frontend
npm install
cd ..

3. Start the application

bash start.sh

Runtime Flows

Watchlist Daily Analysis

1. Add a stock from the Stocks page or POST /api/watchlist.
2. Trigger one stock or the whole watchlist through the analysis endpoints.
3. The backend queues work and runs analysis in FIFO order.
4. The daily-report flow invokes the local claude code.
5. Generated Markdown is parsed into structured report fields and stored in SQLite.
6. Report artifacts are saved under backend/reports/.

Outputs:

• Structured report rows in SQLite
• Markdown report files
• HTML reports when generated
• Analysis status information for the frontend

Specialist Agent Analysis

On the Stocks page, the "strategy specialist analysis" button shown for held positions (frontend/src/views/Stocks.vue) opens the Agent analysis dialog. This flow is integrated from the daily_stock_analysis project (https://github.com/ZhuLinsen/daily_stock_analysis) and uses the project's analysis-strategy Skill definitions together with the Agent runtime code under backend/core/src/agent/.

It depends on these saved settings:

• agent_api_key
• agent_base_url
• agent_model

Notifications and Scheduling

The Settings page persists configuration in the settings table and updates the running scheduler immediately when schedule_time changes.

Supported built-in notification channels in the web settings flow:

• Email
• WeCom webhook

Notification attempts are written to the notification_log table and surfaced by dashboard endpoints.

Configuration

There is no .env.example file in the repository. Most user-facing runtime settings are stored through the Settings page and persisted in SQLite.

Daily Report LLM Settings

Saved fields:

• claude_model
• claude_api_key
• claude_auth_token
• claude_base_url

When these values are updated, the backend rewrites:

• backend/reports/.claude/settings.json

Specialist Agent Settings

Saved fields:

• agent_api_key
• agent_base_url
• agent_model

These are applied to runtime environment variables used by the specialist agent flow.

Notification Settings

Saved fields include:

• smtp_email
• smtp_password
• receiver_email
• wechat_webhook_url
• wechat_msg_type
• schedule_time

TickFlow

Saved field:

• tickflow_api_key

The backend applies this value to the TICKFLOW_API_KEY environment variable at runtime.

Known Boundaries

• Daily report generation depends on a local claude code.
• Interactive analysis also depends on Claude-based local tooling.
• Specialist agent analysis requires OpenAI-compatible credentials saved in Settings.

License

This project can be used freely, including personal use, learning, modification, and redistribution within your own workflow.

There is currently no standalone LICENSE file in the repository root. If you want that permission to be formalized for external distribution or public reuse, add a root-level LICENSE file with the exact terms you want to publish.