Best Practices for Using daily_stock_analysis: Complete Setup Guide

Configure your environment variables in a .env file, install Python 3.11+ dependencies, and execute the main.py entry point to run stock analysis as a CLI job, FastAPI service, or scheduled task.

The daily_stock_analysis repository by ZhuLinsen is a Python-based platform for end-to-end analysis of A-share, Hong Kong, and US stocks. It aggregates market data, technical indicators, fundamentals, news sentiment, and LLM-powered insights into a unified pipeline. This guide covers the essential configuration steps and architectural patterns required to deploy the system effectively.

1. Prepare Your Runtime Environment

1.1 Configure Environment Variables

All application settings are loaded from environment variables managed through a singleton Config object in src/config.py. Start by copying the example configuration file:

cp .env.example .env

Populate the following minimal required variables:

Variable	Purpose	Example
STOCK_LIST	Comma-separated stock codes (uppercase)	600519,000001,300750
LITELLM_MODEL	LLM provider/model format	gemini/gemini-3-flash-preview
GEMINI_API_KEYS	API keys for load balancing (comma-separated)	key1,key2
PROXY_HOST / PROXY_PORT	Optional proxy for data sources	127.0.0.1 / 10809

The configuration loader automatically appends domestic financial domains to NO_PROXY when a proxy is configured, preventing accidental proxying of Chinese data sources like Tushare. This logic is implemented in src/config.py at lines 1000–1025.

1.2 Install Python Dependencies

The project requires Python 3.11 or newer. Create an isolated environment and install pinned dependencies:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Understand the Core Architecture

The platform follows a layered architecture with clear separation of concerns:

Layer	Responsibility	Main Modules
Configuration	Environment loading and validation	src/config.py
Data Fetching	Unified interface with fallback and caching	data_provider/*
Analysis Pipeline	Orchestrates technical and fundamental analysis	src/core/pipeline.py
LLM & Agent	Multi-model support via LiteLLM	src/config.py, src/analyzer.py, src/agent/*
Notification	Multi-channel alerts (Feishu, Telegram, Email, etc.)	src/notification.py
Web UI/API	FastAPI backend and optional React frontend	server.py, webui.py
Scheduler	Daily task execution and monitoring	src/scheduler.py

The primary entry point is main.py, which parses CLI arguments, builds the pipeline via src/core/pipeline.py, and determines the execution mode.

3. Configure Data Provider Fallback

The DataFetcherManager in data_provider/* implements a unified interface for multiple data sources including yfinance, Tushare, and Tickflow. Configure multiple API keys in your .env file to enable automatic fallback and load balancing across providers.

4. Optimize LLM Configuration

The system uses LiteLLM for provider-agnostic LLM calls with multi-model fallback support. Define your primary model in LITELLM_MODEL and provide comma-separated keys in provider-specific variables (e.g., GEMINI_API_KEYS, OPENAI_API_KEYS) to enable automatic key rotation. Temperature and other generation parameters are resolved through the configuration layer in src/config.py.

5. Select Your Execution Mode

5.1 CLI Execution

Run one-off analysis jobs directly from the command line. The main.py script handles argument parsing and pipeline initialization for single executions.

5.2 FastAPI Service

Deploy the analyzer as a REST API using server.py. This mode provides HTTP endpoints for on-demand analysis requests and integrates with the optional React frontend in apps/dsa-web/.

5.3 Scheduled Daily Tasks

Use src/scheduler.py to run the pipeline on a daily schedule. This module supports background event monitoring and is designed for production deployment where automated daily stock reports are required.

6. Set Up Notification Channels

Configure notification channels in your .env file to receive alerts via Feishu webhooks, WeChat, Telegram bots, Email, or Discord. The src/notification.py module handles message formatting including markdown-to-image conversion when rich formatting is required by the target platform.

Summary

• Environment Configuration: Copy .env.example to .env and set STOCK_LIST, LITELLM_MODEL, and API keys before running
• Proxy Handling: The system automatically excludes domestic financial domains from proxy routing when PROXY_HOST is configured
• Data Resilience: The DataFetcherManager provides automatic fallback across yfinance, Tushare, and Tickflow providers
• Execution Flexibility: Run as CLI jobs (main.py), FastAPI services (server.py), or scheduled tasks (src/scheduler.py)
• LLM Reliability: Use comma-separated API keys for load balancing and configure multi-model fallback for resilient LLM operations

Frequently Asked Questions

What Python version is required for daily_stock_analysis?

The project requires Python 3.11 or newer. Dependencies are pinned in requirements.txt to ensure compatibility with this version.

How does the system handle proxy configurations for Chinese data sources?

When you set PROXY_HOST and PROXY_PORT, the configuration loader in src/config.py (lines 1000–1025) automatically appends domestic financial domains to the NO_PROXY environment variable. This prevents Tushare and other Chinese data providers from being routed through your proxy, avoiding connection issues while still allowing international sources like yfinance to use the proxy.

Can I use multiple LLM providers simultaneously?

Yes. The system supports multi-model fallback via LiteLLM. Configure your primary model in LITELLM_MODEL and provide multiple API keys (comma-separated) in provider-specific variables like GEMINI_API_KEYS or OPENAI_API_KEYS. The implementation in src/config.py and src/analyzer.py handles automatic key rotation and provider fallback if the primary model fails.

Which notification channels are supported?

The platform supports Feishu (Lark) webhooks, WeChat, Telegram bots, Email (SMTP), and Discord. Configure the relevant environment variables (e.g., FEISHU_WEBHOOK_URL, TELEGRAM_BOT_TOKEN) in your .env file. The src/notification.py module automatically converts markdown reports to images when the notification channel requires visual formatting instead of raw text.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}