financial-market-analysis

⚠️ DISCLAIMER (IMPORTANT) This project and all its content (including but not limited to: market analysis, holdings advice, add/reduce/take-profit/stop-loss signals, buy/sell directions, new investment directions, decision tracking) are provided for learning, research and technical exchange only, and do NOT constitute investment advice, investment consultation, or any promise of returns. Investing involves risk; be cautious when entering the market. Any investment decision shall be made independently by you, and you shall bear all risks and consequences yourself. The author accepts no liability for any direct or indirect loss arising from the use of this project.

📌 Unrelated to bidding/tender monitoring services This project is a financial market analysis skill. It does NOT provide bidding agency, tender monitoring, bid submission, or any procurement-service business. The words "tender / procurement" appearing herein refer only to public procurement-notice data sources used as industry-sentiment and order-demand research data (e.g. State Grid tender YoY as an equipment-sector activity indicator) — financial/industry research only.

A configurable financial market analysis skill for DeepSeek Harness: A-shares / funds / QDII / gold / Asia-Pacific indices / global markets — macro, policy, geo and rule-based analysis, plus daily reports (morning/evening), indicator tracking and decision tracking.

Zero personal hardcoding: holdings, delivery chats, credentials and data directory are all configured via config/config.example.json; credentials are read from environment variables only.

Features

• Multi-source data: Taikang TIPVS / AIIB / Chery / Geely / UNGM / World Bank WDS / Sina / Yahoo (see references/data-source-gotchas.md).
• Chief-Investment-Analyst style daily reports: directly tells you "where to put the money / when to add/reduce/take profit / new directions / risks".
• Indicator tracking: each day first re-checks the status of all prior trigger conditions (triggered / not-triggered / stop-loss hit), then gives today's instructions.
• Decision tracking: buy/reduce/take-profit recommendations are automatically written to the decisions table; after 30 days you can backfill to verify win rate.
• Data degradation: on data-source failure, automatically falls back per the fallback protocol (DB → backup source → hardcoded last resort).

Installation

git clone <repo-url> market-intel
cd market-intel

# 1) Python dependencies
pip install -r requirements.txt
# or isolated: python3 -m pip install --target .pyreqs requests feedparser

# 2) Configure
cp config/config.example.json config/config.json
vim config/config.json   # data dir, Feishu delivery, portfolio

# 3) Prepare data dir (optional: run P0 once to init DB and ingest)
export MARKET_INTEL_ROOT="$HOME/.dsh/market_intel"
PYTHONPATH=scripts python3 scripts/run_all_p0.py

Usage

Run a morning report manually

export MARKET_INTEL_ROOT="$HOME/.dsh/market_intel"
export FEISHU_APP_ID=cli_xxx
export FEISHU_APP_SECRET=xxx
python3 scripts/render_prompts.py          # render report_prompts → rendered/
dsh --profile headless "$(cat rendered/morning.txt)"

Scheduled tasks (macOS launchd)

See examples/launchd/ (P0 ingest 07:50 / morning 09:00 / evening 16:00, weekdays), or use cron / systemd yourself.

Directory Layout

• SKILL.md — skill description (read by DSH skill scanner)
• scripts/ — data pipeline (_lib.py provides configurable ROOT; run_all_p0.py inits DB and ingests; evaluate_today*.py/ai_compute.py/hot_themes.py etc. analysis; decision_tracker.py decision tracking; render_prompts.py renders report prompts)
• report_prompts/ — morning/evening agent prompt templates ({{placeholders}})
• references/ — data sources, fallback protocol, report template, pitfalls (methodology)
• config/config.example.json — config sample (copy to config.json)
• examples/launchd/ — launchd scheduled-task samples

Configuration (config.json)

Key	Description
market_intel_root	Data dir (contains scripts/db/state); default ~/.dsh/market_intel; can be overridden by env MARKET_INTEL_ROOT
feishu.app_id_env / app_secret_env	Env var names for Feishu credentials (never stored in repo/docs)
feishu.domain	Feishu API domain
feishu.deliver_chat_ids	chat_id list to deliver reports
portfolio.total_capital / positions / cash_equivalent	Total capital, positions (code/name/amount/proxy index), cash
reports	Morning/evening times and weekdays
decision_expires_days	Decision expiry days (default 30)

Working with DeepSeek Harness

• Put the skill in a DSH skill scan root (e.g. ~/.dsh/skills/ or project .dsh/skills/) to load it.
• Scheduled reports can be triggered via dsh --profile headless "$(cat rendered/morning.txt)" (see examples/launchd/).
• To place the data dir anywhere, set MARKET_INTEL_ROOT; all scripts locate files via _lib.ROOT.

Related Projects

• dsh-feishu-bridge — Full Feishu/Lark channel for DeepSeek Harness (two-way chat + Feishu approvals); pair it with this skill to receive reports on Feishu.

Customization & Support

Need it tailored to your scenario (private data sources, multi-account holdings, custom report templates, scheduled deployment)? Contact: p.wangxo@gmail.com (subject prefix "[定制]").

License

MIT