ashare-news-fetcher

A-share news and sentiment scraping

Scrape A-share-related intelligence from multiple public Chinese financial/policy/social sources and output structured entries.

What this skill does

Scrape Chinese financial news flashes (Cailianshe, Wallstreetcn, Jin10, Sina 7x24, Eastmoney Express)
Scrape regulatory policy announcements (CSRC, PBOC, SSE, MOF)
Scrape individual stock sentiment indicators from Eastmoney Guba
Extract relevant A-share codes from titles/summaries
Output JSON or Markdown for subsequent analysis, archiving, or feeding to an LLM

Capability boundaries

What it can do:

Input a stock code (such as 000001,600519) to scrape that stock's Guba sentiment
When no code is input, scrape market-wide financial/policy news flashes
Filter results by keyword
Output JSON (structured) or Markdown (readable)

What it cannot do:

Does not scrape content behind paywalls/login walls
Does not guarantee real-time performance (the sources themselves are public aggregations and may have minute-level delays)
Does not scrape exchange L1/L2 market data (this is news/sentiment scraping, not a market data interface)
Does not replace compliant investment advisory research; output is only for information aggregation reference

Prerequisites

Python 3.10+
requests(required)
beautifulsoup4(optional, only needed when scraping policy sources)
akshare(optional, used for name→code matching; without it, only regular expressions are used to extract codes)
jieba(optional; after installation, name matching uses word segmentation, reducing false matches)
Environment variable JIN10_APP_ID(required by the Jin10 data interface; if not set, that source is skipped and other cn sources are unaffected)

Minimal installation:

pip install requests
# 如需政策来源：
pip install beautifulsoup4
# 如需名称匹配：
pip install akshare
# 如需更精确的名称匹配：
pip install jieba

Usage

Method 1: Call directly in conversation

/ashare-news-fetcher 000001,600519 cn,guba

Parameter description (all optional):

First parameter: stock code, comma-separated, e.g. 000001,600519
Second parameter: source groups, comma-separated; optional values cn(Chinese news flashes),policy(policy),guba(Guba)
When no parameters are passed, defaults to --sources cn --limit 30

Method 2: Run the script directly

python scripts/fetch_intel.py --sources cn,policy --limit 20
python scripts/fetch_intel.py --symbols 000001,600519 --sources guba
python scripts/fetch_intel.py --sources cn --keywords 降准,降息 --format markdown

Common options:

--sources: source groups, default cn
--symbols: stock code, used for guba
--keywords: keyword filter, comma-separated
--limit: maximum number of items returned per group, default 30
--format：json or markdown, default json
--output: output to file; if omitted, print to stdout
--verbose: print debug logs

Source description

Source group	Source	Content type	Whether bs4 is required
`cn`	Cailianshe, Wallstreetcn, Jin10, Sina 7x24, Eastmoney Express	Market news flashes	No
`policy`	CSRC, PBOC, SSE, MOF	Policy/Regulatory Announcements	Yes (except CSRC)
`guba`	East Money Stock Forum	Individual Stock Retail Investor Sentiment	No

For detailed API endpoints and parsing methods, see references/source_apis.md。

Test

For test instructions and how to run, see references/testing.md。

Output Format

JSON

For the structure of each item, see references/output_schema.md. Example:

{
  "item_id": "4ffde7bc7b18b55e",
  "source_id": "cn_sina",
  "source_name": "sina",
  "title": "...",
  "summary": "...",
  "url": "...",
  "category": "market",
  "priority": "normal",
  "tags": ["公司", "市场"],
  "related_symbols": ["000001"],
  "published_at": "2026-06-25 17:50:41",
  "fetched_at": "2026-06-25 09:50:52",
  "extra": {}
}

Markdown

Sorted in reverse chronological order, each entry shows source, time, category, related code, tags, link, and summary.

Notes

Public APIs may change at any time; the script will skip failed sources without affecting other sources.
If you encounter request failures in a domestic network environment, confirm that the system HTTP_PROXY/HTTPS_PROXY points to an available proxy.
guba The source is relatively sensitive to anti-scraping, so it is recommended not to request too many stocks at once; there is already a 0.5s request interval between codes.
Before using Jin10 Data (jin10), it is recommended to configure JIN10_APP_IDin the environment variables; if not configured, this source will be automatically skipped, without affecting other cn sources such as Cailian Press, Wallstreetcn, Sina, East Money, etc.
```
export JIN10_APP_ID="your-app-id"
```

Next Step

After obtaining structured intelligence, the common next step is to hand it to the analysis skill for summarization or daily report generation:

Would you like to summarize these messages and generate an A-share news daily report?

A) Generate a Markdown briefing (recommended)
B) Only do sentiment/keyword aggregation
C) Not needed, the current output is sufficient