# finance data source

[finnlp](https://github.com/AI4Finance-Foundation/FinNLP/tree/main/finnlp/data_sources) github shows a list of sources to read from.

# transcript

[https://discountingcashflows.com/api/transcript/AAPL/q1/2024/](https://discountingcashflows.com/api/transcript/AAPL/q1/2024/)

but it seems always return the latest transcript.

# fmp

[fingpt fmp colab example](https://github.com/AI4Finance-Foundation/FinGPT/blob/master/fingpt/FinGPT_Sentiment_Analysis_v1/FinGPT_v1.0/data_preparations/FMP/FMP.ipynb)

price target and [news](https://site.financialmodelingprep.com/developer/docs#stock-news-news) can't be got for free account.

the cheapest is [$19/month](https://site.financialmodelingprep.com/developer/docs/pricing), subscription yearly. Does news data belong to Market data?  
Conclusion: give up for now.

# marketaux

[documentation](https://www.marketaux.com/documentation)

can get 3 news per request [example in fingpt](https://github.com/AI4Finance-Foundation/FinGPT/blob/master/fingpt/FinGPT_Others/shares_news_sentiment_classify.py)

get news from market aux, then call chatgpt to filter.

But after randomly looking at the news, it is hard to find really useful info. Most news are not really news, they are just repeating similar info, like it's time to buy the stock.

Maybe I haven't found the right way to use the news.

source\_ids, many I don't know.

* [businessinsider.com](http://businessinsider.com)
    
* [seekingalpha.com](http://seekingalpha.com)
    
* [forbes.com](http://forbes.com)
    

example:

```python
def get_news_from_market_aux(api_key: str, data_path: str = 'finance_news_from_market_aux.txt'):
    limit_line = 4

    if os.path.exists(data_path):
        with open(data_path, 'r') as f:
            data = json.load(f)
    else:
        conn = http.client.HTTPSConnection('api.marketaux.com')

        params = urllib.parse.urlencode({
            'api_token': api_key,
            "found": 8,
            "returned": 3,
            "limit": limit_line,
            "symbols": "SMCI",
            "min_match_score": 80,
            "page": 1,
            "source_id": "forbes.com-1",
            "domain": "forbes.com",
            "language": "en",
        })

        conn.request('GET', '/v1/news/all?{}'.format(params))

        data = conn.getresponse()
        data = data.read().decode('utf-8')
        data = json.loads(data)
        with open(data_path, 'w') as f:
            f.write(json.dumps(data, indent=2))

    assert isinstance(data, dict)

    '''concert dict to string (Title: ... Content: ...)'''
    max_num_news = 8
    max_len_title = 32 * 4
    max_len_content = 0 * 4

    data = data['data']

    data_str = ""
    for item in data[:max_num_news]:
        title = item['title'][:max_len_title]
        content = item['description'][:max_len_content]
        data_str += f"{title}, {content}\n"
    return data_str
```

# polygon

use polygon to download historical flatfiles data.

But the cheapest subscription doesn't have news? give up.