This content originally appeared on Level Up Coding - Medium and was authored by Kevin Meneses González
How I reduced analysis time from 40 minutes to 40 seconds using FMP data, news sentiment, and an LLM-powered workflow
he biggest risk is not taking any risk.” — Mark Zuckerberg
1. The Real Problem: Too Much Data, Too Little Clarity
Financial markets produce an insane amount of information every single day:
- Prices
- Volume
- Indicators
- Company earnings
- Analyst opinions
- Macro news
- Social sentiment
- Rumors
- Geopolitical shocks
Most traders don’t fail because they lack tools — they fail because:
They cannot synthesize information fast enough and they make decisions influenced by emotion instead of structured analysis.
Even if you know how to code, backtest, or read charts…
You’re constantly fighting information overload.
2. The Solution: LLMs + FMP Market Data + News Sentiment
Large Language Models (LLMs) solve a simple but painful problem:
They can consume large amounts of data and summarize it instantly.
If you combine:
- Price data + technical indicators
- News + sentiment
- A structured prompt
- A reliable data provider (FMP)
You get something extremely powerful:
➜ A fully automated “market research assistant”
that can break down everything into:
- Clear summaries
- Trend analysis
- Risks
- Opportunities
- Recommendations (not financial advice)
This workflow has helped me personally to:
- Reduce research time from 30–40 minutes to under 40seconds
- Remove emotional bias
- Understand market context better
- Act with more confidence and less stress
And now I’ll show you how to build it too.
3. What We’re Going to Build
A complete pipeline that:
- Fetches price history from FMP
- Computes a moving average indicator
- Fetches recent news + sentiment using FMP’s /stock_news endpoint
- Formats all the data clearly
- Sends it to an LLM
- Receives a detailed, AI-generated trading insight summary
4. Step-by-Step Python Implementation
4.1. Install Dependencies
pip install requests pandas python-dotenv openai
4.2. Load Environment Variables
from dotenv import load_dotenv
import os
load_dotenv()
FMP_API_KEY = os.getenv("FMP_API_KEY")
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
4.3. Fetch Price History from FMP
import requests
import pandas as pd
def get_price_history(symbol: str, limit: int = 120):
url = "https://financialmodelingprep.com/api/v3/historical-price-full/" + symbol.upper()
params = {"timeseries": limit, "apikey": FMP_API_KEY}
r = requests.get(url, params=params)
r.raise_for_status()
df = pd.DataFrame(r.json()["historical"])
df["date"] = pd.to_datetime(df["date"])
return df.sort_values("date").reset_index(drop=True)
What this does:
- Calls FMP
- Returns OHLCV data sorted by date
- Ready for indicator calculation
4.4. Add Simple Indicators
def add_indicators(df):
df = df.copy()
df["ma20"] = df["close"].rolling(20).mean()
return df
4.5. Format Market Context for the LLM
def format_market_context(df, symbol):
latest = df.iloc[-1]
context = f"""
Market Data Summary for {symbol}:Latest Close: {latest['close']:.2f}
20-day MA: {latest['ma20']:.2f}
Trend: {"Bullish" if latest['close'] > latest['ma20'] else "Bearish"}
1-Day Change: {latest['close'] - df.iloc[-2]['close']:.2f}
Last 5 Sessions:
{df.tail(5).to_string(index=False)}
"""
return context
4.6. Fetch News + Sentiment
Using the correct FMP endpoint:
/api/v3/stock_news
def get_news_sentiment(symbol: str, limit: int = 8):
url = "https://financialmodelingprep.com/api/v3/stock_news"
params = {
"tickers": symbol.upper(),
"limit": limit,
"apikey": FMP_API_KEY
}
r = requests.get(url, params=params)
r.raise_for_status()
articles = r.json()
processed = []
for a in articles:
processed.append({
"title": a.get("title"),
"text": a.get("text"),
"site": a.get("site"),
"published": a.get("publishedDate"),
"symbol": a.get("symbol")
})
return processed
4.7. Format News for the LLM
def format_sentiment_context(news_list):
if not news_list:
return "No recent news found."
output = "Recent News:\n"
for n in news_list:
output += f"- {n['published']} | {n['site']} | {n['title']}\n"
return output
4.8. LLM Insight Generator
from openai import OpenAI
client = OpenAI(api_key=OPENAI_API_KEY)
def generate_ai_insights(market_context, news_context):
prompt = f"""
You are a financial analyst.
Based on the MARKET DATA and NEWS provided, generate a clear trading insight summary including:
1. Market context
2. Trend analysis
3. Interpretation of news sentiment
4. Risks or opportunities
5. A simple trading recommendation (NOT financial advice)
MARKET DATA:
{market_context}
NEWS:
{news_context}
"""
response = client.chat.completions.create(
model="gpt-4o-mini",
messages=[{"role": "user", "content": prompt}]
)
return response.choices[0].message.content
4.9. Full Script
symbol = "AAPL"
df = get_price_history(symbol)
df = add_indicators(df)
market_context = format_market_context(df, symbol)
news_list = get_news_sentiment(symbol)
news_context = format_sentiment_context(news_list)
insights = generate_ai_insights(market_context, news_context)
print(insights)
5. Example Output (Typical LLM Summary)
“AAPL shows mild bullish momentum as price trades above the 20-day MA.
Recent news sentiment is moderately positive, with headlines centered on product launches and earnings stability.
Combined, this suggests a constructive short-term trend with opportunities on pullbacks, though watch for macro-driven volatility.”
This is exactly the type of clarity most traders need.
And it takes less than 20 seconds.
6. Tips to Improve Your Workflow
- Add more indicators
RSI, MACD, ATR, volatility bands…
- Compute a “sentiment score”
Count positive vs negative headlines.
- Add fundamentals
FMP lets you fetch:
- P/E
- P/S
- EBITDA
- Debt levels
- Revenue growth
- Analyst ratings
- Let the LLM compare multiple assets
Feed 3 tickers at once and ask:
“Which asset has the strongest combination of trend + sentiment?”
- Automate the system
Run it daily with a cron job and email yourself the report.
7. FAQ
Why use FMP instead of Yahoo Finance?
Because FMP provides:
✔ Reliable historical data
✔ Stock news + timestamps
✔ Fast JSON responses
✔ Crypto, ETFs, Forex, indices
✔ Financial statements
✔ Easy API authentication
All in one place.
Do LLMs replace technical analysis?
No.
They synthesize data — you still decide the final action.
Can this system be automated?
Absolutely:
- Cron jobs
- AWS Lambda
- GitHub Actions
- Airflow
8. Final Ideas
If you want to build powerful AI-driven trading assistants:
👉 Get your FMP API Key — better data = better decisions
👉 Use Python to automate your research
👉 Add news sentiment for deeper context
👉 Use LLMs to generate instant, structured insights
This workflow has helped me think clearly, act more calmly, and make more rational financial decisions.
And you can build it today with less than 100 lines of code.
How to Use LLMs and FMP Market Data to Generate Smart AI-Driven Trading Recommendations was originally published in Level Up Coding on Medium, where people are continuing the conversation by highlighting and responding to this story.
This content originally appeared on Level Up Coding - Medium and was authored by Kevin Meneses González
Kevin Meneses González | Sciencx (2025-11-25T14:51:44+00:00) How to Use LLMs and FMP Market Data to Generate Smart AI-Driven Trading Recommendations. Retrieved from https://www.scien.cx/2025/11/25/how-to-use-llms-and-fmp-market-data-to-generate-smart-ai-driven-trading-recommendations-2/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.