I've been trading stocks and crypto on the side for a while, and I got tired of opening 5 different tabs just to research a ticker before making a move. So I built AI Research Pro — a Telegram bot that brings real-time prices, AI analysis, news, and alerts into one place.
Here's what it does and how I built it.
What it does
- /price AAPL — real-time price, change %, volume, market cap
- /analyze TSLA — Groq-powered AI gives a buy/hold/sell summary with key metrics
- /news BTC — latest headlines for any stock or crypto
- /movers — top gainers and losers across stocks and crypto right now
- /alert NVDA 900 — price alert that notifies you when hit
- /watchlist — track your portfolio of tickers
- /briefing — daily AI-generated market summary delivered to you
Try it here: https://t.me/AIResearchProbot
Tech stack
- python-telegram-bot (v21.6 with job-queue) for the bot framework
- yfinance for stock data (prices, fundamentals, news)
- CoinGecko API for crypto prices
- Groq API (llama-3.1-8b-instant) for the AI analysis — it's blazing fast and free-tier friendly
- SQLite for user data, watchlists, and alerts
- Railway for hosting (auto-restarts on failure)
The trickiest bug I fixed
yfinance changed their news item structure in v0.2.40. Old code accessed item["title"] directly. New structure nests everything under item["content"]["title"]. This silently broke /news — it returned 0 results with no error.
Fix: a helper that detects both formats:
def _parse_news_item(item: dict) -> tuple[str, str, str, str]:
if "content" in item:
c = item["content"]
title = c.get("title", "")
publisher = c.get("provider", {}).get("displayName", "")
link = (c.get("canonicalUrl") or c.get("clickThroughUrl") or {}).get("url", "")
pub_date = c.get("pubDate", "")
try:
dt_str = datetime.fromisoformat(pub_date.replace("Z", "+00:00")).strftime("%b %d, %H:%M")
except Exception:
dt_str = ""
else:
title = item.get("title", "")
publisher = item.get("publisher", "")
link = item.get("link", "")
ts = item.get("providerPublishTime", 0)
dt_str = datetime.fromtimestamp(ts).strftime("%b %d, %H:%M") if ts else ""
return title, publisher, link, dt_str
AI analysis prompt engineering
The AI gets a structured context block with price, P/E, EPS, 52-week range, volume, and recent news headlines. The prompt asks for: current sentiment, key strengths/risks, and a brief buy/hold/sell take. Keeping the context tight keeps Groq's latency under 2 seconds.
What's next
- Options flow data
- Portfolio tracking with P&L
- Premium tier with more analysis depth
If you're into markets or building bots, give it a try: https://t.me/AIResearchProbot
Happy to answer questions about the tech or share more code snippets below.
