Airtel’s 1-year Perplexity Pro offer includes full API access to the sonar-pro model. Instead of using it only for search, I used the free API credits to build an open-source real-time market signal bot that monitors price action, filters noise using LLM reasoning, and sends actionable alerts to Telegram.
The entire system runs on simple components but produces high-signal outputs suited for active traders.
🧱 System Architecture
1. Price Feed (Finnhub)
The bot polls Finnhub’s /quote endpoint for:
- US stocks
- Crypto pairs (e.g.,
BINANCE:BTCUSDT)
Only symbols with >|1.5%| intraday movement move forward.
2. LLM-Based Market Reasoning (Perplexity)
Filtered snapshots are sent to Perplexity’s sonar-pro via:
https://api.perplexity.ai/chat/completions
The prompt instructs the model to return structured JSON:
should_alert: true/false- concise
reason
- sentiment classification (
bullish,bearish,unclear)
This replaces custom heuristics with dynamic context-aware filtering.
3. Alert Formatting
Signals are composed into a short Telegram-friendly message with sentiment emojis.
4. Telegram Push Delivery
Alerts are delivered instantly using Telegram Bot API.
🔄 Core Processing Loop
FETCH prices
→ FILTER >1.5% movers
→ ANALYZE using Perplexity (JSON output)
→ FORMAT alerts
→ SEND to Telegram
This creates a low-overhead 24/7 signal layer driven entirely by free API credits.
🧪 Code Overview (Repository Ready)
Your bot contains four main functions:
get_latest_price()→ real-time quotes
analyze_trends_with_perplexity()→ LLM signal classification
format_alert_message()→ compact alert formatting
send_telegram_message()→ asynchronous delivery
The main loop continuously aggregates movement, analyzes it, and pushes signals.
⚡ Why This Made Productive Use of Airtel’s Free Perplexity Pro
Zero-cost LLM inference for market reasoning- No need for ML models or rule-based systems
LLM filters out noise, preventing alert spam
Open-source and easy to deploy (just env vars + Python)- Works for both crypto and equities
Instead of just using Perplexity Pro for normal search, the API became the intelligence layer of a real-time signal engine.
📦 What should we add next?
- Volume spike detection
- Multi-timeframe trend scoring
- News + price combined signals
- Discord & Slack integrations
- Web dashboard for recent alerts
If you have any Query regarding this post let me know in comment or Tweet me. thank you for reading this.