Not just a price bot.
It reads the news. Understands the context. Speaks your language.
Runs 24/7. Needs no human intervention.
|
DeepSeek model via OpenRouter. Classifies news as Bullish / Bearish / Neutral. Generates summaries in 3 languages.
|
Polls Cointelegraph, Coindesk and more every 10 minutes. TON-keyword filtered. Smart deduplication via ChromaDB.
|
|
Professional charts & infographics — bot sends images, not just text.
|
Scans TON blockchain via TONAPI. Lets users verify token safety directly inside Telegram.
|
┌──────────────────────────────────────────────────────┐
│ Telegram Interface │
│ aiogram 3.x · Button-based UI │
└────────────────────────┬─────────────────────────────┘
│
┌────────────────────────▼─────────────────────────────┐
│ Service Layer │
│ │
│ AIAnalyst.py RSSFetcher.py │
│ AI summaries & News polling & │
│ market forecasts deduplication │
│ │
│ ChartGenerator.py TokenScanner.py │
│ Price charts & TON blockchain │
│ stat infographics token verification │
└────────────────────────┬─────────────────────────────┘
│
┌────────────────────────▼─────────────────────────────┐
│ Data Layer (Hybrid) │
│ │
│ SQLAlchemy + SQLite ChromaDB │
│ Users · Channels Vector Memory │
│ Settings · Forecasts RAG · Deduplication │
└──────────────────────────────────────────────────────┘
Why hybrid DB? SQLite handles structured data. ChromaDB handles semantic memory — it lets the AI "remember" past news as mathematical vectors, not keywords.
Every 10 min RSSFetcher polls news sources
└─ Filter: TON keywords only
└─ ChromaDB: "Seen this before?"
├─ YES → skip
└─ NO → AIAnalyst summarizes
└─ Broadcast to all users & channels
in their preferred language
09:00 UZB Market Overview infographic sent automatically
09:01 UZB AI Daily Forecast generated from morning news
+ vector memory of past events
Custom Channel owners set their own price update interval
interval (e.g. every 15 min) → bot acts as content engine
🧬 RAG Memory — ChromaDB stores news as vectors. AI never repeats old stories; it understands semantic similarity, not just keywords.
⚡ Fully Async — Built on asyncio + aiogram 3.x. Every service runs concurrently without blocking.
🕐 APScheduler — All timed jobs (news fetch, morning brief, price updates) run on a strict schedule, 24/7.
🌍 Multilingual AI — Same news, three outputs. DeepSeek generates EN / RU / UZ summaries in one call.
📡 Channel Engine — Bot owners can plug TonFonbot into any Telegram channel as an automated content pipeline.
live_price: Instant TON/USDT rate on demand
ai_forecast: Daily market prediction from AI
technical_chart: Price chart generated on request
language_switch: EN · RU · UZ — instant UI rebuild
channel_mode: Automated price & news feed for channel owners
token_check: Verify any TON token via blockchain scanframework: aiogram 3.x (async Telegram)
ai: OpenRouter → DeepSeek model
database: SQLAlchemy + SQLite (structured)
vector_db: ChromaDB (RAG / semantic memory)
scheduler: APScheduler
financial: ccxt (live exchange data)
charts: mplfinance (price chart rendering)
blockchain: TONAPI (token scanning)
languages: English · Russian · Uzbekgit clone https://github.com/aza4k/TonFonbot.git && cd TonFonbot
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
# Fill in: BOT_TOKEN, OPENROUTER_API_KEY, TONAPI_KEY
python main.pyPrerequisites: Python 3.11+ · Telegram Bot Token · OpenRouter API Key · TONAPI Key
TonFonbot/
├── services/
│ ├── AIAnalyst.py ← DeepSeek summaries & forecasts
│ ├── RSSFetcher.py ← news polling & deduplication
│ ├── ChartGenerator.py ← mplfinance price charts
│ ├── StatsGenerator.py ← daily stat infographics
│ └── TokenScanner.py ← TONAPI blockchain queries
├── database/
│ ├── models.py ← SQLAlchemy models
│ └── vector_store.py ← ChromaDB interface
├── handlers/ ← aiogram routers
├── scheduler/ ← APScheduler jobs
├── locales/ ← en · ru · uz translations
└── main.py