The bot is still in its early stages and currently rate-limited due to my budget constraints for OpenAI API credits. As a result, response times may be slow and answers may occasionally be inaccurate. Please don't abuse the bot/spam it. Responses may take upto 10s some times.
Repository: https://github.com/Neeraj138/vibe-ai-tg-bot/tree/main
Bot Link: https://t.me/vybe_vibe_ai_bot
This is a conversational Telegram bot that enables natural language interaction with on-chain Solana data. Built using SolanaAgentKit and Vybe Network APIs, the bot supports rich query capabilities across tokens, wallets, programs, and protocols.
- Natural language support for all interactions
- Fully conversational Telegram experience (no commands required)
- Integrates Vybe Network APIs for Solana on-chain data
- Supports both Telegram messages and CLI usage
- Memory and tool chaining via LangGraph + LangChain
- Typed plugin system with validation using Zod
vybe-ai-tg-bot/
├── package.json
├── tsconfig.json
├── vercel.json
├── src/
│ ├── index.ts # Agent REPL (CLI)
│ ├── telegram-server.ts # Telegram webhook server
│ ├── types/ # SolanaAgentKit type extensions
│ └── my-plugin/ # Custom Vybe plugin
│ ├── actions/ # LangChain action definitions
│ ├── tools/ # Tool implementations
│ ├── types/ # Zod validation schemas
│ └── utils/ # Helpers and label normalization
git clone https://github.com/Neeraj138/vibe-ai-tg-bot.git
cd vibe-ai-tg-bot/backend
nvm use --lts
yarn installCreate a .env file with the following values:
TELEGRAM_BOT_TOKEN=<your_telegram_bot_token>
OPENAI_API_KEY=<your_openai_api_key>
VYBE_API_KEY=<your_vybe_api_key>
SOLANA_PRIVATE_KEY=<keypair_in_json_array_format>
RPC_URL=https://api.mainnet-beta.solana.comStart in CLI (terminal prompt):
yarn devBuild and run HTTP server (e.g., for Telegram):
yarn build
yarn startThe project includes a vercel.json config to deploy the Express server as a serverless function on Vercel.
To set up the Telegram webhook after deployment:
curl -X POST "https://api.telegram.org/bot<YOUR_BOT_TOKEN>/setWebhook?url=https://<your-vercel-app>.vercel.app/webhook"The LangGraph-based agent uses OpenAI's GPT model with tools generated via SolanaAgentKit. Vybe API integrations are added via a custom plugin inside src/my-plugin.
Telegram → Express → LangGraph Agent → SolanaAgentKit Tools → Vybe API → Solana
Use natural language questions to trigger tool actions. Example queries by category:
- show known Solana accounts for defi
- get me amm wallets
- fetch hacker addresses
- what tokens does this wallet hold?
- show only verified tokens in this wallet
- which SPL tokens are in this address?
- what NFTs are in this wallet?
- show NFT portfolio value
- list 5 NFTs from this address
- get PnL performance of wallet X for the last 30 days
- which token made the most profit for this wallet?
- show all programs from Star Atlas
- list programs with category DEFI and NFT
- get average number of instructions per user for program X
- show token transfers for SEND token in the last 24 hours
- find all tokens received by a wallet
- list top transfers by USD value
- what’s the current price and market cap of token X?
- show daily volume and holder count of SEND token
- get top token holders by balance
A full list of example queries for each tool is available in the project documentation.
- Per-user memory context is maintained using thread ID (for multi-turn interactions)
- Tool calls and responses are logged to console
- Telegram typing indicator is displayed while agent processes requests
- Rate-limited with basic in-memory throttling
MIT License. Created and maintained by Neeraj138.