Skip to content

HugoWong528/Telegram-bot

BranchesTags

Repository files navigation

AI Telegram Bot

This repository provides a unified Telegram bot that combines general AI chat, multi-role company discussions, autonomous code generation, code review, and Hong Kong weather reminders — all powered by Pollinations AI.

One TELEGRAM_TOKEN is all you need.

For Railway deployment (recommended always-on hosting) see RAILWAY.md.
For other platforms (Linux/VPS, Docker, macOS, Windows, GitHub Actions) see PLATFORM.md.

What's New

  • Telegram bot — replaced the Discord bot; works in direct messages and group chats
  • Live streaming replies — the bot edits its message as the AI streams tokens
  • Automatic code review — after /build or /autorun, a Code Reviewer AI checks for critical issues and auto-regenerates (up to 2 rounds)
  • Universal /followup — context-aware continuation:
    • Active AI stream → cancels it and redirects with your input
    • After /build / /autorun → build follow-up (amend code, ask questions); supports unlimited chained follow-ups
    • Otherwise → continues the general chat conversation
  • /weather — fetches current Hong Kong weather from the HKO official RSS feed and generates AI-powered clothing suggestions (Traditional Chinese + English); optional auto-reminders

Repository Structure

Telegram-bot/
├── bot.py               # ★ Unified Telegram bot — one TELEGRAM_TOKEN, all features
├── requirements.txt     # Python dependencies
├── runtime.txt          # Python version
├── railway.toml         # Railway deployment config
├── README.md
├── RAILWAY.md           # Railway.com deployment guide
├── PLATFORM.md          # All platforms deployment guide
└── APIDOCS.md           # Pollinations AI API reference

Quick Start

1. Create a Telegram Bot & Get Your Token

  1. Open Telegram and start a chat with @BotFather.
  2. Send /newbot and follow the prompts (choose a name and a username ending in bot).
  3. BotFather will send you the bot token — copy it. It looks like 7123456789:AAF....
  4. (Optional) Register your bot commands with BotFather: 
    /setcommands
    Then paste the list from the Commands section below.

2. Get a Pollinations API Key

Go to https://enter.pollinations.ai and copy your POLLINATIONS_TOKEN.

3. Run Locally

git clone https://github.com/HugoWong528/Telegram-bot.git
cd Telegram-bot

pip install -r requirements.txt

export TELEGRAM_TOKEN="your_telegram_bot_token"
export POLLINATIONS_TOKEN="your_pollinations_api_key"
# Optional — needed for /build to commit generated code:
# export GITHUB_TOKEN="your_github_pat"
# export GITHUB_REPOSITORY="owner/repo"
# Optional — auto weather reminders (Hong Kong):
# export WEATHER_CHAT_ID="your_telegram_chat_id"
# export WEATHER_REMINDER_HOURS="8,20"   # HKT hours to post (default: 8)

python bot.py

4. Deploy on Railway (Recommended)

Railway is the easiest way to keep your bot running 24/7. See RAILWAY.md for a step-by-step guide.

All Commands

Register these with BotFather via /setcommands:

start - Show the help guide
about - Show the full help guide
ask - Ask the AI a question (live streaming)
cancel - Cancel your in-progress AI request
models - List all available AI models
settings - View or set your preferred AI model
company - Run a multi-role AI company discussion
build - Developer team discussion + code generation
autorun - Fully autonomous build (AI picks the task)
company_roles - List all available roles
followup - Context-aware continuation
weather - Current HK weather + AI clothing suggestions

General Chat

Command Description
/ask <question> Ask the AI; response streams live
/ask model:<name> <question> Ask with a specific model
/cancel Cancel your current in-progress AI request
/models List all available AI models
/settings <model> Set your preferred AI model
/about Show the full help guide

AI Company / Build

Command Description
/company <task> Multi-role AI company discussion
/company <task> roles:CEO,CTO,Designer interactive:true Interactive mode
/build <task> Developer team discussion + code gen + GitHub commit
/build <task> roles:... interactive:true Interactive build
/autorun Fully autonomous: AI picks task + builds end-to-end
/autorun stack:python Force a specific tech stack
/company_roles List all available roles

Weather

Command Description
/weather Current HK weather (HKO RSS) + AI clothing suggestions

Universal

Command Description
/followup <request> Context-aware: redirects an active stream, continues a build session, or adds to general chat

How /followup Works

/followup is the single command for all "continue the conversation" needs:

  1. If an AI stream is running → cancels it and redirects with your input
  2. If a /build or /autorun session exists in this chat → build follow-up: amend code, ask questions, generate new files and commit them. You can chain as many follow-ups as you like.
  3. Any other time → general chat follow-up: adds your message to the conversation history and gets a streaming AI reply
/followup Add JWT authentication to the API
/followup Now add rate limiting to the auth routes
/followup Can you write unit tests for the middleware?
/followup Actually, make the tests use pytest-asyncio

Hong Kong Weather — /weather

The /weather command fetches the current official weather bulletin from the Hong Kong Observatory RSS feed:

Auto Weather Reminders

Set two environment variables to have the bot post a weather report automatically:

Variable Description
WEATHER_CHAT_ID Telegram chat ID to post reminders to
WEATHER_REMINDER_HOURS Comma-separated hours in HKT to post (default: 8). E.g. 8,20 posts at 08:00 and 20:00 HKT.

To find your chat ID: forward a message to @userinfobot or use @RawDataBot.

Automatic Code Review

After /build or /autorun generates code, the bot automatically:

  1. Format check — verifies ### File: blocks were produced; retries generation up to 3 times if not
  2. Code review — a Code Reviewer AI checks for critical issues
  3. Auto-fix loop — if issues are found, regenerates the code with the reviewer's feedback injected (up to 2 review rounds)
  4. Upload to GitHub — commits the verified code to project/<slug>/ in the repository

Available AI Models

Model Vision Notes
openai-fast ✅ Fast OpenAI model; default first choice
openai ✅ Full OpenAI model
gemini-search ✅ Gemini with Google Search grounding
gemini ✅ Standard Gemini model
gemini-fast ✅ Faster Gemini variant
gemini-large ✅ Larger Gemini variant
claude-fast ✅ Fast Claude model
glm ❌ GLM model
qwen-character ❌ Qwen character model
deepseek ❌ DeepSeek model
qwen-safety ❌ Qwen safety model

Configuration Reference

Variable Required Description
TELEGRAM_TOKEN ✅ Telegram bot token from BotFather
POLLINATIONS_TOKEN ✅ Pollinations AI API key
GITHUB_TOKEN optional GitHub PAT — needed for /build to commit files
GITHUB_REPOSITORY optional owner/repo — needed for /build to commit files
WEATHER_CHAT_ID optional Telegram chat ID for auto weather reminders
WEATHER_REMINDER_HOURS optional HKT hours for auto reminders (default: 8). E.g. 8,20

Notes

  • Sessions are stored in memory — restart the bot and sessions reset. Use Railway or a VPS for persistent uptime.
  • If all AI models fail: "Sorry, all AI models are currently unavailable."
  • Responses exceeding Telegram's 4 096-character limit are automatically split into multiple messages.
  • Build-session /followup calls are serialised per chat — if you fire two /followup commands rapidly, the second waits for the first to finish.

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages