How to Build a Telegram Bot with a Free LLM API in Python (2025)

Why Telegram Bots + Free LLM APIs Are a Perfect Pair

Telegram bots are one of the easiest ways to give an AI assistant a real interface that people actually use. Unlike building a full web app, a Telegram bot requires almost no frontend work — Telegram handles the UI entirely. Combined with a free LLM API key from FreeLLMKeys, you can ship a personal AI assistant in under 30 minutes at zero cost.

This guide walks you through everything: setting up the bot on Telegram, wiring it to GPT-4o via FreeLLMKeys, handling conversation history, and deploying it so it runs 24/7.

What You Will Build

• A Telegram bot that responds to messages using GPT-4o
• Per-user conversation memory (the bot remembers what you said)
• A /reset command to clear conversation history
• A /model command to switch between GPT-4o, Claude, and DeepSeek on the fly

Step 1 — Create Your Telegram Bot

1. Open Telegram and search for @BotFather
2. Send /newbot
3. Give it a name and a username (must end in bot)
4. BotFather gives you a token like 7123456789:AAHxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
5. Save this token — it is your bot's API key

Step 2 — Install Dependencies

pip install python-telegram-bot openai

Step 3 — Write the Bot

import os
from openai import OpenAI
from telegram import Update
from telegram.ext import ApplicationBuilder, CommandHandler, MessageHandler, filters, ContextTypes

# ── Config ────────────────────────────────────────────────
TELEGRAM_TOKEN = "YOUR_TELEGRAM_BOT_TOKEN"
LLM_API_KEY    = "sk-your-freellmkeys-key"
BASE_URL       = "https://aiapiv2.pekpik.com/v1"
DEFAULT_MODEL  = "gpt-4o"

client = OpenAI(base_url=BASE_URL, api_key=LLM_API_KEY)

# Per-user conversation history and model selection
user_histories: dict[int, list] = {}
user_models: dict[int, str]     = {}

SYSTEM_PROMPT = "You are a helpful AI assistant. Be concise and friendly."

# ── Command handlers ──────────────────────────────────────

async def start(update: Update, ctx: ContextTypes.DEFAULT_TYPE):
    await update.message.reply_text(
        "👋 Hi! I'm an AI assistant powered by GPT-4o.

"
        "Just send me a message to get started.
"
        "Commands:
"
        "/reset — Clear conversation history
"
        "/model gpt-4o|deepseek-chat|claude-opus-4-7 — Switch model"
    )

async def reset(update: Update, ctx: ContextTypes.DEFAULT_TYPE):
    uid = update.effective_user.id
    user_histories[uid] = []
    await update.message.reply_text("✅ Conversation cleared.")

async def set_model(update: Update, ctx: ContextTypes.DEFAULT_TYPE):
    uid   = update.effective_user.id
    args  = ctx.args
    valid = ["gpt-4o", "deepseek-chat", "claude-opus-4-7", "gemini-2.5-flash"]
    if not args or args[0] not in valid:
        await update.message.reply_text(f"Usage: /model {'|'.join(valid)}")
        return
    user_models[uid] = args[0]
    await update.message.reply_text(f"✅ Switched to {args[0]}")

# ── Message handler ───────────────────────────────────────

async def handle_message(update: Update, ctx: ContextTypes.DEFAULT_TYPE):
    uid     = update.effective_user.id
    text    = update.message.text
    model   = user_models.get(uid, DEFAULT_MODEL)
    history = user_histories.setdefault(uid, [])

    # Add user message to history
    history.append({"role": "user", "content": text})

    # Keep last 20 messages to avoid token overflow
    if len(history) > 20:
        history = history[-20:]
        user_histories[uid] = history

    # Show typing indicator
    await update.message.chat.send_action("typing")

    try:
        response = client.chat.completions.create(
            model=model,
            messages=[{"role": "system", "content": SYSTEM_PROMPT}] + history,
            max_tokens=1024,
        )
        reply = response.choices[0].message.content
        history.append({"role": "assistant", "content": reply})
        await update.message.reply_text(reply)
    except Exception as e:
        await update.message.reply_text(f"⚠️ Error: {e}")

# ── Main ──────────────────────────────────────────────────

def main():
    app = ApplicationBuilder().token(TELEGRAM_TOKEN).build()
    app.add_handler(CommandHandler("start",  start))
    app.add_handler(CommandHandler("reset",  reset))
    app.add_handler(CommandHandler("model",  set_model))
    app.add_handler(MessageHandler(filters.TEXT & ~filters.COMMAND, handle_message))
    print("Bot running...")
    app.run_polling()

if __name__ == "__main__":
    main()

Step 4 — Run It

python bot.py

Open Telegram, find your bot by its username, send a message. You should get an AI response within 2–3 seconds.

Step 5 — Deploy for 24/7 Uptime (Free)

To keep the bot running when your laptop is closed, deploy it for free on Railway or Render:

1. Create a requirements.txt: pip freeze > requirements.txt
2. Push your code to a GitHub repo
3. Connect the repo on railway.app or render.com
4. Add environment variables: TELEGRAM_TOKEN and LLM_API_KEY
5. Deploy — your bot runs 24/7 on their free tier

Adding Image Support

GPT-4o is multimodal. You can add image understanding with a few extra lines:

async def handle_photo(update: Update, ctx: ContextTypes.DEFAULT_TYPE):
    photo = update.message.photo[-1]  # highest resolution
    file  = await photo.get_file()
    url   = file.file_path  # Telegram CDN URL

    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[{
            "role": "user",
            "content": [
                {"type": "image_url", "image_url": {"url": url}},
                {"type": "text",      "text": update.message.caption or "What is in this image?"}
            ]
        }]
    )
    await update.message.reply_text(response.choices[0].message.content)

# Register in main():
app.add_handler(MessageHandler(filters.PHOTO, handle_photo))

Total cost for a fully functional AI Telegram bot with conversation memory, model switching, and image understanding: $0. Grab your FreeLLMKeys key from the homepage and have it running in 30 minutes.