Skip to main content

A standardized framework for building gRPC-based Telegram agents

Project description

AnyAgent Framework

Simple. Fast. Production-ready.

Build Telegram agents with minimal code. Like Express.js for chatbots.

Installation

pip install anyagent

Quick Start

from anyagent import BaseAgent, AgentRequest, AgentResponse, TelegramMessage, TextContent

class MyAgent(BaseAgent):
    def __init__(self):
        super().__init__()
    
    async def execute(self, request):
        yield AgentResponse(
            telegram_message=TelegramMessage(
                text=TextContent(text=f"Echo: {request.telegram_message.text.text}")
            )
        )
    
    async def help(self, request):
        yield AgentResponse(
            telegram_message=TelegramMessage(
                text=TextContent(text="I echo your messages!")
            )
        )

# Run it
from anyagent import AgentServer
AgentServer(MyAgent()).run()  # Starts on port 50051

Features

  • 🚀 Zero config - Just inherit and implement 2 methods
  • 💰 Built-in payments - Pay-per-use with credits
  • 📱 All Telegram types - Text, images, video, audio, documents, location
  • 🎛️ Interactive buttons - Callbacks and keyboards
  • 🔄 Streaming responses - Real-time message streaming
  • 🐳 Docker ready - Production deployment included
  • ⚡ gRPC based - High performance protocol

Payment Requests

from anyagent import UsagePaymentRequest

async def execute(self, request):
    # Request payment for processing
    if not request.paid:
        yield AgentResponse(
            payment_request=UsagePaymentRequest(
                key="text_analysis",  # Payment key (pricing configured in web console)
                quantity=1  # Quantity of operations (1 text analysis)
            )
        )
    
    # Process after payment
    result = analyze_text(request.telegram_message.text.text)
    yield AgentResponse(
        telegram_message=TelegramMessage(
            text=TextContent(text=result)
        )
    )

Interactive Buttons

from anyagent import InlineKeyboard

async def execute(self, request):
    # Handle button clicks
    if request.callback_query:
        data = request.callback_query.callback_data
        if data == "action1":
            yield AgentResponse(
                telegram_message=TelegramMessage(
                    text=TextContent(text="Button 1 clicked!")
                )
            )
        return
    
    # Send message with buttons
    yield AgentResponse(
        telegram_message=TelegramMessage(
            text=TextContent(text="Choose an action:"),
            inline_keyboard=InlineKeyboard(rows=[
                {"buttons": [
                    {"text": "Action 1", "callback_data": "action1"},
                    {"text": "Action 2", "callback_data": "action2"}
                ]}
            ])
        )
    )

All Message Types

async def execute(self, request):
    message = request.telegram_message
    
    if message.text:
        # Handle text
        text = message.text.text
        yield AgentResponse(...)
    
    elif message.image:
        # Handle image
        image_data = message.image.image_data
        filename = message.image.filename
        yield AgentResponse(...)
    
    elif message.video:
        # Handle video
        video_data = message.video.video_data
        yield AgentResponse(...)
    
    elif message.audio:
        # Handle audio
        audio_data = message.audio.audio_data
        yield AgentResponse(...)
    
    elif message.document:
        # Handle document
        file_data = message.document.file_data
        yield AgentResponse(...)
    
    elif message.location:
        # Handle location
        lat = message.location.latitude
        lon = message.location.longitude
        yield AgentResponse(...)

Deployment

Docker

FROM python:3.11-slim
COPY . /app
WORKDIR /app
RUN pip install anyagent
CMD ["python", "agent.py"]
docker build -t my-agent .
docker run -p 50051:50051 my-agent

Docker Compose

version: '3.8'
services:
  agent:
    build: .
    ports:
      - "50051:50051"
    restart: unless-stopped

Architecture

Client (Telegram Bot) 
    ↓ gRPC
Your Agent (Python)
    ↓ 
AnyAgent Framework
    ↓ Protocol Buffers
Agent Server (gRPC)

API Reference

BaseAgent

class BaseAgent:
    def __init__(self)
    async def execute(self, request: AgentRequest) -> AsyncGenerator[AgentResponse, None]
    async def help(self, request: AgentRequest) -> AsyncGenerator[AgentResponse, None]

AgentRequest

class AgentRequest:
    telegram_message: Optional[TelegramMessage]  # User's message
    callback_query: Optional[CallbackQuery]      # Button clicks
    user_id: int                                 # User identifier
    paid: bool                                   # Payment status
    language_code: Optional[str]                 # User's language
    context: Optional[Context]                   # Conversation history

AgentResponse

class AgentResponse:
    telegram_message: Optional[TelegramMessage]     # Message to send
    payment_request: Optional[UsagePaymentRequest]  # Request payment
    memory: Optional[ContextMessage]                # Store meaningful content in conversation memory

Memory Usage: Use the memory field only for meaningful conversational content (final answers, analysis results). Don't store progress updates, loading messages, or temporary UI elements.

TelegramMessage

class TelegramMessage:
    text: Optional[TextContent]
    image: Optional[ImageContent]
    video: Optional[VideoContent]
    audio: Optional[AudioContent]
    document: Optional[DocumentContent]
    location: Optional[LocationContent]
    inline_keyboard: Optional[InlineKeyboard]
    action: Optional[TelegramAction]

UsagePaymentRequest

class UsagePaymentRequest:
    key: str      # Payment key identifier (configured in web console)
    quantity: int # Number of operations (e.g., 60 for 60 minutes of audio processing)

Examples

See the echo_agent directory for a complete example demonstrating all features.

Testing

import grpc
from anyagent.proto import agent_pb2, agent_pb2_grpc

# Test your agent
channel = grpc.aio.insecure_channel("localhost:50051")
stub = agent_pb2_grpc.AgentServiceStub(channel)

request = agent_pb2.AgentRequest(
    user_id=12345,
    paid=False,
    telegram_message=agent_pb2.TelegramMessage(
        text=agent_pb2.TextContent(text="Hello!")
    )
)

async for response in stub.ExecuteStream(iter([request])):
    print(response)

Philosophy

AnyAgent follows the "numpy approach" - minimal required parameters, maximum flexibility. No forced metadata, no complex configuration. Just implement execute() and help(), and you're done.

License

MIT License - build anything you want.

Links

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

anyagent_framework-1.0.0.tar.gz (30.0 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

anyagent_framework-1.0.0-py3-none-any.whl (20.0 kB view details)

Uploaded Python 3

File details

Details for the file anyagent_framework-1.0.0.tar.gz.

File metadata

  • Download URL: anyagent_framework-1.0.0.tar.gz
  • Upload date:
  • Size: 30.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.10.12

File hashes

Hashes for anyagent_framework-1.0.0.tar.gz
Algorithm Hash digest
SHA256 7ef1d1e1e323c84c24ffbabf77e0adc0fefb8822340bef047bbb21e2c62359a1
MD5 0942deb28df017e9a1e668e00e57c658
BLAKE2b-256 aa0c589902553a08f268930ceab70ed4230e86a701b0702d2b4ac2728144b752

See more details on using hashes here.

File details

Details for the file anyagent_framework-1.0.0-py3-none-any.whl.

File metadata

File hashes

Hashes for anyagent_framework-1.0.0-py3-none-any.whl
Algorithm Hash digest
SHA256 6336268be2b93639cd63ed0c3582120eb56fcce4a97a89c2f9b70d057a3be886
MD5 e3fa650b9a71f94d097e28be02c76fad
BLAKE2b-256 7dc7a4a3fe95311ba049ac0eba067f906de7d90fad96dbe8e612f67d330039c1

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page