Python Basic SIP/RTP Server

A lightweight SIP (Session Initiation Protocol) and RTP (Real-time Transport Protocol) server implemented in Python. This server acts as a basic echo server for voice calls, receiving SIP INVITE messages, establishing RTP streams, and echoing back received audio packets.

Why This Instead of Asterisk or FreeSWITCH?

When building a voice bot or IVR system, traditional media servers like Asterisk or FreeSWITCH add significant complexity:

• Steep learning curve and configuration overhead
• Resource-intensive setup and maintenance
• Overkill for simple voice processing applications

This Python implementation provides a direct, minimal path from Cisco (or other SIP endpoints) to your Python application, allowing seamless integration with:

• Speech-to-Text (STT) services
• GPT/LLM for natural language processing
• Text-to-Speech (TTS) for responses

Instead of managing a full media server, you get a simple Python script that handles the SIP signaling and RTP transport, letting you focus on the voice AI logic.

Features

• SIP signaling support (INVITE, BYE, OPTIONS, UPDATE)
• RTP audio streaming (PCMA codec)
• Automatic RTP port allocation
• Thread-safe call management
• Basic call logging

How It Works

1. SIP Handling: Listens on UDP port 5060 for SIP messages
2. Call Establishment: Responds to INVITE with 200 OK and SDP offer
3. RTP Echo: Starts an RTP server that echoes received audio packets back to the caller
4. Call Termination: Handles BYE messages to clean up resources

Running the Server

Local Development

python server.py

The server will bind to 127.0.0.1:5060 by default.

Environment Variables

• BIND_IP: IP address to bind the server (default: 0.0.0.0 for containers)
• ADVERTISED_IP: IP address advertised in SDP (default: 127.0.0.1)

Modifying for Voice Bot

To transform this into a voice bot, modify the rtp_echo_server function in rtp.py:

1. Decode Audio: Convert incoming PCMA RTP packets to PCM audio
2. Speech-to-Text: Send audio chunks to STT service (e.g., Google Speech, Azure, Whisper)
3. Process with GPT: Send transcribed text to GPT/LLM for response generation
4. Text-to-Speech: Convert response text to audio using TTS service
5. Encode and Send: Convert TTS audio back to PCMA and send via RTP

Example modification:

def rtp_voice_bot(server_ip, rtp_port, remote_sdp, stop_event):
    # Parse remote SDP for IP/port
    # ...

    sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
    sock.bind((server_ip, rtp_port))

    audio_buffer = []

    while not stop_event.is_set():
        data, addr = sock.recvfrom(2048)
        # Decode RTP payload to PCM
        pcm_data = decode_pcma(data)
        audio_buffer.append(pcm_data)

        # Process in chunks
        if len(audio_buffer) > CHUNK_SIZE:
            text = stt.process(audio_buffer)
            response = gpt.generate_response(text)
            tts_audio = tts.synthesize(response)
            # Encode to PCMA RTP and send
            rtp_packets = encode_pcma_rtp(tts_audio)
            for packet in rtp_packets:
                sock.sendto(packet, addr)

Dependencies

• Python 3.7+
• No external dependencies (uses only standard library)

Testing

Run the unit tests:

python -m unittest tests/test_sip_message.py

Architecture

• server.py: Main SIP server loop and call management
• sip_message.py: SIP message parsing
• sip_response.py: SIP response building
• sip_utils.py: Utility functions for extracting caller/callee info
• rtp.py: RTP packet handling (currently echo server)
• tests/: Unit tests

Limitations

• Single codec support (PCMA/G.711)
• No security (no authentication, encryption)
• Basic error handling
• Not production-ready (for development/learning only)

Contributing

Feel free to extend this for your voice bot needs. Key areas for enhancement:

• Multiple codec support
• DTLS-SRTP for encryption
• Call recording
• Integration with popular STT/TTS/GPT APIs