Siphon is an open-source Python framework that abstracts the complexity of building AI calling agents, letting developers ship voice AI in hours instead of weeks.

THE DEVELOPER PROBLEM:
Building AI calling agents requires deep expertise in:
• SIP signaling and telephony infrastructure
• Real-time WebRTC audio streaming
• Orchestrating LLM, STT, and TTS services with sub-500ms latency
• Managing interruptions, state, and async event handling
• Implementing call persistence and monitoring

Most dev teams spend 2-4 weeks on infrastructure before writing conversation logic.

THE SOLUTION:
Siphon provides a batteries-included framework. Here's a complete AI receptionist:

from siphon.agent import Agent
from siphon.plugins import openai, cartesia, deepgram

agent = Agent(
agent_name="receptionist",
llm=openai.LLM(),
tts=cartesia.TTS(),
stt=deepgram.STT(),
system_instructions="You are a helpful receptionist..."
)

if name == "__main__":
agent.start() # Handles real PSTN calls

That's production-ready. Deploy it and it answers phone calls.

ARCHITECTURE:
• Built on LiveKit (WebRTC transport layer)
• Native SIP integration for PSTN connectivity
• Streaming audio pipeline for <500ms voice-to-voice latency
• Plugin system for AI providers (OpenAI, Anthropic, Groq, etc.)
• Horizontal auto-scaling with zero configuration
• Automatic transcript/recording/metadata persistence

TECHNICAL HIGHLIGHTS:
• Vendor-agnostic: Swap LLM/STT/TTS providers via config
• Self-hosted: Apache 2.0 license, you own the infrastructure
• Production-ready: Built-in monitoring, error handling, state management
• Developer-friendly: Type hints, clear APIs, comprehensive docs

PERFECT FOR BUILDING:
• Customer support automation
• Appointment scheduling systems
• Sales qualification pipelines
• Healthcare intake workflows
• Any telephony + AI use case

WHY WE BUILT IT:
After months spent wrangling SIP infrastructure for a voice AI project, we realized every developer building calling agents hits the same walls. Siphon removes these barriers.

REPO & DOCS:
GitHub: https://github.com/blackdwarftech/siphon
Docs: https://siphon.blackdwarf.in/docs
Install: pip install siphon-ai

SIPHON

About