Compared to alternatives

This open source Go library provides a few different things to enable you to build Sovereign AI.

  • Routing: Easily switch your LLM, STT, TTS, decision and other models
  • Acceleration: Run edge servers in the cloud to accelerate your agents. Video/Audio RTC, skills, tool calling etc
  • SDKs for backend & client side AI development
  • Persistence: Storage, memory and sandbox integrations
  • Voice: Modern telephony/voice integrations
  • CLI/Dashboard & Skills to quickly build

Let's discuss how this compares to the various alternatives.

OpenAI /Anthropic Direct

Many companies integrate directly with OpenAI or Anthropic.

Sovereign AI - Advantages:

  • You can easily switch models. Frontier, open models or your own
  • Building AI is faster
  • Video & Chat SDKs
  • Global acceleration edge
  • Control over costs
  • Decide if you want to share IP with OpenAI/Anthropic or not

OpenAI/Anthropic direct - Advantages

  • One layer of abstraction less
  • Simpler

We had this ourselves as well. We integrated a state of the art model. 6 months later it fell behind. But it was hard to change to a new model. So we ended up spending more money, for a lower quality than the state of the art.

There are a couple of reasons why building with this framework is faster than openAI or anthropic direct.

  • Chat and video/voice SDKs for iOS, android, flutter, react native, react etc.
  • Skills that makes it faster to build Agents
  • SDKs offer client, stateflow as well as UI components
  • Knowledge and skills are a simple sync command (vs a zip upload and environment in OpenAI)

Historically going direct had the advantage that features where available immediately. This repo uses a software factory approach. So it's typically following in hours after a release.

Eve and AI.SDK

Eve and AI SDK are great frameworks focused on the JS ecosystem. Vercel also offers Routing and Connect in addition to this. The frameworks are mostly focused on text based chat and have limited support for voice & video.

Sovereign AI - Advantages

  • SDKs for all languages
  • Global acceleration edge
  • Chat & Video SDKs
  • Voice & Video support

Eve/AI SDK

  • If you only do JS this approach can be simpler
  • Zero markup on the router/gateway. Similar to running open source, Cheaper than our hosted

Both:

  • Similar benefits in terms of routing and connector management.
  • Sandbox and subagent support
  • Guardrail support

Routers: OpenRouter, BiFrost and LiteLLM

Sovereign AI - Advantages

  • Acceleration
  • Video/Voice support
  • Persistence
  • Open Source
  • Chat and Video SDKs

BiFrost/LiteLLM/ OpenRouter

  • Routing only, can be an advantage if you want only routing
  • Bifrost has MCP and semantic caching

Voice AI frameworks: Livekit, Pipecat, Retell, Agora or Vapi

Sovereign AI - Advantages

  • Full text AI feature set
  • Global acceleration edge
  • SDKs for all languages
  • Go instead of Python
  • Skill to quickly build

Voice AI frameworks - Advantages

  • If you specifically want Python
  • Very mature ecosystem

Both/similar: Great telephony, voice loop

Support SAAS: Sierra and Decagon

Sovereign AI - Advantages

  • Infrastructure, not an app
  • SDKs for all languages
  • Open source
  • Easy to customize
  • No Lock in

Sierra/Decagon - Advantages

  • Success based fee. (At a very high premium)
  • AI agent that reviews your agent

Next steps