SuperOptiX + TurboAgents

SuperOptiX 0.2.22

This release makes TurboAgents a first-class retrieval option in SuperOptiX, updates local model defaults to Qwen, and refreshes the docs and demo surfaces to
match the current runtime story.

Highlights

• Added TurboAgents-backed retrieval support to the shared RAG path
• Added TurboAgents GEPA vector-store integration
• Added support for:
  • turboagents-chroma
  • turboagents-faiss
  • turboagents-lancedb
  • turboagents-surrealdb
• Updated demo playbooks and docs to reflect the new retrieval options
• Switched smaller local defaults to qwen3.5:2b
• Switched larger local defaults to qwen3.5:9b
• Aligned Chroma support to chromadb 1.5.5
• Blocked compromised LiteLLM releases 1.82.7 and 1.82.8

TurboAgents Integration

SuperOptiX now uses TurboAgents as the first full reference integration layer for compressed retrieval and reranking. This includes:

• Shared RAG integration in rag_mixin
• GEPA vector-store wrappers
• Chroma, LanceDB, and SurrealDB demo coverage
• Updated integration guide and example docs

Validated Paths

• Focused TurboAgents integration tests passed
• CLI and pipeline integration tests passed
• Docs build passed
• Real rag_lancedb_demo validation passed against the published turboagents 0.1.0a2 package

Dependency Alignment

• SuperOptiX is aligned to published turboagents 0.1.0a2
• Chroma is aligned to chromadb 1.5.5
• LiteLLM dependency metadata now excludes the compromised March 2026 releases

Install

uv add superoptix
uv add "superoptix[turboagents]"

Links

• Docs: https://superagenticai.github.io/superoptix/
• PyPI: https://pypi.org/project/superoptix/0.2.22/
• TurboAgents docs: https://superagenticai.github.io/turboagents/
• Repository: https://github.com/SuperagenticAI/superoptix

🧬 A2A v1 Support

🧬 A2A v1 Support

SuperOptiX now has first-class A2A v1 support.

This release adds native A2A interoperability so SuperOptiX agents can be exposed over A2A and SuperOptiX can also call external
A2A agents.

Included

• Core A2A v1 protocol support
• Native A2A client and server bridge
• A2A v1 Agent Card generation
• A2A task/message operations including SendMessage, GetTask, ListTasks, CancelTask, and SubscribeToTask
• Framework-neutral runtime layer for A2A
• Runtime adapters for compiled pipelines, DSPy, Pydantic AI, and Google ADK
• CLI serving with super agent serve --protocol a2a
• Pullable and packaged A2A demos for DSPy, Pydantic AI, and Google ADK
• Dedicated A2A docs and website updates

Notes

This release delivers core A2A v1 interoperability in SuperOptiX. More advanced enterprise A2A features are planned for later.

⛎ SurrealDB Integration v0.2.9

SuperOptiX v0.2.9 - SurrealDB Integration

This release adds first-class SurrealDB support for SuperOptiX RAG workflows and framework demos.

Added

• Native surrealdb retriever support in runner-managed RAG.
• SurrealDB vector store adapter for GEPA RAG.
• SuperSpec schema and validator support for surrealdb retriever type.
• New SurrealDB demo agents:
  • rag_surrealdb_demo (embedded/local)
  • rag_surrealdb_docker_demo (Docker/auth mode)
  • rag_surrealdb_pydanticai_demo
  • rag_surrealdb_crewai_demo
  • rag_surrealdb_adk_demo
• New SurrealDB docs:
  • Embedded demo guide
  • Docker demo guide
  • Framework guide for DSPy, PydanticAI, CrewAI, and Google ADK

Improved

• Minimal pipeline templates now inject retrieved RAG context for:
  • PydanticAI
  • CrewAI
  • Google ADK
• Better SurrealDB URL handling and auth mode behavior for embedded and server setups.
• Beginner-friendly framework documentation and troubleshooting flow.

Validation

• Added tests for SurrealDB vector store behavior:
  • tests/test_surrealdb_vector_store.py

Notes

• provider: ollama controls generation model.
• embedding_model: sentence-transformers/all-MiniLM-L6-v2 is used for embeddings and may download once on first
  run, then use local cache.