This update 0.3.2 introduces support for Text formats/Response formats support PyDantic JSON model schemas:

• Refinements to endpoint: /v1/responses
  • Adds parameters for test_format (Response formats) supporting PyDantic model JSON schemas.
  • Adds response parameter: output_parsed.
• Refinements to default parameters on endpoint: /v1/chat/completions
• Additional schema models.

Continue reading

This update 0.3.0 introduces bi-directional function calling (client and server-side) tool calling:

• Additions to schemas.
• Refinements to endpoint: /v1/responses
  • Adds parameters for tool/function calling.
  • Tool calling supported on client and server side.

Continue reading

Following on our release for the Rhea: Latent Loom project this update 0.2.5 introduces:

• Merges ingestion endpoints in to a single endpoint.
• Additions to schemas.
• New endpoint: /v1/rag/synthetic for generating synthetic questions used during knowledge ingestion.
• Refinements to: /v1/search allowing vector distance threshold customisations.

Continue reading

Announcing the Rhea: Latent Loom project a locally hosted LLM services that can be ran independently or with the Rhea Generative Framework. This post introduces the Rhea: Latent Loom API specification in OpenAPI (3.0.1) format. A Swagger UI Bundle is used to explore the API.

Rhea: Latent Loom can serve as drop-in replacement for AI cloud providers. Request responses are compatible with Chat Completion and Responses object types.

Continue reading