Expand description
API protocol conversion and inference output parsing.
request::ProtocolRequest::convert produces a shared
request::ConversationRequest. stream::StreamProcessor parses backend
output, passes it to the selected protocol’s chunk generator
(stream::ChunkGenerator), and accumulates completion usage. The caller
renders and encodes the prompt, supplies the backend and its prompt usage,
and transports the protocol responses.
Modules§
- anthropic
- Anthropic-compatible protocol types and conversion.
- error_
response - Error response bodies for conversion errors.
- openai
- request
- Shared request data and the conversion contract for protocol adapters.
- response
- Protocol response construction and chunk accumulation.
- stream
- Parse model output and turn it into protocol response chunks.