v0.6.0
v0.6.0 — Public Connector API
Highlights
- Public connector API —
hidden_dim,num_layers,context_length,vocab_size,device,dtype,tokenize(),detokenize(),apply_chat_template(),stop_token_ids,stop_stringson the EngineConnector ABC - OutputType/PayloadType split — API-level enum (
OutputType) separated from wire-level enum (PayloadType).PayloadType.AUTOremoved. - LlamaCpp latent primitives —
create_inference_context(),run_latent_steps(),keep_context=True,grammar=for constrained generation - 11 bug fixes including critical HF
inject_and_generate()crash and LlamaCpp use-after-free
Breaking changes
tokenize()returnsList[int](wasAny/torch.Tensor)PayloadType.AUTOremoved — useOutputType.AUTOextract_hidden_state()/inject_and_generate()removed from ABC- LlamaCpp
tokenize()no longer adds BOS
See CHANGELOG.md for full details.