acme-engine
v0.1.6
Published
ACME model execution runtime for provider-neutral AI systems
Readme
acme-engine
Provider-neutral model execution runtime for embedding ACME directly in Node applications.
Install
npm install acme-engineUse
import { createAcmeModelRuntime } from 'acme-engine';acme-engine is the convenience facade for the supported in-process
@acme-engine/model-runtime package. It does not own application cognition,
memory, prompts, tool policy, or model-selection strategy.
Current release: 0.1.5.
Release lineage after the repaired registry baseline:
- 0.1.1 is the first installable registry release of the importable library API.
- 0.1.2 preserves explicit non-stream execution intent.
- 0.1.3 preserves explicit transport timeouts after response start.
- 0.1.4 adds profile-selected Chat Completions output-token wire naming.
- 0.1.6 adds lossless lowering of provably disjoint
oneOfschema branches, includingboolean | string, for the native OpenAI strict structured-output path. - 0.1.5 adds native OpenAI Responses ordered user text/image input.
Version 0.1.0 was published with unresolved workspace protocol dependencies and must not be used.
