Skip to content

v0.4.2

Choose a tag to compare

View all tags
@carefreelife98 carefreelife98 released this 27 Oct 11:49
· 29 commits to main since this release
v0.4.2
26cf8fe

openapi-flow 0.4.2

Changes

  • Presentation reasons are separated from functional validation in API selection,
    request media/literal generation, native selection/dependencies/configuration,
    API-by-call bindings and connection planning. Invalid display text is omitted;
    executable choices, values, source provenance, ownership and DAG checks remain.
  • Goal/native reviews use code-attached original source text instead of asking a
    model to copy quotations. Native review can explicitly cover every selected role,
    including selections without reported gaps.
  • planWorkflowIntent describes business scope before concrete API/native
    selection. Required, conditional and excluded tasks and business flow intentions
    remain separate from actual OAS ports, data paths and execution edges.
  • Explicit producer-owned native settings avoid a redundant model decision when
    the registered capability contract determines the complete configuration.
  • The optional consumer composition settles unused model-call reservations using
    actual calls, preserving the original bounded budget and failure causes.

Packages

Compatibility

Custom model doubles must follow the supplied current output schema. Native review
proposals no longer contain sourceQuote; old review checkpoints are not silently
migrated. Goal-review original quotations are attached by code using actual fact
identities. Review and binding/literal record reasons may be absent: consumers
must handle optional display text instead of assuming a nonempty string. Functional
literal-source verification is retained and is distinct from display quotations.

The new intent planner is an optional primitive, not a package-owned execution
pipeline or retry loop. The host owns when to invoke it and how to preserve its
context across stages, revisions, HITL and checkpoints. It does not replace actual
OAS/native or whole-graph validation, and is not semantic correctness proof.

Verification and limits

Strict type/build/Lint/format checks passed. Enabled tests: 1,793/1,794, with the
retained Ajv failure and no skips/cancellations. Four original OAS files passed
59 checks; three selected isolated engine suites and the strict independent packed
consumer passed. Candidate verification and publication status are tracked in the
release record. Controlled actual-model checks of one unchanged original
scenario passed the new intent/API/native selection and full-role-review path.
That initial check is not a complete JSON/engine/UI result or a five-scenario quality score.

A later fresh host run produced SDK JSON for the same unchanged scenario. The
isolated n8n execution completed ten local GETs and three original-response groups.
However, the model promoted an illustrative scenario value into a concrete input.
Exact source/type checks do not prove that a user selected a value. This result is
not semantic acceptance; UI verification and the independent five-scenario
comparison remain unfinished. A prompt-only experiment did not resolve the issue
and was not adopted or published. Review generated inputs before execution.

The previously disclosed Ajv properties.__proto__ limitation remains; the original
OAS and runtime data are not rewritten. The current upstream SDK dependency audit
reports High 5 and no other severities. No clean-audit or
all-tests-passing claim is made. Company host installation, service deployment and
workflow execution are separate from this library release.

Assets 2
Loading