Documentation
¶
Overview ¶
Package validityeval contains experimental pair validity diagnostics only. PairVerifier does not implement semantic.Verifier and cannot serve an answer, create a semantic alias, or participate in application reuse resolution.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type BatchVerificationResult ¶
type BatchVerificationResult struct {
Decisions []semantic.ValidityResult `json:"decisions"`
Work semantic.ValidityResult `json:"work"`
}
BatchVerificationResult records one diagnostic inference job in Work and zero-incremental-usage decisions in original caller order. Neither is an application reuse resolution or authority to deliver a cached answer.
type PairVerifier ¶
type PairVerifier struct {
DailyTokenBudget *inference.TokenBudget
// contains filtered or unexported fields
}
PairVerifier runs a diagnostic model comparison with durable inference receipts. Decisions are experimental measurements, never serving authority.
func NewPairVerifier ¶
func NewPairVerifier(ctx context.Context, broker inference.Broker, baseURL string, c semantic.ReuseConfig) (*PairVerifier, error)
NewPairVerifier discovers only already-installed literal-loopback weights. It performs no generation and offers no standalone Verify method.
func (*PairVerifier) Close ¶
func (v *PairVerifier) Close()
func (*PairVerifier) Identity ¶
func (v *PairVerifier) Identity() string
func (*PairVerifier) VerifyBatch ¶
func (v *PairVerifier) VerifyBatch(ctx context.Context, requests []semantic.ValidityRequest) (out BatchVerificationResult, err error)
VerifyBatch validates exactly two compatible requests in one durable model job. The six non-Question fields must match exactly after JSON encoding; questions are required to be unique and sorted for canonical job membership. No source facts are omitted. Malformed complete output abstains for every member; unknown work returns no decisions and must not be automatically retried or regrouped.
Source Files