Remove confidence scoring and refine correction prompts

This commit is contained in:
2026-11-11 17:05:52 +00:00
parent bbf89bb4f5
commit d155bf4cb3
92 changed files with 448 additions and 786 deletions

View File

@@ -13,10 +13,8 @@ import (
"time"
"gitea.maximumdirect.net/eric/audita/v3/internal/core/config"
"gitea.maximumdirect.net/eric/audita/v3/internal/core/glossary"
"gitea.maximumdirect.net/eric/audita/v3/internal/core/schema"
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/contracts"
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/modules"
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/outcomes"
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/proposal_generation"
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/proposals"
@@ -57,7 +55,7 @@ func TestRunnerRecoverableProposalFailureAllowsLaterModule(t *testing.T) {
return contracts.ProposalResult{Operations: []outcomes.Operation{{Reference: outcomes.Reference{ModuleKey: "first", ModuleInstance: "first", Kind: outcomes.OperationProposal}, FailureReason: outcomes.ReasonTransportFailed}}}, nil
}}
second := fakeModule{key: "second", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}}, nil
}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{"first": failed, "second": second}})
out, err := r.Run(context.Background(), RunInput{Transcript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "teh"}}}, ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "first", InstanceName: "first"}, {ModuleKey: "second", InstanceName: "second"}}})
@@ -94,9 +92,9 @@ func TestRunnerOnlyFullyValidatedCandidatesApplyAndRecoveriesMerge(t *testing.T)
}}
module := outcomeModule{fakeModule: fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{validators.ProposalShapeValidator{}, partial, last}}, proposeResult: func(req contracts.ProposalRequest) (contracts.ProposalResult, error) {
return contracts.ProposalResult{Proposals: []proposals.CorrectionProposal{
{TargetSegmentID: 0, OriginalText: "bad", CorrectedText: "x", Confidence: 1},
{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1},
{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: 1},
{TargetSegmentID: 0, OriginalText: "bad", CorrectedText: "x"},
{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"},
{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog"},
}, Operations: []outcomes.Operation{{Reference: outcomes.Reference{Kind: outcomes.OperationProposal}, Succeeded: true, RecoveryReasons: []outcomes.RecoveryReason{outcomes.RecoveryOutputRepaired}}}}, nil
}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{"m": module}})
@@ -119,7 +117,7 @@ func TestRunnerOnlyFullyValidatedCandidatesApplyAndRecoveriesMerge(t *testing.T)
func TestFatalValidatorNeverReturnsUncheckedCandidates(t *testing.T) {
origin := errors.New("validator setup failed")
proposal := proposals.EnrichedCorrectionProposal{CorrectionProposal: proposals.CorrectionProposal{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}, ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 7}}
proposal := proposals.EnrichedCorrectionProposal{CorrectionProposal: proposals.CorrectionProposal{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}, ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 7}}
broken := fakeValidator{name: "broken", validateF: func(req contracts.ValidationRequest) (validators.Result, error) {
return validators.Result{Operations: []outcomes.Operation{{Reference: outcomes.Reference{Kind: outcomes.OperationValidation}, FailureReason: outcomes.ReasonSetupFailure}}}, origin
}}
@@ -133,7 +131,7 @@ func TestCandidateCaptureIsExplicitAndIncludesFilteredProposals(t *testing.T) {
module := fakeModule{key: "grammar", policy: proposals.ReplacementPolicyRequireUnique,
validators: []contracts.Validator{validators.ProposalShapeValidator{}},
proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 0, OriginalText: "bad", CorrectedText: "good", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 0, OriginalText: "bad", CorrectedText: "good"}}, nil
}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{"grammar": module}})
input := RunInput{Transcript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "bad"}}}, ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "grammar", InstanceName: "grammar"}}}
@@ -236,7 +234,7 @@ func TestRunnerOneModuleAppliesProposal(t *testing.T) {
transcript := &schema.Transcript{Segments: []schema.Segment{{ID: 1, Speaker: "A", Start: 0, End: 1, Text: "teh cat"}}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"grammar": fakeModule{key: "grammar", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 0.9}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}}, nil
}},
}})
@@ -256,13 +254,13 @@ func TestRunnerModulesRunSequentially(t *testing.T) {
transcript := &schema.Transcript{Segments: []schema.Segment{{ID: 1, Speaker: "A", Start: 0, End: 1, Text: "teh cat"}}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m1": fakeModule{key: "m1", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}}, nil
}},
"m2": fakeModule{key: "m2", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
if req.WorkingTranscript.Segments[0].Text != "the cat" {
t.Fatalf("second module did not see first module changes: %q", req.WorkingTranscript.Segments[0].Text)
}
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog"}}, nil
}},
}})
@@ -293,7 +291,7 @@ func TestRunnerKeepsSectionMembershipAfterTextGrowth(t *testing.T) {
if req.Section.Index != 0 {
return nil, nil
}
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "a", CorrectedText: updatedText, Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "a", CorrectedText: updatedText}}, nil
}},
"inspect": fakeModule{key: "inspect", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
ids := make([]int, len(req.WorkingTranscript.Segments))
@@ -307,7 +305,7 @@ func TestRunnerKeepsSectionMembershipAfterTextGrowth(t *testing.T) {
return nil, errors.New("later module did not see updated section text")
}
if req.Section.Index == 0 {
return []proposals.CorrectionProposal{{TargetSegmentID: 3, OriginalText: "c", CorrectedText: "C", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 3, OriginalText: "c", CorrectedText: "C"}}, nil
}
return nil, nil
}},
@@ -449,7 +447,7 @@ func TestRunnerProposalIndexOrderingIsDeterministicAcrossParallelSections(t *tes
started <- seg.ID
<-releaseByID[seg.ID]
return []proposals.CorrectionProposal{
{TargetSegmentID: seg.ID, OriginalText: seg.Text, CorrectedText: strings.ToUpper(seg.Text), Confidence: 1},
{TargetSegmentID: seg.ID, OriginalText: seg.Text, CorrectedText: strings.ToUpper(seg.Text)},
}, nil
}},
}})
@@ -584,7 +582,7 @@ func TestRunnerValidationStartsBeforeAllSectionProposalsComplete(t *testing.T) {
<-releaseSectionOne
}
return []proposals.CorrectionProposal{
{TargetSegmentID: seg.ID, OriginalText: seg.Text, CorrectedText: strings.ToUpper(seg.Text), Confidence: 1},
{TargetSegmentID: seg.ID, OriginalText: seg.Text, CorrectedText: strings.ToUpper(seg.Text)},
}, nil
},
},
@@ -630,7 +628,7 @@ func (c *captureContractStructuredClient) CompleteStructured(ctx context.Context
if target, ok := out.(*validators.LLMValidationResponse); ok {
*target = validators.LLMValidationResponse{
Validations: []validators.LLMValidationDecision{
{CorrectionIndex: 1, Approved: true, Confidence: 0.9, Reason: "ok"},
{CorrectionIndex: 1, Approved: true, Reason: "ok"},
},
}
}
@@ -662,7 +660,7 @@ func (c *stageAwareStructuredClient) CompleteStructured(ctx context.Context, req
target := out.(*validators.LLMValidationResponse)
*target = validators.LLMValidationResponse{
Validations: []validators.LLMValidationDecision{
{CorrectionIndex: 1, Approved: true, Confidence: 0.99, Reason: "ok"},
{CorrectionIndex: 1, Approved: true, Reason: "ok"},
},
}
return contracts.StructuredCompletionResponse{}, nil
@@ -672,7 +670,7 @@ func TestRunnerSkippedRecorded(t *testing.T) {
transcript := &schema.Transcript{Segments: []schema.Segment{{ID: 1, Speaker: "A", Start: 0, End: 1, Text: "word word"}}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "word", CorrectedText: "term", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "word", CorrectedText: "term"}}, nil
}},
}})
out, err := r.Run(context.Background(), RunInput{Transcript: transcript, ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "m", InstanceName: "m"}}})
@@ -688,7 +686,7 @@ func TestRunnerFailureReturnsPartialProgress(t *testing.T) {
transcript := &schema.Transcript{Segments: []schema.Segment{{ID: 1, Speaker: "A", Start: 0, End: 1, Text: "teh cat"}}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m1": fakeModule{key: "m1", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}}, nil
}},
"m2": fakeModule{key: "m2", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return nil, errors.New("boom")
@@ -712,7 +710,7 @@ func TestRunnerDoesNotMutateInputTranscript(t *testing.T) {
before := transcript.Segments[0].Text
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}}, nil
}},
}})
@@ -747,7 +745,7 @@ func TestRunnerValidatorApprovedProposalApplied(t *testing.T) {
policy: proposals.ReplacementPolicyRequireUnique,
validators: []contracts.Validator{allowAll},
proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}}, nil
},
},
}})
@@ -775,7 +773,7 @@ func TestRunnerValidatorRejectedProposalNotApplied(t *testing.T) {
}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{rejectAll}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}}, nil
}},
}})
out, err := r.Run(context.Background(), RunInput{Transcript: transcript, ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "m", InstanceName: "m"}}})
@@ -813,8 +811,8 @@ func TestRunnerMultipleValidatorsRunInOrderAndFilterSurvivors(t *testing.T) {
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{first, second}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{
{TargetSegmentID: 1, OriginalText: "one", CorrectedText: "ONE", Confidence: 1},
{TargetSegmentID: 1, OriginalText: "two", CorrectedText: "TWO", Confidence: 1},
{TargetSegmentID: 1, OriginalText: "one", CorrectedText: "ONE"},
{TargetSegmentID: 1, OriginalText: "two", CorrectedText: "TWO"},
}, nil
}},
}})
@@ -831,14 +829,14 @@ func TestRunnerMultipleValidatorsRunInOrderAndFilterSurvivors(t *testing.T) {
func TestRunnerValidatorCardinalityErrorStopsPipelineWithPartialProgress(t *testing.T) {
transcript := &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "teh cat"}}}
good := fakeModule{key: "m1", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}}, nil
}}
badValidator := fakeValidator{name: "bad", validateF: func(req contracts.ValidationRequest) (validators.Result, error) {
// Missing one decision triggers cardinality error.
return validators.Result{ValidatorName: "bad", Decisions: nil}, nil
}}
bad := fakeModule{key: "m2", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{badValidator}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog"}}, nil
}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{"m1": good, "m2": bad}})
@@ -861,7 +859,7 @@ func TestRunnerApplicationSkipAfterValidatorApprovalReported(t *testing.T) {
}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{allow}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "word", CorrectedText: "term", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "word", CorrectedText: "term"}}, nil
}},
}})
out, err := r.Run(context.Background(), RunInput{Transcript: transcript, ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "m", InstanceName: "m"}}})
@@ -884,8 +882,8 @@ func TestRunnerApplyConflictAfterTwoInitiallyValidDeletions(t *testing.T) {
proposeResult: func(req contracts.ProposalRequest) (contracts.ProposalResult, error) {
return contracts.ProposalResult{
Proposals: []proposals.CorrectionProposal{
{TargetSegmentID: 1, OriginalText: "a", CorrectedText: "", Confidence: 1},
{TargetSegmentID: 1, OriginalText: "b", CorrectedText: "", Confidence: 1},
{TargetSegmentID: 1, OriginalText: "a", CorrectedText: ""},
{TargetSegmentID: 1, OriginalText: "b", CorrectedText: ""},
},
Operations: []outcomes.Operation{{Reference: outcomes.Reference{Kind: outcomes.OperationProposal}, Succeeded: true}},
}, nil
@@ -953,7 +951,7 @@ func (f *fakeStructuredClient) CompleteStructured(ctx context.Context, req contr
}
func TestRunnerLLMValidatorApprovalApplied(t *testing.T) {
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: true, Confidence: 0.9, Reason: "ok"}}}}}
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: true, Reason: "ok"}}}}}
llmValidator, err := validators.NewLLMBackedValidator("spoken_form_plausibility_review", validators.LLMValidatorTypeSpokenFormPlausibility)
if err != nil {
t.Fatalf("NewLLMBackedValidator: %v", err)
@@ -961,7 +959,7 @@ func TestRunnerLLMValidatorApprovalApplied(t *testing.T) {
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{llmValidator}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters"}}, nil
}},
}})
cfg := runnerTestConfig()
@@ -979,11 +977,11 @@ func TestRunnerLLMValidatorApprovalApplied(t *testing.T) {
}
func TestRunnerLLMValidatorRejectionPreventsApplication(t *testing.T) {
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: false, Confidence: 0.95, Reason: "reject"}}}}}
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: false, Reason: "reject"}}}}}
llmValidator, _ := validators.NewLLMBackedValidator("spoken_form_plausibility_review", validators.LLMValidatorTypeSpokenFormPlausibility)
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{llmValidator}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters"}}, nil
}},
}})
cfg := runnerTestConfig()
@@ -1004,14 +1002,14 @@ func TestRunnerLLMValidatorRejectionPreventsApplication(t *testing.T) {
}
func TestRunnerLLMValidatorMalformedResponseRejectsBatchAndKeepsPartialProgress(t *testing.T) {
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 99, Approved: true, Confidence: 0.9, Reason: "bad index"}}}}}
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 99, Approved: true, Reason: "bad index"}}}}}
llmValidator, _ := validators.NewLLMBackedValidator("spoken_form_plausibility_review", validators.LLMValidatorTypeSpokenFormPlausibility)
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m1": fakeModule{key: "m1", policy: proposals.ReplacementPolicyRequireUnique, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"}}, nil
}},
"m2": fakeModule{key: "m2", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{llmValidator}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog"}}, nil
}},
}})
cfg := runnerTestConfig()
@@ -1036,9 +1034,9 @@ func TestRunnerLLMValidatorMalformedResponseRejectsBatchAndKeepsPartialProgress(
func TestRunnerReconciledSurvivorsCompleteLaterValidation(t *testing.T) {
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{
{CorrectionIndex: 1, Approved: true, Confidence: .9, Reason: "ok"},
{CorrectionIndex: 3, Approved: true, Confidence: .9, Reason: "ok"},
{CorrectionIndex: 99, Approved: true, Confidence: .9, Reason: "unexpected"},
{CorrectionIndex: 1, Approved: true, Reason: "ok"},
{CorrectionIndex: 3, Approved: true, Reason: "ok"},
{CorrectionIndex: 99, Approved: true, Reason: "unexpected"},
}}}}
first, _ := validators.NewLLMBackedValidator("editorial_review", validators.LLMValidatorTypeEditorialReview)
var laterIndices []int
@@ -1057,9 +1055,9 @@ func TestRunnerReconciledSurvivorsCompleteLaterValidation(t *testing.T) {
}}
module := outcomeModule{fakeModule: fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{first, validatormetadata.Wrap(later, validatormetadata.ExecutionClassLLMBacked)}}, proposeResult: func(contracts.ProposalRequest) (contracts.ProposalResult, error) {
return contracts.ProposalResult{Proposals: []proposals.CorrectionProposal{
{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 1},
{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: 1},
{TargetSegmentID: 1, OriginalText: "bad", CorrectedText: "good", Confidence: 1},
{TargetSegmentID: 1, OriginalText: "teh", CorrectedText: "the"},
{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog"},
{TargetSegmentID: 1, OriginalText: "bad", CorrectedText: "good"},
}, Operations: []outcomes.Operation{{Reference: outcomes.Reference{Kind: outcomes.OperationProposal}, Succeeded: true}}}, nil
}}
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{"m": module}})
@@ -1086,7 +1084,7 @@ func TestRunnerLLMValidatorMissingDecisionRejectsBatch(t *testing.T) {
llmValidator, _ := validators.NewLLMBackedValidator("spoken_form_plausibility_review", validators.LLMValidatorTypeSpokenFormPlausibility)
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{llmValidator}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters"}}, nil
}},
}})
cfg := runnerTestConfig()
@@ -1102,13 +1100,13 @@ func TestRunnerLLMValidatorMissingDecisionRejectsBatch(t *testing.T) {
func TestRunnerLLMValidatorDuplicateDecisionRejectsBatch(t *testing.T) {
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{
{CorrectionIndex: 1, Approved: true, Confidence: 0.9, Reason: "ok"},
{CorrectionIndex: 1, Approved: false, Confidence: 0.9, Reason: "dup"},
{CorrectionIndex: 1, Approved: true, Reason: "ok"},
{CorrectionIndex: 1, Approved: false, Reason: "dup"},
}}}}
llmValidator, _ := validators.NewLLMBackedValidator("spoken_form_plausibility_review", validators.LLMValidatorTypeSpokenFormPlausibility)
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{llmValidator}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters"}}, nil
}},
}})
cfg := runnerTestConfig()
@@ -1128,8 +1126,8 @@ func (lenEstimator) EstimateTokens(text string) int { return len(text) }
func TestRunnerLLMValidatorBatching(t *testing.T) {
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{
{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: true, Confidence: 0.9, Reason: "ok"}}},
{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: true, Confidence: 0.9, Reason: "ok"}}},
{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: true, Reason: "ok"}}},
{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: true, Reason: "ok"}}},
}}
llmValidator, _ := validators.NewLLMBackedValidator("spoken_form_plausibility_review", validators.LLMValidatorTypeSpokenFormPlausibility)
llmValidator.SetTokenEstimator(lenEstimator{})
@@ -1139,8 +1137,8 @@ func TestRunnerLLMValidatorBatching(t *testing.T) {
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyReplaceAll, validators: []contracts.Validator{llmValidator}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{
{TargetSegmentID: 1, OriginalText: "alpha", CorrectedText: strings.Repeat("B", 40), Confidence: 1},
{TargetSegmentID: 1, OriginalText: "gamma", CorrectedText: strings.Repeat("D", 40), Confidence: 1},
{TargetSegmentID: 1, OriginalText: "alpha", CorrectedText: strings.Repeat("B", 40)},
{TargetSegmentID: 1, OriginalText: "gamma", CorrectedText: strings.Repeat("D", 40)},
}, nil
}},
}})
@@ -1164,7 +1162,7 @@ func TestRunnerLLMValidatorBatching(t *testing.T) {
func TestRunnerLLMValidatorDiagnosticsWrittenAndRedacted(t *testing.T) {
primarySecret := "runner-primary-secret"
validationSecret := "runner-validation-secret"
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: true, Confidence: 0.9, Reason: validationSecret}}}}}
client := &fakeStructuredClient{responses: []validators.LLMValidationResponse{{Validations: []validators.LLMValidationDecision{{CorrectionIndex: 1, Approved: true, Reason: validationSecret}}}}}
llmValidator, _ := validators.NewLLMBackedValidator("spoken_form_plausibility_review", validators.LLMValidatorTypeSpokenFormPlausibility)
client.secrets = []string{primarySecret}
client.preparedSecret = validationSecret
@@ -1172,7 +1170,7 @@ func TestRunnerLLMValidatorDiagnosticsWrittenAndRedacted(t *testing.T) {
diagDir := t.TempDir()
r := New(fakeFactory{modules: map[string]contracts.TranscriptModule{
"m": fakeModule{key: "m", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{llmValidator}, proposeF: func(req contracts.ProposalRequest) ([]proposals.CorrectionProposal, error) {
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters", Confidence: 1}}, nil
return []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "gestures", CorrectedText: "Jesters"}}, nil
}},
}})
out, err := r.Run(context.Background(), RunInput{
@@ -1262,7 +1260,7 @@ func TestRunnerProposalGenerationHelperFlowsThroughPipeline(t *testing.T) {
responses: []proposal_generation.StructuredCorrectionSet{
{
Corrections: []proposal_generation.StructuredCorrectionProposal{
{SegmentReference: 1, OriginalText: "teh", CorrectedText: "the", Confidence: 0.9},
{SegmentReference: 1, OriginalText: "teh", CorrectedText: "the"},
},
},
},
@@ -1278,7 +1276,7 @@ func TestRunnerProposalGenerationHelperFlowsThroughPipeline(t *testing.T) {
out, err := r.Run(context.Background(), RunInput{
Config: &cfg,
Transcript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "teh cat"}}},
ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "m", InstanceName: "m", ProposalClient: client, Settings: config.Stage{ConfidenceThreshold: cfg.Thresholds.Grammar}}},
ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "m", InstanceName: "m", ProposalClient: client, Settings: config.Stage{}}},
ProposalDiagnosticsDir: diagDir,
})
if err != nil {
@@ -1299,70 +1297,6 @@ func TestRunnerProposalGenerationHelperFlowsThroughPipeline(t *testing.T) {
}
}
func TestRunnerGrammarModuleUsesConfidenceThreshold(t *testing.T) {
client := &fakeProposalStructuredClient{
responses: []proposal_generation.StructuredCorrectionSet{
{
Corrections: []proposal_generation.StructuredCorrectionProposal{
{SegmentReference: 1, OriginalText: "hello ,world", CorrectedText: "Hello, world", Confidence: 0.5},
},
},
},
}
cfg := runnerTestConfig()
cfg.Thresholds.Grammar = 0.9
factory := modules.NewFactory()
out, err := New(factory).Run(context.Background(), RunInput{
Config: &cfg,
Transcript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "hello ,world"}}},
ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "grammar", InstanceName: "grammar", ProposalClient: client, Settings: config.Stage{ConfidenceThreshold: cfg.Thresholds.Grammar}}},
})
if err != nil {
t.Fatalf("Run error: %v", err)
}
if out.FinalTranscript.Segments[0].Text != "hello ,world" {
t.Fatalf("expected no changes due to confidence threshold, got %q", out.FinalTranscript.Segments[0].Text)
}
if len(out.ModuleResults) != 1 || len(out.ModuleResults[0].ValidatorRejected) == 0 {
t.Fatalf("expected validator rejection, got %+v", out.ModuleResults)
}
if out.ModuleResults[0].ValidatorRejected[0].ReasonCode != validators.ReasonLowConfidence {
t.Fatalf("expected low confidence reason, got %+v", out.ModuleResults[0].ValidatorRejected[0])
}
}
func TestRunnerGlossaryModuleUsesConfidenceThreshold(t *testing.T) {
client := &fakeProposalStructuredClient{
responses: []proposal_generation.StructuredCorrectionSet{
{
Corrections: []proposal_generation.StructuredCorrectionProposal{
{SegmentReference: 1, OriginalText: "gestures", CorrectedText: "Jesters", Confidence: 0.5},
},
},
},
}
cfg := runnerTestConfig()
cfg.Thresholds.Glossary = 0.9
factory := modules.NewFactory()
out, err := New(factory).Run(context.Background(), RunInput{
Config: &cfg,
Transcript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "There were gestures"}}},
ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "glossary", InstanceName: "glossary", Glossary: []glossary.Entry{{Name: "Jesters", Category: "faction", Description: "Faction"}}, ProposalClient: client, Settings: config.Stage{ConfidenceThreshold: cfg.Thresholds.Glossary}}},
})
if err != nil {
t.Fatalf("Run error: %v", err)
}
if out.FinalTranscript.Segments[0].Text != "There were gestures" {
t.Fatalf("expected no changes due to glossary confidence threshold, got %q", out.FinalTranscript.Segments[0].Text)
}
if len(out.ModuleResults) != 1 || len(out.ModuleResults[0].ValidatorRejected) == 0 {
t.Fatalf("expected validator rejection, got %+v", out.ModuleResults)
}
if out.ModuleResults[0].ValidatorRejected[0].ReasonCode != validators.ReasonLowConfidence {
t.Fatalf("expected low confidence reason, got %+v", out.ModuleResults[0].ValidatorRejected[0])
}
}
func TestFatalPartialDecisionsRejectUnknownAndDuplicateIdentity(t *testing.T) {
origin := errors.New("original fatal failure")
candidate := proposals.EnrichedCorrectionProposal{ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 7}}