185 lines
12 KiB
Go
185 lines
12 KiB
Go
package runner
|
|
|
|
import (
|
|
"context"
|
|
"encoding/json"
|
|
"errors"
|
|
"testing"
|
|
|
|
"gitea.maximumdirect.net/eric/audita/v3/internal/core/config"
|
|
"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/outcomes"
|
|
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/proposal_generation"
|
|
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/proposals"
|
|
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/structuredoutput"
|
|
"gitea.maximumdirect.net/eric/audita/v3/internal/framework/validators"
|
|
)
|
|
|
|
type repairClient struct {
|
|
outputs []string
|
|
calls []structuredoutput.Request
|
|
err error
|
|
}
|
|
|
|
func (c *repairClient) CompleteStructured(_ context.Context, req structuredoutput.Request, out any) (structuredoutput.Response, error) {
|
|
if req.Replay != nil {
|
|
copyReplay := *req.Replay
|
|
copyReplay.Turns = append([]structuredoutput.Message(nil), req.Replay.Turns...)
|
|
req.Replay = ©Replay
|
|
}
|
|
c.calls = append(c.calls, req)
|
|
if c.err != nil {
|
|
return structuredoutput.Response{}, c.err
|
|
}
|
|
value := c.outputs[len(c.calls)-1]
|
|
if err := json.Unmarshal([]byte(value), out); err != nil {
|
|
return structuredoutput.Response{}, err
|
|
}
|
|
return structuredoutput.Response{Content: json.RawMessage(value), FinalContent: value, RepairInstruction: []structuredoutput.Message{{Role: "user", Content: "instruction"}}}, nil
|
|
}
|
|
|
|
func TestRepairBranchesRetargetAndPreserveOtherApprovals(t *testing.T) {
|
|
client := &repairClient{outputs: []string{`{"proposal_ref":2,"action":"revise","correction":{"id":2,"original_text":"uh ","corrected_text":"","confidence":0.9}}`}}
|
|
candidates := []proposals.EnrichedCorrectionProposal{
|
|
{CorrectionProposal: proposals.CorrectionProposal{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: .9}, ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 4}},
|
|
{CorrectionProposal: proposals.CorrectionProposal{TargetSegmentID: 0, OriginalText: "x", CorrectedText: "y", Confidence: .9}, ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 5}},
|
|
}
|
|
input := validateSectionCandidatesInput{Spec: contracts.ModuleRunSpec{ModuleKey: "grammar", RepairClient: client, Settings: config.Stage{ConfidenceThreshold: .8, LLM: config.LLM{CorrectionRepairAttempts: 1}}}, Policy: proposals.ReplacementPolicyRequireUnique, WorkingTranscript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "cat"}, {ID: 2, Text: "uh hello"}}}, Validators: []contracts.Validator{validators.ProposalShapeValidator{}, validators.NoEffectValidator{}, validators.OriginalTextPresenceValidator{}, validators.ConfidenceThresholdValidator{}, validators.NonEmptyResultingSegmentValidator{}}, SectionEnriched: candidates, Exchange: &structuredoutput.Exchange{PromptID: "proposal", FinalContent: "original response"}}
|
|
initial, err := validateSectionCandidates(context.Background(), input)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
result, branches, err := repairCandidates(context.Background(), input, []int{1, 9}, initial, "repair")
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
if len(result.approved) != 2 || result.approved[1].ProposalIndex != 5 || result.approved[1].TargetSegmentID != 2 || result.approved[1].CorrectedText != "" || len(branches) != 1 || branches[0].Outcome != "approved" || len(result.filtered) != 0 {
|
|
t.Fatalf("result=%+v branches=%+v", result, branches)
|
|
}
|
|
var payload proposal_generation.RepairRequest
|
|
if err := json.Unmarshal([]byte(client.calls[0].Variables["RepairRequestJSON"]), &payload); err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
if payload.ProposalRef != 2 || payload.Correction.SegmentReference != 9 {
|
|
t.Fatalf("request: %+v", payload)
|
|
}
|
|
}
|
|
|
|
func TestRepairBranchTerminalOutcomes(t *testing.T) {
|
|
for _, tc := range []struct {
|
|
name string
|
|
outputs []string
|
|
err error
|
|
budget int
|
|
outcome string
|
|
calls int
|
|
}{
|
|
{"withdraw", []string{`{"proposal_ref":1,"action":"withdraw","correction":null}`}, nil, 1, "withdrawn", 1},
|
|
{"exhausted", []string{`{"proposal_ref":1,"action":"revise","correction":{"id":1,"original_text":"cat","corrected_text":"dog","confidence":0.1}}`, `{"proposal_ref":1,"action":"revise","correction":{"id":1,"original_text":"cat","corrected_text":"dog","confidence":0.2}}`}, nil, 2, "exhausted", 2},
|
|
{"wrong reference", []string{`{"proposal_ref":2,"action":"withdraw","correction":null}`}, nil, 1, "unavailable", 1},
|
|
{"provider failure", nil, &structuredoutput.Failure{Class: structuredoutput.FailureRecoverable, Reason: outcomes.ReasonProviderFailed, Message: "provider failed"}, 1, "unavailable", 1},
|
|
} {
|
|
t.Run(tc.name, func(t *testing.T) {
|
|
client := &repairClient{outputs: tc.outputs, err: tc.err}
|
|
candidate := proposals.EnrichedCorrectionProposal{CorrectionProposal: proposals.CorrectionProposal{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: .1}, ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 0}}
|
|
input := validateSectionCandidatesInput{Spec: contracts.ModuleRunSpec{RepairClient: client, Settings: config.Stage{ConfidenceThreshold: .8, LLM: config.LLM{CorrectionRepairAttempts: tc.budget}}}, Policy: proposals.ReplacementPolicyRequireUnique, WorkingTranscript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "cat"}}}, Validators: []contracts.Validator{validators.ProposalShapeValidator{}, validators.NoEffectValidator{}, validators.OriginalTextPresenceValidator{}, validators.ConfidenceThresholdValidator{}}, SectionEnriched: []proposals.EnrichedCorrectionProposal{candidate}, Exchange: &structuredoutput.Exchange{PromptID: "proposal"}}
|
|
initial, err := validateSectionCandidates(context.Background(), input)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
_, branches, err := repairCandidates(context.Background(), input, []int{1}, initial, "repair")
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
if len(branches) != 1 || branches[0].Outcome != tc.outcome || len(client.calls) != tc.calls {
|
|
t.Fatalf("branches=%+v calls=%d", branches, len(client.calls))
|
|
}
|
|
if tc.calls == 2 && len(client.calls[1].Replay.Turns) != 2 {
|
|
t.Fatalf("history: %+v", client.calls[1].Replay.Turns)
|
|
}
|
|
})
|
|
}
|
|
}
|
|
|
|
type repairModule struct{ outcomeModule }
|
|
|
|
func (m repairModule) RepairPromptID() string { return "repair" }
|
|
|
|
func TestRunnerAppliesRevisionsInOriginalOrder(t *testing.T) {
|
|
sectionIndex := 0
|
|
candidates := []proposals.CorrectionProposal{{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: .9}, {TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "fox", Confidence: .1}}
|
|
client := &repairClient{outputs: []string{`{"proposal_ref":2,"action":"revise","correction":{"id":1,"original_text":"cat","corrected_text":"fox","confidence":0.9}}`}}
|
|
module := repairModule{outcomeModule{fakeModule: fakeModule{key: "grammar", policy: proposals.ReplacementPolicyRequireUnique, validators: []contracts.Validator{validators.ProposalShapeValidator{}, validators.NoEffectValidator{}, validators.OriginalTextPresenceValidator{}, validators.ConfidenceThresholdValidator{}}}, proposeResult: func(req contracts.ProposalRequest) (contracts.ProposalResult, error) {
|
|
return contracts.ProposalResult{Proposals: candidates, SegmentReferences: []int{1, 1}, Exchange: &structuredoutput.Exchange{PromptID: "proposal"}}, nil
|
|
}}}
|
|
runner := New(fakeFactory{modules: map[string]contracts.TranscriptModule{"grammar": module}})
|
|
out, err := runner.Run(context.Background(), RunInput{Transcript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "cat"}}}, ModuleSpecs: []contracts.ModuleRunSpec{{ModuleKey: "grammar", InstanceName: "grammar", RepairClient: client, Settings: config.Stage{ConfidenceThreshold: .8, LLM: config.LLM{CorrectionRepairAttempts: 1}}}}})
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
if out.FinalTranscript.Segments[0].Text != "dog" || len(out.WorkUnits) != 1 || len(out.WorkUnits[0].Repairs) != 1 || out.WorkUnits[0].Repairs[0].Outcome != "approved_skipped" || out.WorkUnits[0].CorrectionsApplied != 1 || len(out.ModuleResults[0].SkippedChanges) != 1 || out.ModuleResults[0].SkippedChanges[0].ProposalIndex != 1 {
|
|
t.Fatalf("output=%+v section=%d", out, sectionIndex)
|
|
}
|
|
if len(out.WorkUnits[0].Operations) != 1 || out.WorkUnits[0].Operations[0].Reference.Kind != outcomes.OperationCorrectionRepair || len(out.WorkUnits[0].Operations[0].RecoveryReasons) != 1 || out.WorkUnits[0].Operations[0].RecoveryReasons[0] != outcomes.RecoveryCorrectionRevised {
|
|
t.Fatalf("repair operations: %+v", out.WorkUnits[0].Operations)
|
|
}
|
|
}
|
|
|
|
func TestRepairRevalidatesEarlierApprovalAndRetainsUnrelatedApprovalOnFailure(t *testing.T) {
|
|
candidates := []proposals.EnrichedCorrectionProposal{
|
|
{CorrectionProposal: proposals.CorrectionProposal{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: .9}, ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 0}},
|
|
{CorrectionProposal: proposals.CorrectionProposal{TargetSegmentID: 1, OriginalText: "bird", CorrectedText: "owl", Confidence: .1}, ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 1}},
|
|
}
|
|
for _, tc := range []struct {
|
|
name string
|
|
failure bool
|
|
}{{"revision", false}, {"provider failure", true}} {
|
|
t.Run(tc.name, func(t *testing.T) {
|
|
calls := 0
|
|
check := fakeValidator{name: "custom", validateF: func(req contracts.ValidationRequest) (validators.Result, error) {
|
|
var decisions []validators.Decision
|
|
for _, candidate := range req.CandidateProposal {
|
|
calls++
|
|
decisions = append(decisions, validators.Decision{ProposalIndex: candidate.ProposalIndex, Approved: true, Disposition: validators.DispositionApproved})
|
|
}
|
|
return validators.Result{Decisions: decisions}, nil
|
|
}}
|
|
client := &repairClient{outputs: []string{`{"proposal_ref":2,"action":"revise","correction":{"id":1,"original_text":"bird","corrected_text":"owl","confidence":0.9}}`}}
|
|
if tc.failure {
|
|
client.err = &structuredoutput.Failure{Class: structuredoutput.FailureRecoverable, Reason: outcomes.ReasonProviderFailed, Message: "failed"}
|
|
}
|
|
input := validateSectionCandidatesInput{Spec: contracts.ModuleRunSpec{RepairClient: client, Settings: config.Stage{ConfidenceThreshold: .8, LLM: config.LLM{CorrectionRepairAttempts: 1}}}, Policy: proposals.ReplacementPolicyRequireUnique, WorkingTranscript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "cat bird"}}}, Validators: []contracts.Validator{check, validators.ConfidenceThresholdValidator{}}, SectionEnriched: candidates, Exchange: &structuredoutput.Exchange{PromptID: "proposal"}}
|
|
initial, err := validateSectionCandidates(context.Background(), input)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
got, branches, err := repairCandidates(context.Background(), input, []int{1, 1}, initial, "repair")
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
if tc.failure {
|
|
if len(got.approved) != 1 || got.approved[0].ProposalIndex != 0 || branches[0].Outcome != "unavailable" || calls != 2 {
|
|
t.Fatalf("result=%+v calls=%d", got, calls)
|
|
}
|
|
} else if len(got.approved) != 2 || branches[0].Outcome != "approved" || calls != 3 {
|
|
t.Fatalf("result=%+v calls=%d", got, calls)
|
|
}
|
|
})
|
|
}
|
|
}
|
|
|
|
func TestRepairInterruptionRetainsAttempt(t *testing.T) {
|
|
client := &repairClient{err: &structuredoutput.Failure{Class: structuredoutput.FailureCanceled, Reason: outcomes.ReasonCanceled, Message: "canceled", Cause: context.Canceled}}
|
|
candidate := proposals.EnrichedCorrectionProposal{CorrectionProposal: proposals.CorrectionProposal{TargetSegmentID: 1, OriginalText: "cat", CorrectedText: "dog", Confidence: .1}, ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 0}}
|
|
input := validateSectionCandidatesInput{Spec: contracts.ModuleRunSpec{RepairClient: client, Settings: config.Stage{ConfidenceThreshold: .8, LLM: config.LLM{CorrectionRepairAttempts: 1}}}, Policy: proposals.ReplacementPolicyRequireUnique, WorkingTranscript: &schema.Transcript{Segments: []schema.Segment{{ID: 1, Text: "cat"}}}, Validators: []contracts.Validator{validators.ConfidenceThresholdValidator{}}, SectionEnriched: []proposals.EnrichedCorrectionProposal{candidate}, Exchange: &structuredoutput.Exchange{PromptID: "proposal"}}
|
|
initial, err := validateSectionCandidates(context.Background(), input)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
_, branches, err := repairCandidates(context.Background(), input, []int{1}, initial, "repair")
|
|
if !errors.Is(err, context.Canceled) || len(branches) != 1 || branches[0].Outcome != "interrupted" || len(branches[0].Attempts) != 1 {
|
|
t.Fatalf("branches=%+v error=%v", branches, err)
|
|
}
|
|
}
|