BCAI/easyai-ai-gateway
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat: record task attempt chains

Browse Source
This commit is contained in:
wangbo
2026-05-11 22:39:45 +08:00
parent 4c2de4b4c9
commit 0049b246c1
9 changed files with 492 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
apps/api/internal/runner/recording.go
+67 -1
View File
@@ -40,7 +40,8 @@ func buildSuccessRecord(task store.GatewayTask, user *auth.User, body map[string
}
func taskMetrics(task store.GatewayTask, user *auth.User, body map[string]any, candidate store.RuntimeModelCandidate, response clients.Response, simulated bool) map[string]any {
metrics := map[string]any{
metrics := attemptMetrics(candidate, 0, simulated)
for key, value := range map[string]any{
"kind": task.Kind,
"runMode": task.RunMode,
"requestedModel": task.Model,
@@ -57,6 +58,8 @@ func taskMetrics(task store.GatewayTask, user *auth.User, body map[string]any, c
"queueKey": candidate.QueueKey,
"requestId": response.RequestID,
"simulated": simulated,
} {
metrics[key] = value
}
if user != nil {
metrics["apiKeyId"] = user.APIKeyID
@@ -94,6 +97,29 @@ func taskMetrics(task store.GatewayTask, user *auth.User, body map[string]any, c
return metrics
}
func attemptMetrics(candidate store.RuntimeModelCandidate, attemptNo int, simulated bool) map[string]any {
metrics := map[string]any{
"resolvedModel": candidate.ModelName,
"modelName": candidate.ModelName,
"modelAlias": candidate.ModelAlias,
"providerModel": candidate.ProviderModelName,
"canonicalModel": candidate.CanonicalModelKey,
"modelType": candidate.ModelType,
"provider": candidate.Provider,
"platformId": candidate.PlatformID,
"platformKey": candidate.PlatformKey,
"platformName": candidate.PlatformName,
"platformModelId": candidate.PlatformModelID,
"clientId": candidate.ClientID,
"queueKey": candidate.QueueKey,
"simulated": simulated,
}
if attemptNo > 0 {
metrics["attempt"] = attemptNo
}
return metrics
}
func usageToMap(usage clients.Usage) map[string]any {
out := map[string]any{}
if usage.InputTokens > 0 {
@@ -172,6 +198,46 @@ func failureMetrics(err error, simulated bool) (string, map[string]any, time.Tim
return meta.RequestID, metrics, meta.ResponseStartedAt, meta.ResponseFinishedAt, meta.ResponseDurationMS
}
func mergeMetrics(values ...map[string]any) map[string]any {
out := map[string]any{}
for _, value := range values {
for key, item := range value {
out[key] = item
}
}
return out
}
func summarizeAttempts(attempts []store.TaskAttempt) []map[string]any {
items := make([]map[string]any, 0, len(attempts))
for _, attempt := range attempts {
item := map[string]any{
"attempt": attempt.AttemptNo,
"status": attempt.Status,
"platformId": attempt.PlatformID,
"platformName": attempt.PlatformName,
"provider": attempt.Provider,
"platformModelId": attempt.PlatformModelID,
"modelName": attempt.ModelName,
"providerModelName": attempt.ProviderModelName,
"modelAlias": attempt.ModelAlias,
"modelType": attempt.ModelType,
"clientId": attempt.ClientID,
"queueKey": attempt.QueueKey,
"requestId": attempt.RequestID,
"retryable": attempt.Retryable,
"simulated": attempt.Simulated,
"errorCode": attempt.ErrorCode,
"errorMessage": attempt.ErrorMessage,
"responseDurationMs": attempt.ResponseDurationMS,
"startedAt": attempt.StartedAt,
"finishedAt": attempt.FinishedAt,
}
items = append(items, item)
}
return items
}
func messageCount(body map[string]any) int {
messages, _ := body["messages"].([]any)
return len(messages)
apps/api/internal/runner/service.go
+24 -4
View File
@@ -85,6 +85,7 @@ func (s *Service) execute(ctx context.Context, task store.GatewayTask, user *aut
if err == nil {
billings := s.billings(ctx, user, task.Kind, body, candidate, response, isSimulation(task, candidate))
record := buildSuccessRecord(task, user, body, candidate, response, billings, isSimulation(task, candidate))
record.Metrics = s.withAttemptHistory(ctx, task.ID, record.Metrics)
finished, finishErr := s.store.FinishTaskSuccess(ctx, store.FinishTaskSuccessInput{
TaskID: task.ID,
Result: response.Result,
@@ -161,6 +162,7 @@ func (s *Service) runCandidate(ctx context.Context, task store.GatewayTask, user
Status: "running",
Simulated: simulated,
RequestSnapshot: body,
Metrics: attemptMetrics(candidate, attemptNo, simulated),
})
if err != nil {
return clients.Response{}, err
@@ -172,7 +174,7 @@ func (s *Service) runCandidate(ctx context.Context, task store.GatewayTask, user
AttemptID: attemptID,
Status: "failed",
Retryable: false,
Metrics: map[string]any{"error": err.Error(), "candidateModel": candidate.ModelName, "clientId": candidate.ClientID},
Metrics: mergeMetrics(attemptMetrics(candidate, attemptNo, simulated), map[string]any{"error": err.Error(), "retryable": false}),
ErrorCode: "rate_limit",
ErrorMessage: err.Error(),
})
@@ -224,6 +226,7 @@ func (s *Service) runCandidate(ctx context.Context, task store.GatewayTask, user
responseDurationMS = 0
}
}
metrics = mergeMetrics(attemptMetrics(candidate, attemptNo, simulated), metrics)
_ = s.store.FinishTaskAttempt(ctx, store.FinishTaskAttemptInput{
AttemptID: attemptID,
Status: "failed",
@@ -242,9 +245,10 @@ func (s *Service) runCandidate(ctx context.Context, task store.GatewayTask, user
}
uploadedResult, err := s.uploadGeneratedAssets(ctx, response.Result)
if err != nil {
metrics := taskMetrics(task, user, body, candidate, response, simulated)
metrics["error"] = err.Error()
metrics["retryable"] = clients.IsRetryable(err)
metrics := mergeMetrics(taskMetrics(task, user, body, candidate, response, simulated), map[string]any{
"error": err.Error(),
"retryable": clients.IsRetryable(err),
})
_ = s.store.FinishTaskAttempt(ctx, store.FinishTaskAttemptInput{
AttemptID: attemptID,
Status: "failed",
@@ -299,6 +303,7 @@ func (s *Service) clientFor(candidate store.RuntimeModelCandidate, simulated boo
func (s *Service) failTask(ctx context.Context, taskID string, code string, message string, simulated bool, cause error) (store.GatewayTask, error) {
requestID, metrics, responseStartedAt, responseFinishedAt, responseDurationMS := failureMetrics(cause, simulated)
metrics = s.withAttemptHistory(ctx, taskID, metrics)
failed, err := s.store.FinishTaskFailure(ctx, store.FinishTaskFailureInput{
TaskID: taskID,
Code: code,
@@ -318,6 +323,21 @@ func (s *Service) failTask(ctx context.Context, taskID string, code string, mess
return failed, nil
}
func (s *Service) withAttemptHistory(ctx context.Context, taskID string, metrics map[string]any) map[string]any {
attempts, err := s.store.ListTaskAttempts(ctx, taskID)
if err != nil {
s.logger.Warn("list task attempts for metrics failed", "taskID", taskID, "error", err)
return metrics
}
if len(attempts) == 0 {
return metrics
}
metrics = mergeMetrics(metrics)
metrics["attemptCount"] = len(attempts)
metrics["attempts"] = summarizeAttempts(attempts)
return metrics
}
func (s *Service) emit(ctx context.Context, taskID string, eventType string, status string, phase string, progress float64, message string, payload map[string]any, simulated bool) error {
event, err := s.store.AddTaskEvent(ctx, taskID, eventType, status, phase, progress, message, payload, simulated)
if err != nil {