feat: expose streaming test run results

2026-04-07 18:30:08 +03:00
parent 35e9c8c754
commit 1a4f0ea6f3
9 changed files with 865 additions and 91 deletions
@@ -2,18 +2,18 @@

 ## Current

-### `feat/websocket-wizard-foundation`
+### `feat/streaming-test-run-results`

 Status: completed

 DoD:
- wizard step 1 exposes WebSocket as a selectable protocol
- wizard step 3 collects WebSocket endpoint, subprotocols and subscribe/unsubscribe templates
- execution config serializes `ProtocolOptions.websocket` runtime tuning
+- `POST /operations/{operation_id}/test-runs` returns mode-aware streaming metadata
+- window test-runs expose bounded result flags for the wizard status block
+- session and async-job test-runs create persisted runtime state and return references for follow-up pages

 ## Next

- `feat/live-authenticated-staging-entry`
+- `feat/soap-test-run-polish`

 ## Backlog

@@ -221,9 +221,10 @@ mod tests {
    use axum::{Json, Router, routing::post};
    use crank_adapter_grpc::test_support as grpc_test_support;
    use crank_core::{
-        DescriptorId, ExecutionConfig, GraphqlOperationType, GraphqlTarget, GrpcTarget, HttpMethod,
-        MembershipRole, Protocol, RestTarget, SecretKind, SoapBindingStyle, SoapOperationMetadata,
-        SoapTarget, SoapVersion, Target, ToolDescription, WorkspaceId,
+        DescriptorId, ExecutionConfig, ExecutionMode, GraphqlOperationType, GraphqlTarget,
+        GrpcTarget, HttpMethod, MembershipRole, Protocol, RestTarget, SecretKind, SoapBindingStyle,
+        SoapOperationMetadata, SoapTarget, SoapVersion, Target, ToolDescription, TransportBehavior,
+        WorkspaceId,
    };
    use crank_mapping::{MappingRule, MappingSet};
    use crank_registry::PostgresRegistry;
@@ -1441,6 +1442,177 @@ mod tests {
        assert_eq!(test_run["response_preview"]["message"], "hello");
    }

+    #[tokio::test(flavor = "multi_thread")]
+    #[serial]
+    async fn returns_window_metadata_for_streaming_test_runs() {
+        let registry = test_registry().await;
+        let storage_root = test_storage_root("grpc_window_test_run");
+        let server_addr = grpc_test_support::spawn_unary_echo_server().await;
+        let base_url = spawn_admin_api(build_test_app(registry, storage_root)).await;
+        let client = authorized_client(&base_url).await;
+
+        let created = client
+            .post(format!("{base_url}/operations"))
+            .json(&test_grpc_window_operation_payload(
+                &server_addr,
+                "echo_window_runtime",
+            ))
+            .send()
+            .await
+            .unwrap()
+            .json::<Value>()
+            .await
+            .unwrap();
+        let operation_id = created["operation_id"].as_str().unwrap().to_owned();
+
+        let test_run = client
+            .post(format!("{base_url}/operations/{operation_id}/test-runs"))
+            .json(&json!({
+                "version": 1,
+                "input": { "message": "hello" }
+            }))
+            .send()
+            .await
+            .unwrap()
+            .json::<Value>()
+            .await
+            .unwrap();
+
+        assert_eq!(test_run["ok"], true);
+        assert_eq!(test_run["mode"], "window");
+        assert!(test_run["window"].is_object());
+        assert!(test_run["window"]["window_complete"].is_boolean());
+        assert!(test_run["window"]["truncated"].is_boolean());
+        assert!(test_run["window"]["has_more"].is_boolean());
+        assert!(test_run["response_preview"]["items"].is_array());
+    }
+
+    #[tokio::test(flavor = "multi_thread")]
+    #[serial]
+    async fn starts_stream_session_from_streaming_test_runs() {
+        let registry = test_registry().await;
+        let storage_root = test_storage_root("grpc_session_test_run");
+        let server_addr = grpc_test_support::spawn_unary_echo_server().await;
+        let base_url = spawn_admin_api(build_test_app(registry, storage_root)).await;
+        let client = authorized_client(&base_url).await;
+
+        let created = client
+            .post(format!("{base_url}/operations"))
+            .json(&test_grpc_session_operation_payload(
+                &server_addr,
+                "echo_session_runtime",
+            ))
+            .send()
+            .await
+            .unwrap()
+            .json::<Value>()
+            .await
+            .unwrap();
+        let operation_id = created["operation_id"].as_str().unwrap().to_owned();
+
+        let test_run = client
+            .post(format!("{base_url}/operations/{operation_id}/test-runs"))
+            .json(&json!({
+                "version": 1,
+                "input": { "message": "hello" }
+            }))
+            .send()
+            .await
+            .unwrap()
+            .json::<Value>()
+            .await
+            .unwrap();
+        let session_id = test_run["stream_session"]["session_id"]
+            .as_str()
+            .unwrap()
+            .to_owned();
+
+        let session = client
+            .get(format!("{base_url}/stream-sessions/{session_id}"))
+            .send()
+            .await
+            .unwrap()
+            .json::<Value>()
+            .await
+            .unwrap();
+
+        assert_eq!(test_run["ok"], true);
+        assert_eq!(test_run["mode"], "session");
+        assert_eq!(test_run["stream_session"]["status"], "running");
+        assert!(test_run["response_preview"]["preview"]["items"].is_array());
+        assert_eq!(session["id"], session_id);
+        assert_eq!(session["status"], "running");
+    }
+
+    #[tokio::test(flavor = "multi_thread")]
+    #[serial]
+    async fn starts_async_job_from_streaming_test_runs() {
+        let registry = test_registry().await;
+        let storage_root = test_storage_root("async_job_test_run");
+        let upstream_base_url = spawn_upstream_server().await;
+        let base_url = spawn_admin_api(build_test_app(registry, storage_root)).await;
+        let client = authorized_client(&base_url).await;
+
+        let created = client
+            .post(format!("{base_url}/operations"))
+            .json(&test_rest_async_job_operation_payload(
+                &upstream_base_url,
+                "crm_async_job_runtime",
+            ))
+            .send()
+            .await
+            .unwrap()
+            .json::<Value>()
+            .await
+            .unwrap();
+        let operation_id = created["operation_id"].as_str().unwrap().to_owned();
+
+        let test_run = client
+            .post(format!("{base_url}/operations/{operation_id}/test-runs"))
+            .json(&json!({
+                "version": 1,
+                "input": { "email": "user@example.com" }
+            }))
+            .send()
+            .await
+            .unwrap()
+            .json::<Value>()
+            .await
+            .unwrap();
+        let job_id = test_run["async_job"]["job_id"].as_str().unwrap().to_owned();
+
+        let mut job = Value::Null;
+        for _ in 0..20 {
+            job = client
+                .get(format!("{base_url}/async-jobs/{job_id}"))
+                .send()
+                .await
+                .unwrap()
+                .json::<Value>()
+                .await
+                .unwrap();
+            if job["status"] == "completed" {
+                break;
+            }
+            tokio::time::sleep(std::time::Duration::from_millis(25)).await;
+        }
+
+        let result = client
+            .get(format!("{base_url}/async-jobs/{job_id}/result"))
+            .send()
+            .await
+            .unwrap()
+            .json::<Value>()
+            .await
+            .unwrap();
+
+        assert_eq!(test_run["ok"], true);
+        assert_eq!(test_run["mode"], "async_job");
+        assert_eq!(test_run["async_job"]["status"], "running");
+        assert_eq!(job["status"], "completed");
+        assert_eq!(result["id"], "lead_123");
+    }
+
    #[tokio::test(flavor = "multi_thread")]
    #[serial]
    async fn uploads_wsdl_and_tests_soap_operation() {
@@ -2317,6 +2489,98 @@ mod tests {
        }
    }

+    fn test_grpc_window_operation_payload(server_addr: &str, name: &str) -> OperationPayload {
+        let mut payload = test_grpc_operation_payload(server_addr, name);
+        payload.display_name = "Server Echo Window".to_owned();
+        payload.target = Target::Grpc(GrpcTarget {
+            server_addr: server_addr.to_owned(),
+            package: "echo".to_owned(),
+            service: "EchoService".to_owned(),
+            method: "ServerEcho".to_owned(),
+            descriptor_ref: DescriptorId::new("desc_echo_stream"),
+            descriptor_set_b64: grpc_test_support::descriptor_set_b64(),
+        });
+        payload.execution_config.streaming = Some(crank_core::StreamingConfig {
+            mode: ExecutionMode::Window,
+            transport_behavior: TransportBehavior::ServerStream,
+            window_duration_ms: Some(100),
+            poll_interval_ms: None,
+            upstream_timeout_ms: Some(1_000),
+            idle_timeout_ms: None,
+            max_session_lifetime_ms: None,
+            max_items: Some(2),
+            max_bytes: None,
+            aggregation_mode: crank_core::AggregationMode::RawItems,
+            summary_path: None,
+            items_path: Some("$.response.body.items".to_owned()),
+            cursor_path: None,
+            status_path: None,
+            done_path: Some("$.response.body.done".to_owned()),
+            redacted_paths: Vec::new(),
+            truncate_item_fields: false,
+            max_field_length: None,
+            drop_duplicates: false,
+            sampling_rate: None,
+            tool_family: crank_core::ToolFamilyConfig::default(),
+        });
+        payload
+    }
+
+    fn test_grpc_session_operation_payload(server_addr: &str, name: &str) -> OperationPayload {
+        let mut payload = test_grpc_window_operation_payload(server_addr, name);
+        payload.display_name = "Server Echo Session".to_owned();
+        if let Some(streaming) = payload.execution_config.streaming.as_mut() {
+            streaming.mode = ExecutionMode::Session;
+            streaming.poll_interval_ms = Some(1_000);
+            streaming.max_session_lifetime_ms = Some(60_000);
+            streaming.tool_family = crank_core::ToolFamilyConfig {
+                start_tool_name: Some("echo_session_start".to_owned()),
+                poll_tool_name: Some("echo_session_poll".to_owned()),
+                stop_tool_name: Some("echo_session_stop".to_owned()),
+                status_tool_name: None,
+                result_tool_name: None,
+                cancel_tool_name: None,
+            };
+        }
+        payload
+    }
+
+    fn test_rest_async_job_operation_payload(base_url: &str, name: &str) -> OperationPayload {
+        let mut payload = test_operation_payload(base_url, name);
+        payload.display_name = "Create Lead Async Job".to_owned();
+        payload.execution_config.streaming = Some(crank_core::StreamingConfig {
+            mode: ExecutionMode::AsyncJob,
+            transport_behavior: TransportBehavior::DeferredResult,
+            window_duration_ms: None,
+            poll_interval_ms: Some(1_000),
+            upstream_timeout_ms: Some(1_000),
+            idle_timeout_ms: None,
+            max_session_lifetime_ms: Some(300_000),
+            max_items: None,
+            max_bytes: None,
+            aggregation_mode: crank_core::AggregationMode::SummaryPlusSamples,
+            summary_path: None,
+            items_path: None,
+            cursor_path: None,
+            status_path: None,
+            done_path: None,
+            redacted_paths: Vec::new(),
+            truncate_item_fields: false,
+            max_field_length: None,
+            drop_duplicates: false,
+            sampling_rate: None,
+            tool_family: crank_core::ToolFamilyConfig {
+                start_tool_name: Some("crm_async_start".to_owned()),
+                poll_tool_name: None,
+                stop_tool_name: None,
+                status_tool_name: Some("crm_async_status".to_owned()),
+                result_tool_name: Some("crm_async_result".to_owned()),
+                cancel_tool_name: Some("crm_async_cancel".to_owned()),
+            },
+        });
+        payload
+    }
+
    fn test_soap_operation_payload(endpoint: &str, name: &str) -> OperationPayload {
        OperationPayload {
            name: name.to_owned(),
@@ -20,14 +20,15 @@ use crank_core::{
 use crank_mapping::{JsonPathRoot, MappingSet, infer_mapping_from_samples};
 use crank_proto::{ProtoService, services_from_descriptor_set_bytes};
 use crank_registry::{
-    AgentSummary, AgentVersionRecord, AsyncJobFilter, CreateAgentRequest, CreateInvitationRequest,
-    CreateInvocationLogRequest, CreatePlatformApiKeyRequest, CreateSecretRequest,
-    CreateVersionRequest, CreateWorkspaceRequest, InvitationRecord, InvocationLogRecord,
-    ListInvocationLogsQuery, MembershipRecord, OperationAgentRef, OperationSampleMetadata,
-    OperationSummary, OperationUsageSummary, OperationVersionRecord, Page, PlatformApiKeyRecord,
-    PostgresRegistry, PublishAgentRequest, PublishRequest, RegistryError, RegistryOperation,
-    RotateSecretRequest, SampleKind, SaveAgentBindingsRequest, SaveAuthProfileRequest,
-    SaveDescriptorMetadataRequest, SaveSampleMetadataRequest, StreamSessionFilter,
+    AgentSummary, AgentVersionRecord, AsyncJobFilter, CreateAgentRequest, CreateAsyncJobRequest,
+    CreateInvitationRequest, CreateInvocationLogRequest, CreatePlatformApiKeyRequest,
+    CreateSecretRequest, CreateStreamSessionRequest, CreateVersionRequest, CreateWorkspaceRequest,
+    InvitationRecord, InvocationLogRecord, ListInvocationLogsQuery, MembershipRecord,
+    OperationAgentRef, OperationSampleMetadata, OperationSummary, OperationUsageSummary,
+    OperationVersionRecord, Page, PlatformApiKeyRecord, PostgresRegistry, PublishAgentRequest,
+    PublishRequest, RegistryError, RegistryOperation, RotateSecretRequest, SampleKind,
+    SaveAgentBindingsRequest, SaveAuthProfileRequest, SaveDescriptorMetadataRequest,
+    SaveSampleMetadataRequest, StreamSessionFilter, UpdateAsyncJobStatusRequest,
    UpdateWorkspaceRequest, UsageAgentBreakdown, UsageBucket, UsageOperationBreakdown, UsageQuery,
    UsageSummary, UsageTimelinePoint, WorkspaceMembershipRecord, WorkspaceRecord,
 };
@@ -128,9 +129,123 @@ pub struct TestRunPayload {
 #[derive(Clone, Debug, Serialize)]
 pub struct TestRunResult {
    pub ok: bool,
+    pub mode: ExecutionMode,
    pub request_preview: Value,
    pub response_preview: Value,
    pub errors: Vec<Value>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub window: Option<WindowTestRunView>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub stream_session: Option<StreamSessionStartView>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub async_job: Option<AsyncJobStartView>,
+}
+
+#[derive(Clone, Debug, Serialize)]
+pub struct WindowTestRunView {
+    pub window_complete: bool,
+    pub truncated: bool,
+    pub has_more: bool,
+    pub cursor: Option<Value>,
+}
+
+#[derive(Clone, Debug, Serialize)]
+pub struct StreamSessionStartView {
+    pub session_id: String,
+    pub status: StreamStatus,
+    pub expires_at: String,
+    pub poll_after_ms: u64,
+    pub preview: Value,
+}
+
+#[derive(Clone, Debug, Serialize)]
+pub struct AsyncJobStartView {
+    pub job_id: String,
+    pub status: JobStatus,
+    pub progress: Value,
+}
+
+#[derive(Debug, Serialize, Deserialize)]
+struct StoredSessionState {
+    input: Value,
+    summary: Value,
+    items: Vec<Value>,
+    next_index: usize,
+    batch_size: usize,
+}
+
+enum TestRunOutcome {
+    Unary {
+        output: Value,
+    },
+    Window {
+        output: crank_runtime::WindowExecutionResult,
+    },
+    Session {
+        output: StreamSessionStartView,
+    },
+    AsyncJob {
+        output: AsyncJobStartView,
+    },
+}
+
+impl TestRunOutcome {
+    fn into_result_views(
+        self,
+    ) -> (
+        &'static str,
+        Value,
+        Option<WindowTestRunView>,
+        Option<StreamSessionStartView>,
+        Option<AsyncJobStartView>,
+    ) {
+        match self {
+            Self::Unary { output } => ("admin test run completed", output, None, None, None),
+            Self::Window { output } => (
+                "admin window test run completed",
+                json!({
+                    "summary": output.summary,
+                    "items": output.items,
+                    "cursor": output.cursor,
+                    "window_complete": output.window_complete,
+                    "truncated": output.truncated,
+                    "has_more": output.has_more,
+                }),
+                Some(WindowTestRunView {
+                    window_complete: output.window_complete,
+                    truncated: output.truncated,
+                    has_more: output.has_more,
+                    cursor: output.cursor,
+                }),
+                None,
+                None,
+            ),
+            Self::Session { output } => (
+                "admin session test run started",
+                json!({
+                    "session_id": output.session_id,
+                    "status": output.status,
+                    "expires_at": output.expires_at,
+                    "poll_after_ms": output.poll_after_ms,
+                    "preview": output.preview,
+                }),
+                None,
+                Some(output),
+                None,
+            ),
+            Self::AsyncJob { output } => (
+                "admin async job test run started",
+                json!({
+                    "job_id": output.job_id,
+                    "status": output.status,
+                    "progress": output.progress,
+                }),
+                None,
+                None,
+                Some(output),
+            ),
+        }
+    }
 }

 #[derive(Clone, Debug, Deserialize)]
@@ -1889,6 +2004,12 @@ impl AdminService {
            .get_operation_version(workspace_id, operation_id, payload.version)
            .await?;
        let runtime = RuntimeOperation::from(record.snapshot.clone());
+        let mode = runtime
+            .execution_config
+            .streaming
+            .as_ref()
+            .map(|streaming| streaming.mode)
+            .unwrap_or(ExecutionMode::Unary);
        let request_preview =
            match build_request_preview(&record.snapshot.input_mapping, &payload.input) {
                Ok(preview) => preview,
@@ -1910,11 +2031,15 @@ impl AdminService {
                    .await?;
                    return Ok(TestRunResult {
                        ok: false,
+                        mode,
                        request_preview: Value::Null,
                        response_preview: Value::Null,
                        errors: vec![crate::error::runtime_test_failure(&RuntimeError::Mapping(
                            error,
                        ))],
+                        window: None,
+                        stream_session: None,
+                        async_job: None,
                    });
                }
            };
@@ -1924,16 +2049,39 @@ impl AdminService {
            .await;
        let started_at = std::time::Instant::now();
        match match resolved_auth {
-            Ok(resolved_auth) => {
-                self.runtime
+            Ok(resolved_auth) => match mode {
+                ExecutionMode::Unary => self
+                    .runtime
                    .execute_with_auth(&runtime, &payload.input, resolved_auth.as_ref())
                    .await
-            }
+                    .map(|output| TestRunOutcome::Unary { output }),
+                ExecutionMode::Window => self
+                    .runtime
+                    .execute_window_with_auth(&runtime, &payload.input, resolved_auth.as_ref())
+                    .await
+                    .map(|output| TestRunOutcome::Window { output }),
+                ExecutionMode::Session => self
+                    .start_stream_session_test(
+                        workspace_id,
+                        &record.snapshot,
+                        &runtime,
+                        &payload.input,
+                        resolved_auth.as_ref(),
+                    )
+                    .await
+                    .map(|output| TestRunOutcome::Session { output }),
+                ExecutionMode::AsyncJob => self
+                    .start_async_job_test(workspace_id, &record.snapshot, &runtime, &payload.input)
+                    .await
+                    .map(|output| TestRunOutcome::AsyncJob { output }),
+            },
            Err(error) => Err(error),
        } {
-            Ok(output) => {
+            Ok(outcome) => {
                let duration_ms =
                    u64::try_from(started_at.elapsed().as_millis()).unwrap_or(u64::MAX);
+                let (message, response_preview, window, stream_session, async_job) =
+                    outcome.into_result_views();
                self.record_invocation(InvocationRecordRequest {
                    workspace_id,
                    agent_id: None,
@@ -1941,19 +2089,23 @@ impl AdminService {
                    source: InvocationSource::AdminTestRun,
                    level: InvocationLevel::Info,
                    status: InvocationStatus::Ok,
-                    message: "admin test run completed".to_owned(),
+                    message: message.to_owned(),
                    status_code: None,
                    error_kind: None,
                    duration_ms,
                    request_preview: request_preview.clone(),
-                    response_preview: output.clone(),
+                    response_preview: response_preview.clone(),
                })
                .await?;
                Ok(TestRunResult {
                    ok: true,
+                    mode,
                    request_preview,
-                    response_preview: output,
+                    response_preview,
                    errors: Vec::new(),
+                    window,
+                    stream_session,
+                    async_job,
                })
            }
            Err(error) => {
@@ -1976,9 +2128,13 @@ impl AdminService {
                .await?;
                Ok(TestRunResult {
                    ok: false,
+                    mode,
                    request_preview,
                    response_preview: Value::Null,
                    errors: vec![crate::error::runtime_test_failure(&error)],
+                    window: None,
+                    stream_session: None,
+                    async_job: None,
                })
            }
        }
@@ -2009,6 +2165,187 @@ impl AdminService {
            .map(Some)
    }

+    async fn start_stream_session_test(
+        &self,
+        workspace_id: &WorkspaceId,
+        operation: &RegistryOperation,
+        runtime: &RuntimeOperation,
+        input: &Value,
+        resolved_auth: Option<&ResolvedAuth>,
+    ) -> Result<StreamSessionStartView, RuntimeError> {
+        let streaming = runtime.execution_config.streaming.as_ref().ok_or_else(|| {
+            RuntimeError::MissingStreamingConfig {
+                operation_id: runtime.operation_id.as_str().to_owned(),
+            }
+        })?;
+        let seed = self
+            .runtime
+            .execute_session_seed_with_auth(runtime, input, resolved_auth)
+            .await?;
+        let batch_size = streaming.max_items.unwrap_or(10).max(1) as usize;
+        let preview_count = seed.items.len().min(batch_size);
+        let preview_items = seed.items[..preview_count].to_vec();
+        let next_index = preview_count;
+        let created_at = now_string().map_err(|error| RuntimeError::SecretCrypto {
+            details: error.to_string(),
+        })?;
+        let expires_at = add_millis(
+            &created_at,
+            streaming.max_session_lifetime_ms.unwrap_or(60_000),
+        )
+        .map_err(|error| RuntimeError::SecretCrypto {
+            details: error.to_string(),
+        })?;
+        let session = StreamSession {
+            id: crank_core::StreamSessionId::new(new_prefixed_id("sess")),
+            workspace_id: workspace_id.clone(),
+            agent_id: None,
+            operation_id: operation.id.clone(),
+            protocol: operation.protocol,
+            mode: ExecutionMode::Session,
+            status: StreamStatus::Running,
+            cursor: (next_index < seed.items.len()).then(|| json!(next_index)),
+            state: json!(StoredSessionState {
+                input: input.clone(),
+                summary: seed.summary.clone(),
+                items: seed.items.clone(),
+                next_index,
+                batch_size,
+            }),
+            expires_at: expires_at.clone(),
+            last_poll_at: Some(created_at.clone()),
+            created_at,
+            closed_at: None,
+        };
+        self.registry
+            .create_stream_session(CreateStreamSessionRequest { session: &session })
+            .await
+            .map_err(|error| RuntimeError::SecretCrypto {
+                details: error.to_string(),
+            })?;
+
+        Ok(StreamSessionStartView {
+            session_id: session.id.as_str().to_owned(),
+            status: session.status,
+            expires_at,
+            poll_after_ms: streaming.poll_interval_ms.unwrap_or(1_000),
+            preview: json!({
+                "summary": seed.summary,
+                "items": preview_items,
+            }),
+        })
+    }
+
+    async fn start_async_job_test(
+        &self,
+        workspace_id: &WorkspaceId,
+        operation: &RegistryOperation,
+        runtime: &RuntimeOperation,
+        input: &Value,
+    ) -> Result<AsyncJobStartView, RuntimeError> {
+        let created_at = now_string().map_err(|error| RuntimeError::SecretCrypto {
+            details: error.to_string(),
+        })?;
+        let expires_at =
+            add_millis(&created_at, 300_000).map_err(|error| RuntimeError::SecretCrypto {
+                details: error.to_string(),
+            })?;
+        let job = AsyncJobHandle {
+            id: crank_core::AsyncJobId::new(new_prefixed_id("job")),
+            workspace_id: workspace_id.clone(),
+            agent_id: None,
+            operation_id: operation.id.clone(),
+            status: JobStatus::Running,
+            progress: json!({ "pct": 0 }),
+            result: None,
+            error: None,
+            expires_at: Some(expires_at),
+            created_at: created_at.clone(),
+            updated_at: created_at.clone(),
+            finished_at: None,
+        };
+        self.registry
+            .create_async_job(CreateAsyncJobRequest { job: &job })
+            .await
+            .map_err(|error| RuntimeError::SecretCrypto {
+                details: error.to_string(),
+            })?;
+
+        let registry = self.registry.clone();
+        let secret_crypto = self.secret_crypto.clone();
+        let runtime_for_task = self.runtime.clone();
+        let workspace_for_task = workspace_id.clone();
+        let operation_for_task = runtime.clone();
+        let input_for_task = input.clone();
+        let job_id = job.id.clone();
+        tokio::spawn(async move {
+            let resolved_auth = resolve_runtime_auth_for_task(
+                &registry,
+                &secret_crypto,
+                &workspace_for_task,
+                &operation_for_task.execution_config,
+            )
+            .await;
+            let result = match resolved_auth {
+                Ok(resolved_auth) => {
+                    runtime_for_task
+                        .execute_with_auth(
+                            &operation_for_task,
+                            &input_for_task,
+                            resolved_auth.as_ref(),
+                        )
+                        .await
+                }
+                Err(error) => Err(error),
+            };
+            let finished_at = match now_string() {
+                Ok(value) => value,
+                Err(_) => return,
+            };
+            let _ = match result {
+                Ok(output) => {
+                    registry
+                        .update_async_job_status(UpdateAsyncJobStatusRequest {
+                            job_id: &job_id,
+                            current_status: JobStatus::Running,
+                            next_status: JobStatus::Completed,
+                            progress: &json!({ "pct": 100 }),
+                            result: Some(&output),
+                            error: None,
+                            expires_at: None,
+                            updated_at: &finished_at,
+                            finished_at: Some(&finished_at),
+                        })
+                        .await
+                }
+                Err(error) => {
+                    registry
+                        .update_async_job_status(UpdateAsyncJobStatusRequest {
+                            job_id: &job_id,
+                            current_status: JobStatus::Running,
+                            next_status: JobStatus::Failed,
+                            progress: &json!({ "pct": 100 }),
+                            result: None,
+                            error: Some(&json!({
+                                "code": runtime_error_code(&error),
+                                "message": error.to_string(),
+                            })),
+                            expires_at: None,
+                            updated_at: &finished_at,
+                            finished_at: Some(&finished_at),
+                        })
+                        .await
+                }
+            };
+        });
+
+        Ok(AsyncJobStartView {
+            job_id: job.id.as_str().to_owned(),
+            status: job.status,
+            progress: job.progress,
+        })
+    }
+
    async fn resolve_auth_profile(
        &self,
        workspace_id: &WorkspaceId,
@@ -4267,6 +4604,74 @@ fn now_string() -> Result<String, ApiError> {
        .map_err(|error| ApiError::internal(error.to_string()))
 }

+fn add_millis(timestamp: &str, millis: u64) -> Result<String, ApiError> {
+    let parsed = OffsetDateTime::parse(timestamp, &Rfc3339)
+        .map_err(|error| ApiError::internal(error.to_string()))?;
+    parsed
+        .checked_add(time::Duration::milliseconds(millis as i64))
+        .ok_or_else(|| ApiError::internal("failed to add millis to timestamp"))?
+        .format(&Rfc3339)
+        .map_err(|error| ApiError::internal(error.to_string()))
+}
+
+async fn resolve_runtime_auth_for_task(
+    registry: &PostgresRegistry,
+    secret_crypto: &SecretCrypto,
+    workspace_id: &WorkspaceId,
+    execution_config: &crank_core::ExecutionConfig,
+) -> Result<Option<ResolvedAuth>, RuntimeError> {
+    let Some(auth_profile_id) = execution_config.auth_profile_ref.as_ref() else {
+        return Ok(None);
+    };
+
+    let auth_profile = registry
+        .get_auth_profile(workspace_id, auth_profile_id)
+        .await
+        .map_err(|error| RuntimeError::SecretCrypto {
+            details: error.to_string(),
+        })?
+        .ok_or_else(|| RuntimeError::MissingAuthProfile {
+            auth_profile_id: auth_profile_id.as_str().to_owned(),
+        })?;
+
+    let mut secrets = BTreeMap::new();
+    let used_at = now_string().map_err(|error| RuntimeError::SecretCrypto {
+        details: error.to_string(),
+    })?;
+
+    for secret_id in auth_profile.config.secret_ids() {
+        let secret = registry
+            .get_secret(workspace_id, secret_id)
+            .await
+            .map_err(|error| RuntimeError::SecretCrypto {
+                details: error.to_string(),
+            })?
+            .ok_or_else(|| RuntimeError::MissingSecret {
+                secret_id: secret_id.as_str().to_owned(),
+            })?;
+        let version = registry
+            .get_current_secret_version(workspace_id, secret_id)
+            .await
+            .map_err(|error| RuntimeError::SecretCrypto {
+                details: error.to_string(),
+            })?
+            .ok_or_else(|| RuntimeError::MissingSecretVersion {
+                secret_id: secret_id.as_str().to_owned(),
+                version: secret.secret.current_version,
+            })?;
+        let plaintext = secret_crypto.decrypt(&version.secret_version.ciphertext)?;
+        registry
+            .touch_secret(workspace_id, secret_id, &used_at)
+            .await
+            .map_err(|error| RuntimeError::SecretCrypto {
+                details: error.to_string(),
+            })?;
+        secrets.insert(secret_id.clone(), plaintext);
+    }
+
+    ResolvedAuth::from_profile(&auth_profile, &secrets).map(Some)
+}
+
 fn default_invitation_expiry() -> Result<String, ApiError> {
    OffsetDateTime::now_utc()
        .checked_add(time::Duration::days(7))
@@ -831,10 +831,18 @@ var TRANSLATIONS = {
    'wizard.test.failed': 'Test run returned errors',
    'wizard.test.completed_body': 'The current draft executed successfully against the upstream.',
    'wizard.test.failed_body': 'The draft executed, but the backend returned validation or runtime errors.',
+    'wizard.test.window_completed': 'Window test completed',
+    'wizard.test.window_completed_body': 'Window complete: {window_complete}. Truncated: {truncated}. Has more: {has_more}.',
+    'wizard.test.session_started': 'Session test started',
+    'wizard.test.session_started_body': 'Session {session_id} was created. Poll after {poll_after_ms} ms from the stream sessions view.',
+    'wizard.test.async_job_started': 'Async job test started',
+    'wizard.test.async_job_started_body': 'Async job {job_id} was created. Track status and result from the async jobs view.',
    'wizard.test.no_response': 'No test response yet',
    'wizard.test.no_response_body': 'Run a test first, then copy the response preview into the output sample editor.',
    'wizard.test.response_copied': 'Response copied',
    'wizard.test.response_copied_body': 'The latest test response was copied into the output sample editor.',
+    'wizard.boolean.yes': 'yes',
+    'wizard.boolean.no': 'no',
    'wizard.yaml.exported': 'YAML exported',
    'wizard.yaml.exported_body': 'The current draft YAML was downloaded from the backend.',
    'wizard.yaml.none': 'No YAML provided',
@@ -1813,10 +1821,18 @@ var TRANSLATIONS = {
    'wizard.test.failed': 'Тестовый прогон вернул ошибки',
    'wizard.test.completed_body': 'Текущий черновик успешно выполнился против upstream.',
    'wizard.test.failed_body': 'Черновик выполнился, но backend вернул ошибки валидации или runtime.',
+    'wizard.test.window_completed': 'Оконный тест завершен',
+    'wizard.test.window_completed_body': 'Окно завершено: {window_complete}. Усечено: {truncated}. Есть продолжение: {has_more}.',
+    'wizard.test.session_started': 'Session-тест запущен',
+    'wizard.test.session_started_body': 'Создана session {session_id}. Следующий poll можно делать через {poll_after_ms} мс из представления stream sessions.',
+    'wizard.test.async_job_started': 'Async job тест запущен',
+    'wizard.test.async_job_started_body': 'Создан async job {job_id}. Отслеживайте статус и результат через представление async jobs.',
    'wizard.test.no_response': 'Тестового ответа пока нет',
    'wizard.test.no_response_body': 'Сначала выполните тест, затем скопируйте response preview в editor output sample.',
    'wizard.test.response_copied': 'Ответ скопирован',
    'wizard.test.response_copied_body': 'Последний тестовый ответ скопирован в editor output sample.',
+    'wizard.boolean.yes': 'да',
+    'wizard.boolean.no': 'нет',
    'wizard.yaml.exported': 'YAML экспортирован',
    'wizard.yaml.exported_body': 'YAML текущего черновика скачан из backend.',
    'wizard.yaml.none': 'YAML не предоставлен',
@@ -2546,6 +2546,47 @@ function setTextareaValue(id, value) {
  element.value = safeStringify(value);
 }

+function describeWizardTestResult(result) {
+  if (result.stream_session && result.stream_session.session_id) {
+    return {
+      title: tKey('wizard.test.session_started'),
+      body: tfKey('wizard.test.session_started_body', {
+        session_id: result.stream_session.session_id,
+        poll_after_ms: result.stream_session.poll_after_ms
+      }),
+      isError: false,
+    };
+  }
+
+  if (result.async_job && result.async_job.job_id) {
+    return {
+      title: tKey('wizard.test.async_job_started'),
+      body: tfKey('wizard.test.async_job_started_body', {
+        job_id: result.async_job.job_id
+      }),
+      isError: false,
+    };
+  }
+
+  if (result.window) {
+    return {
+      title: result.ok ? tKey('wizard.test.window_completed') : tKey('wizard.test.failed'),
+      body: tfKey('wizard.test.window_completed_body', {
+        truncated: result.window.truncated ? tKey('wizard.boolean.yes') : tKey('wizard.boolean.no'),
+        has_more: result.window.has_more ? tKey('wizard.boolean.yes') : tKey('wizard.boolean.no'),
+        window_complete: result.window.window_complete ? tKey('wizard.boolean.yes') : tKey('wizard.boolean.no')
+      }),
+      isError: !result.ok,
+    };
+  }
+
+  return {
+    title: result.ok ? tKey('wizard.test.completed') : tKey('wizard.test.failed'),
+    body: result.ok ? tKey('wizard.test.completed_body') : tKey('wizard.test.failed_body'),
+    isError: !result.ok,
+  };
+}
+
 async function uploadInputSampleFromWizard() {
  await persistCurrentDraft(true);
  await window.CrankApi.uploadInputSample(
@@ -2586,12 +2627,11 @@ async function runWizardTest() {
  setTextareaValue('wizard-test-request-preview', result.request_preview);
  setTextareaValue('wizard-test-response-preview', result.response_preview);
  setTextareaValue('wizard-test-errors', result.errors && result.errors.length ? result.errors : []);
+  var status = describeWizardTestResult(result);
  showWizardLiveStatus(
-    result.ok ? tKey('wizard.test.completed') : tKey('wizard.test.failed'),
-    result.ok
-      ? tKey('wizard.test.completed_body')
-      : tKey('wizard.test.failed_body'),
-    !result.ok
+    status.title,
+    status.body,
+    status.isError
  );
 }

@@ -3872,7 +3872,9 @@ fn map_secret_version_record(row: &PgRow) -> Result<SecretVersionRecord, Registr
            ciphertext: row.try_get("ciphertext")?,
            key_version: row.try_get("key_version")?,
            created_at: row.try_get("created_at")?,
-            created_by: row.try_get::<Option<String>, _>("created_by")?.map(UserId::new),
+            created_by: row
+                .try_get::<Option<String>, _>("created_by")?
+                .map(UserId::new),
        },
    })
 }
@@ -127,6 +127,11 @@
 - `POST /descriptors/wsdl` принимает raw WSDL payload и сохраняет inspection metadata для SOAP;
 - `POST /descriptors/xsd` принимает supporting XSD artifacts для SOAP operation;
 - `GET /soap/services` возвращает нормализованный список `service -> port -> operation` из последнего WSDL текущего draft version.
+- `POST /operations/{operation_id}/test-runs` остается единым test-run endpoint для `unary`, `window`, `session` и `async_job` execution modes;
+- test-run response всегда возвращает `mode`, `request_preview`, `response_preview`, `errors`, а для streaming modes дополняется:
+  - `window` с `window_complete`, `truncated`, `has_more`, `cursor`;
+  - `stream_session` с `session_id`, `status`, `expires_at`, `poll_after_ms`, `preview`;
+  - `async_job` с `job_id`, `status`, `progress`.

 ### 5.5. Upstream auth profiles

@@ -264,11 +264,11 @@

 ## 8. Streaming operation test-runs

-### `POST /api/admin/workspaces/{workspace_id}/operations/{operation_id}/stream-test-runs`
+### `POST /api/admin/workspaces/{workspace_id}/operations/{operation_id}/test-runs`

 Назначение:

- выполнить `window`, `session` или `async_job` test-run для draft version;
+- выполнить `unary`, `window`, `session` или `async_job` test-run для draft version;
 - показать оператору runtime behavior до publish.

 Тело:
@@ -280,30 +280,57 @@
    "service": "billing",
    "level": "error"
  },
-  "test_mode": "window",
-  "overrides": {
-    "streaming": {
-      "window_duration_ms": 3000,
-      "max_items": 50
-    }
+  "version": 4,
+  "input": {
+    "service": "billing",
+    "level": "error"
  }
 }
 ```

+Успех для `unary`:
+
+```json
+{
+  "ok": true,
+  "mode": "unary",
+  "request_preview": {
+    "service": "billing",
+    "level": "error"
+  },
+  "response_preview": {
+    "summary": "completed"
+  },
+  "errors": [],
+  "window": null,
+  "stream_session": null,
+  "async_job": null
+}
+```
+
 Успех для `window`:

 ```json
 {
-  "run_id": "strun_01j0test",
+  "ok": true,
  "mode": "window",
-  "status": "completed",
-  "window_complete": true,
-  "truncated": false,
-  "has_more": false,
-  "summary": {},
-  "items": [],
-  "cursor": null,
-  "duration_ms": 2871
+  "request_preview": {
+    "service": "billing",
+    "level": "error"
+  },
+  "response_preview": {
+    "summary": {},
+    "items": []
+  },
+  "errors": [],
+  "window": {
+    "window_complete": true,
+    "truncated": false,
+    "has_more": false,
+    "cursor": null
+  },
+  "stream_session": null,
+  "async_job": null
 }
 ```

@@ -311,18 +338,29 @@

 ```json
 {
-  "run_id": "strun_01j0test",
+  "ok": true,
  "mode": "session",
-  "status": "running",
-  "session": {
-    "session_id": "sess_01j0stream",
-    "expires_at": "2026-04-06T12:05:00Z",
-    "poll_after_ms": 2000
+  "request_preview": {
+    "service": "billing",
+    "level": "error"
  },
-  "preview": {
+  "response_preview": {
    "summary": {},
    "items": []
-  }
+  },
+  "errors": [],
+  "window": null,
+  "stream_session": {
+    "session_id": "sess_01j0stream",
+    "status": "running",
+    "expires_at": "2026-04-06T12:05:00Z",
+    "poll_after_ms": 2000,
+    "preview": {
+      "summary": {},
+      "items": []
+    }
+  },
+  "async_job": null
 }
 ```

@@ -330,10 +368,19 @@

 ```json
 {
-  "run_id": "strun_01j0test",
+  "ok": true,
  "mode": "async_job",
-  "status": "running",
-  "job": {
+  "request_preview": {
+    "service": "billing",
+    "level": "error"
+  },
+  "response_preview": {
+    "job": "started"
+  },
+  "errors": [],
+  "window": null,
+  "stream_session": null,
+  "async_job": {
    "job_id": "job_01j0deploy",
    "status": "running",
    "progress": {
@@ -343,47 +390,34 @@
 }
 ```

-### `POST /api/admin/workspaces/{workspace_id}/operations/{operation_id}/stream-test-runs/{run_id}/poll`
-
-Назначение:
-
- продолжить session-oriented test-run.
-
-Тело:
+Ошибка:

 ```json
 {
-  "session_id": "sess_01j0stream"
+  "ok": false,
+  "mode": "window",
+  "request_preview": {
+    "service": "billing",
+    "level": "error"
+  },
+  "response_preview": null,
+  "errors": [
+    {
+      "code": "runtime_test_failure",
+      "message": "upstream timeout"
+    }
+  ],
+  "window": null,
+  "stream_session": null,
+  "async_job": null
 }
 ```

-Ответ:
+Примечания:

-```json
-{
-  "run_id": "strun_01j0test",
-  "mode": "session",
-  "status": "running",
-  "window_complete": false,
-  "truncated": false,
-  "has_more": true,
-  "summary": {},
-  "items": [],
-  "cursor": "next_cursor"
-}
-```
-
-### `POST /api/admin/workspaces/{workspace_id}/operations/{operation_id}/stream-test-runs/{run_id}/stop`
-
-Назначение:
-
- остановить test session.
-
-### `GET /api/admin/workspaces/{workspace_id}/operations/{operation_id}/stream-test-runs/{run_id}/result`
-
-Назначение:
-
- получить финальный результат async-job test-run.
+- endpoint не использует отдельные `stream-test-runs/*` subresources;
+- session и async-job test-runs создают обычные runtime resources в `stream_sessions` и `async_jobs`;
+- дальнейшее наблюдение идет через `GET /stream-sessions`, `GET /stream-sessions/{session_id}`, `POST /stream-sessions/{session_id}/stop`, `GET /async-jobs`, `GET /async-jobs/{job_id}`, `GET /async-jobs/{job_id}/result`, `POST /async-jobs/{job_id}/cancel`.

 ## 9. Runtime session resources

@@ -27,6 +27,14 @@
 - `Stream Sessions`
 - `Async Jobs`

+Для test-run UX:
+
+- wizard status block должен различать `unary`, `window`, `session`, `async_job`;
+- `window` показывает bounded flags `window_complete`, `truncated`, `has_more`;
+- `session` показывает `session_id`, `poll_after_ms` и переводит оператора к странице `Stream Sessions`;
+- `async_job` показывает `job_id` и переводит оператора к странице `Async Jobs`;
+- request/response preview textareas остаются общими для всех режимов.
+
 ## 3. Wizard information architecture

 ## 3.1. Shared wizard structure