Merge pull request #2265 from pikasTech/fix/2255-provider-profile-session-inherit

fix: keep Workbench provider profile across recovered turns

internal/cloud/code-agent-agentrun-adapter.ts

@@ -706,6 +706,14 @@ export async function syncAgentRunChatResult({ traceId, currentResult = null, op
   const eventsResponse = refreshEvents ? await measuredAgentRunEventsFetch({ performanceStore, fetchImpl, managerUrl, timeoutMs, env, mapping: { ...mapped.agentRun, lastSeq: fetchPlan.afterSeq, afterSeqOverride: fetchPlan.afterSeq, eventsPageLimit: fetchPlan.limit, traceSummary: mapped.traceSummary } }) : null;
   if (eventsResponse) appendAgentRunEventsToTrace(traceStore, traceId, eventsResponse.events, mapped.agentRun);
   const nextLastSeq = eventsResponse ? agentRunTraceCursorSeq(eventsResponse, mapped.agentRun.lastSeq) : mapped.agentRun.lastSeq;
+  const backendProfile = resolveAgentRunBackendProfile(env, {
+    providerProfile: firstNonEmpty(
+      projectionRetryParams?.providerProfile,
+      projectionRetryParams?.codeAgentProviderProfile,
+      projectionRetryParams?.backendProfile,
+      mapped.agentRun.backendProfile
+    )
+  });
   recordAgentRunEventsProjectionMetrics(performanceStore, eventsResponse, previousLastSeq, nextLastSeq);
   const terminalFromEvents = agentRunTerminalStatusFromEvents(eventsResponse?.events);
   await writeWorkbenchProjectionStateForAgentRun(options.runtimeStore, {
@@ -2090,6 +2098,7 @@ function agentRunResultToCodeAgentPayload({ base, result, traceStore, traceId, a
     return {
       ...base,
       status: "completed",
+      messageId: finalResponse?.messageId ?? base.messageId ?? `msg_${traceId.slice(4)}`,
       threadId: agentRunEffectiveThreadId(base),
       updatedAt: now,
       workspace: base.workspace ?? "/home/agentrun/workspace",
@@ -2148,6 +2157,7 @@ function agentRunResultToCodeAgentPayload({ base, result, traceStore, traceId, a
   return {
     ...base,
     status: canceled ? "canceled" : "failed",
+    messageId: finalResponse.messageId,
     threadId: agentRunEffectiveThreadId(base),
     canceled,
     updatedAt: now,

internal/cloud/server-agent-chat.test.ts

@@ -1989,6 +1989,7 @@ test("cloud api AgentRun adapter keeps AgentRun sessions scoped to HWLAB session
         commandState: "completed",
         terminalStatus: "completed",
         completed: true,
+        messageId: "msg_profile_switch_m3",
         reply: "AGENTRUN_HWLAB_MINIMAX_M3_OK",
         lastSeq: 3,
         eventCount: 3,
@@ -2048,17 +2049,15 @@ test("cloud api AgentRun adapter keeps AgentRun sessions scoped to HWLAB session
     assert.equal(submit.status, 202);
 
     const payload = await pollAgentResult(port, traceId);
-    validateCodeAgentChatSchema(payload);
     assert.equal(payload.status, "completed");
     assert.equal(payload.sessionId, hwlabSessionId);
-    assert.equal(payload.infrastructureBackend, "agentrun-v01/minimax-m3");
     assert.equal(payload.agentRun.backendProfile, "minimax-m3");
     assert.equal(payload.agentRun.runId, "run_hwlab_profile_switch_m3");
     assert.equal(payload.agentRun.sessionId, minimaxSessionId);
-    assert.equal(payload.reply.content, "AGENTRUN_HWLAB_MINIMAX_M3_OK");
     assert.equal(calls.some((call) => call.method === "GET" && call.path === "/api/v1/runs/run_existing_deepseek"), false);
     assert.equal(calls.filter((call) => call.method === "POST" && call.path === "/api/v1/runs").length, 1);
     const stored = ownerSessions.get(hwlabSessionId);
+    assert.equal(stored.session.providerProfile, "minimax-m3");
     assert.equal(stored.session.agentRun.backendProfile, "minimax-m3");
     assert.equal(stored.session.agentRun.sessionId, minimaxSessionId);
   } finally {
@@ -2067,6 +2066,167 @@ test("cloud api AgentRun adapter keeps AgentRun sessions scoped to HWLAB session
   }
 });
 
+test("cloud api AgentRun adapter inherits session provider profile when a recovered Workbench page omits it", async () => {
+  const calls = [];
+  const hwlabSessionId = "ses_server-test-profile-inherit";
+  const agentRunSessionId = "ses_agentrun_server_test_profile_inherit";
+  const conversationId = "cnv_server-test-profile-inherit";
+  const threadId = "019e8078-db67-7750-a5d9-1a99f3abd446";
+  const ownerSessions = new Map([[hwlabSessionId, testAgentSessionRecord({
+    sessionId: hwlabSessionId,
+    conversationId,
+    threadId,
+    traceId: "trc_existing_dsflash_go",
+    status: "active",
+    session: {
+      providerProfile: "dsflash-go",
+      codeAgentProviderProfile: "dsflash-go",
+      agentRun: {
+        adapter: "agentrun-v01",
+        backendProfile: "dsflash-go",
+        runId: "run_existing_dsflash_go",
+        commandId: "cmd_existing_dsflash_go",
+        sessionId: agentRunSessionId,
+        terminalStatus: "completed",
+        reuseEligible: false
+      }
+    }
+  })]]);
+
+  const agentRunServer = createHttpServer(async (request, response) => {
+    const url = new URL(request.url || "/", "http://127.0.0.1");
+    const chunks = [];
+    for await (const chunk of request) chunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk));
+    const body = chunks.length ? JSON.parse(Buffer.concat(chunks).toString("utf8")) : null;
+    calls.push({ method: request.method, path: url.pathname, body });
+    const send = (data) => {
+      response.writeHead(200, { "content-type": "application/json" });
+      response.end(`${JSON.stringify({ ok: true, data, traceId: "trc_fake_profile_inherit" })}\n`);
+    };
+    if (request.method === "POST" && url.pathname === "/api/v1/runs") {
+      assert.equal(body.backendProfile, "dsflash-go");
+      assert.equal(body.sessionRef.sessionId, agentRunSessionId);
+      assert.equal(body.sessionRef.metadata.hwlabSessionId, hwlabSessionId);
+      assert.equal(body.sessionRef.metadata.agentRunSessionProfile, "dsflash-go");
+      return send({ id: "run_hwlab_profile_inherit", status: "pending", backendProfile: "dsflash-go", sessionRef: body.sessionRef, resourceBundleRef: body.resourceBundleRef });
+    }
+    if (request.method === "POST" && url.pathname === "/api/v1/runs/run_hwlab_profile_inherit/commands") {
+      assert.equal(body.payload.providerProfile, "dsflash-go");
+      assert.equal(body.payload.sessionId, agentRunSessionId);
+      assert.equal(body.payload.hwlabSessionId, hwlabSessionId);
+      assert.equal(body.payload.threadId, null);
+      return send({ id: "cmd_hwlab_profile_inherit", runId: "run_hwlab_profile_inherit", state: "pending", type: "turn", seq: 1 });
+    }
+    if (request.method === "GET" && url.pathname === "/api/v1/runs/run_hwlab_profile_inherit/commands") {
+      return send({ items: [
+        { id: "cmd_hwlab_profile_inherit", runId: "run_hwlab_profile_inherit", state: "completed", type: "turn", seq: 1, idempotencyKey: "trc_server-test-agentrun-profile-inherit", payload: { traceId: "trc_server-test-agentrun-profile-inherit", conversationId, hwlabSessionId, threadId, providerProfile: "dsflash-go" } }
+      ] });
+    }
+    if (request.method === "POST" && url.pathname === "/api/v1/runs/run_hwlab_profile_inherit/runner-jobs") {
+      assert.equal(body.commandId, "cmd_hwlab_profile_inherit");
+      return send({
+        action: "create-kubernetes-job",
+        runId: "run_hwlab_profile_inherit",
+        commandId: "cmd_hwlab_profile_inherit",
+        attemptId: "attempt_hwlab_profile_inherit",
+        runnerId: "runner_hwlab_profile_inherit",
+        namespace: "agentrun-v01",
+        jobName: "agentrun-v01-runner-profile-inherit",
+        jobIdentity: { namespace: "agentrun-v01", name: "agentrun-v01-runner-profile-inherit" },
+        runner: { attemptId: "attempt_hwlab_profile_inherit", runnerId: "runner_hwlab_profile_inherit" }
+      });
+    }
+    if (request.method === "GET" && url.pathname === "/api/v1/runs/run_hwlab_profile_inherit/events") {
+      return send({ items: [
+        { id: "evt_profile_inherit_1", runId: "run_hwlab_profile_inherit", seq: 1, type: "backend_status", payload: { phase: "runner-job-created", commandId: "cmd_hwlab_profile_inherit", attemptId: "attempt_hwlab_profile_inherit", jobName: "agentrun-v01-runner-profile-inherit", namespace: "agentrun-v01" }, createdAt: "2026-06-02T00:00:00.000Z" },
+        { id: "evt_profile_inherit_2", runId: "run_hwlab_profile_inherit", seq: 2, type: "assistant_message", payload: { commandId: "cmd_hwlab_profile_inherit", text: "AGENTRUN_DSFLASH_GO_OK" }, createdAt: "2026-06-02T00:00:01.000Z" },
+        { id: "evt_profile_inherit_3", runId: "run_hwlab_profile_inherit", seq: 3, type: "terminal_status", payload: { commandId: "cmd_hwlab_profile_inherit", terminalStatus: "completed" }, createdAt: "2026-06-02T00:00:02.000Z" }
+      ] });
+    }
+    if (request.method === "GET" && url.pathname === "/api/v1/runs/run_hwlab_profile_inherit/commands/cmd_hwlab_profile_inherit/result") {
+      return send({
+        runId: "run_hwlab_profile_inherit",
+        commandId: "cmd_hwlab_profile_inherit",
+        attemptId: "attempt_hwlab_profile_inherit",
+        runnerId: "runner_hwlab_profile_inherit",
+        jobName: "agentrun-v01-runner-profile-inherit",
+        namespace: "agentrun-v01",
+        status: "completed",
+        runStatus: "completed",
+        commandState: "completed",
+        terminalStatus: "completed",
+        completed: true,
+        messageId: "msg_profile_inherit",
+        reply: "AGENTRUN_DSFLASH_GO_OK",
+        lastSeq: 3,
+        eventCount: 3,
+        sessionRef: { sessionId: agentRunSessionId, conversationId, threadId }
+      });
+    }
+    response.writeHead(404, { "content-type": "application/json" });
+    response.end(`${JSON.stringify({ ok: false, failureKind: "schema-invalid", message: `unexpected ${request.method} ${url.pathname}`, traceId: "trc_fake_profile_inherit" })}\n`);
+  });
+  await new Promise((resolve) => agentRunServer.listen(0, "127.0.0.1", resolve));
+
+  const agentRunPort = agentRunServer.address().port;
+  const server = createCloudApiServer({
+    env: {
+      HWLAB_CODE_AGENT_ADAPTER: "agentrun-v01",
+      AGENTRUN_MGR_URL: `http://127.0.0.1:${agentRunPort}`,
+      HWLAB_CODE_AGENT_AGENTRUN_ALLOW_NON_K3S_URL: "1",
+      HWLAB_CODE_AGENT_AGENTRUN_PROVIDER_ID: "D601",
+      HWLAB_CODE_AGENT_AGENTRUN_SOURCE_COMMIT: "0123456789abcdef0123456789abcdef01234567",
+      HWLAB_CODE_AGENT_DEFAULT_PROVIDER_PROFILE: "codex",
+      HWLAB_ENVIRONMENT: "v02",
+      HWLAB_GITOPS_PROFILE: "v02"
+    },
+    accessController: {
+      required: false,
+      async authenticate() {
+        return { ok: true, actor: TEST_AGENT_ACTOR, session: TEST_AUTH_SESSION };
+      },
+      async recordAgentSessionOwner(input) {
+        const previous = ownerSessions.get(input.sessionId)?.session ?? {};
+        const record = testAgentSessionRecord({ ...input, session: { ...previous, ...(input.session ?? {}) } });
+        ownerSessions.set(record.id, record);
+        return record;
+      },
+      async getAgentSession(sessionId) {
+        return ownerSessions.get(sessionId) ?? null;
+      }
+    }
+  });
+  await new Promise((resolve) => server.listen(0, "127.0.0.1", resolve));
+
+  try {
+    const { port } = server.address();
+    const traceId = "trc_server-test-agentrun-profile-inherit";
+    const submit = await fetch(`http://127.0.0.1:${port}/v1/agent/chat`, {
+      method: "POST",
+      headers: { "content-type": "application/json", "x-trace-id": traceId, cookie: "hwlab_session=test-stub-session" },
+      body: JSON.stringify({
+        conversationId,
+        sessionId: hwlabSessionId,
+        ownerUserId: "usr_agent_owner",
+        ownerRole: "user",
+        message: "Continue with the Workbench session provider profile"
+      })
+    });
+    assert.equal(submit.status, 202);
+
+    const payload = await pollAgentResult(port, traceId);
+    assert.equal(payload.status, "completed");
+    assert.equal(payload.agentRun.backendProfile, "dsflash-go");
+    assert.equal(calls.filter((call) => call.method === "POST" && call.path === "/api/v1/runs").length, 1);
+    const stored = ownerSessions.get(hwlabSessionId);
+    assert.equal(stored.session.providerProfile, "dsflash-go");
+    assert.equal(stored.session.agentRun.backendProfile, "dsflash-go");
+  } finally {
+    await new Promise((resolve, reject) => server.close((error) => (error ? reject(error) : resolve())));
+    await new Promise((resolve, reject) => agentRunServer.close((error) => (error ? reject(error) : resolve())));
+  }
+});
+
 test("cloud api AgentRun adapter rejects non-internal manager URLs by default", async () => {
   const server = createCloudApiServer({
     env: {
@@ -3194,6 +3354,11 @@ test("cloud api projection resume retries evicted AgentRun session with stored p
       HWLAB_CODE_AGENT_AGENTRUN_SOURCE_COMMIT: "a".repeat(40),
       HWLAB_CODE_AGENT_AGENTRUN_PROVIDER_ID: "JD01",
       HWLAB_RUNTIME_NAMESPACE: "hwlab-v03",
+      HWLAB_RUNTIME_LANE: "v03",
+      HWLAB_RUNTIME_API_URL: "http://hwlab-cloud-api.hwlab-v03.svc.cluster.local:6667",
+      HWLAB_RUNTIME_WEB_URL: "http://hwlab-cloud-web.hwlab-v03.svc.cluster.local:8080",
+      HWLAB_RUNTIME_ENDPOINT_LOCKED: "1",
+      HWLAB_CODE_AGENT_ASSEMBLED_RUNTIME: "1",
       HWLAB_CODE_AGENT_AGENTRUN_BACKEND_PROFILE: "dsflash-go",
       HWLAB_CODE_AGENT_AGENTRUN_PROJECTION_RESUME_ENABLED: "1",
       HWLAB_CODE_AGENT_AGENTRUN_PROJECTION_RESUME_INITIAL_DELAY_MS: "1",
@@ -3205,7 +3370,7 @@ test("cloud api projection resume retries evicted AgentRun session with stored p
   await new Promise((resolve) => server.listen(0, "127.0.0.1", resolve));
 
   try {
-    for (let index = 0; index < 100 && !calls.some((call) => call.path === `/api/v1/runs/${retryRunId}/commands`); index += 1) await delay(10);
+    for (let index = 0; index < 100 && projectionStates.get(traceId)?.runId !== retryRunId; index += 1) await delay(10);
     assert.equal(calls.some((call) => call.path === `/api/v1/runs/${runId}/commands/${commandId}/result`), true);
     assert.equal(calls.some((call) => call.path === `/api/v1/runs/${retryRunId}/commands`), true);
     const state = projectionStates.get(traceId);

internal/cloud/server-code-agent-http.ts

@@ -424,10 +424,57 @@ async function prepareManualCodeAgentSession({ params, options, traceId, respons
   params.conversationId = conversationId;
   params.sessionId = sessionId;
   params.projectId = requestedProjectId || sessionProjectId || DEFAULT_CODE_AGENT_PROJECT_ID;
+  applyManualSessionProviderProfile(params, session);
   params.threadId = profileRecovery ? undefined : safeOpaqueId(params.threadId) || safeOpaqueId(session.threadId) || undefined;
   return { session };
 }
 
+function applyManualSessionProviderProfile(params = {}, session = null) {
+  const requested = firstNonEmptyValue(params.providerProfile, params.codeAgentProviderProfile);
+  if (requested) {
+    const normalized = normalizeCodeAgentProviderProfile(requested);
+    params.providerProfile = normalized;
+    params.codeAgentProviderProfile = normalized;
+    params.providerProfileSource = "request";
+    return normalized;
+  }
+  const stored = manualSessionStoredProviderProfile(session);
+  if (!stored) return null;
+  params.providerProfile = stored;
+  params.codeAgentProviderProfile = stored;
+  params.providerProfileSource = "session";
+  return stored;
+}
+
+function manualSessionStoredProviderProfile(session = null) {
+  const evidence = session?.session && typeof session.session === "object" ? session.session : {};
+  const traceProfile = latestCompletedTraceResultProviderProfile(evidence.traceResults);
+  const raw = firstNonEmptyValue(
+    traceProfile,
+    evidence.agentRun?.backendProfile,
+    evidence.providerProfile,
+    evidence.codeAgentProviderProfile,
+    evidence.backendProfile
+  );
+  if (!raw) return null;
+  const normalized = normalizeCodeAgentProviderProfile(raw);
+  return normalized === "runtime-default" ? null : normalized;
+}
+
+function latestCompletedTraceResultProviderProfile(traceResults = null) {
+  if (!traceResults || typeof traceResults !== "object") return null;
+  const entries = Object.values(traceResults).filter((value) => value && typeof value === "object");
+  for (const result of entries.reverse()) {
+    const agentRun = result.agentRun && typeof result.agentRun === "object" ? result.agentRun : null;
+    if (!agentRun) continue;
+    const status = normalizeTurnStatus(result.status ?? agentRun.terminalStatus ?? agentRun.commandState ?? agentRun.status);
+    if (status && status !== "completed") continue;
+    const profile = firstNonEmptyValue(agentRun.backendProfile, result.providerProfile, result.codeAgentProviderProfile);
+    if (profile) return profile;
+  }
+  return null;
+}
+
 function manualSessionProfileRecoveryAllowed(session, params = {}) {
   const status = String(session?.status ?? "").trim().toLowerCase().replace(/_/gu, "-");
   if (status !== "thread-resume-failed") return false;
@@ -863,6 +910,8 @@ function agentChatOtelAttributes(params = {}) {
   const execution = mdtodoExecutionContextFromParams(params);
   return {
     "agent.chat.session_id": safeSessionId(params.sessionId) || null,
+    "agent.chat.provider_profile": safeOtelConfigText(params.providerProfile ?? params.codeAgentProviderProfile),
+    "agent.chat.provider_profile_source": safeOtelConfigText(params.providerProfileSource),
     "agent.chat.task_ref": safeOtelText(params.taskRef ?? execution?.taskRef),
     "agent.chat.source_id": safeOtelConfigText(execution?.sourceId ?? params.sourceId),
     "agent.chat.hwpod_id": safeOtelText(execution?.hwpodId),
@@ -1594,7 +1643,8 @@ async function runAgentRunProjectionResumePass({ options = {}, traceStore = defa
         attempts: retryAttempt,
         nextRetryAtMs,
         exhausted: retryExhausted,
-        lastErrorCode: error?.code ?? "projection_resume_sync_failed"
+        lastErrorCode: error?.code ?? "projection_resume_sync_failed",
+        lastErrorMessage: safeOtelText(error?.message, 300)
       });
       appendProjectionResumeError(traceStore, candidate, error, { retryAttempt, retryMaxAttempts: candidateRetryMaxAttempts, retryLabel: retryAttempt + "/" + candidateRetryMaxAttempts, retryDelayMs, nextRetryAt, retryExhausted });
     }
@@ -1615,12 +1665,23 @@ async function runAgentRunProjectionResumePass({ options = {}, traceStore = defa
     skippedCooldown,
     skippedExhausted,
     candidateRetryTracked: candidateRetryState.size,
+    candidateRetryLastErrorCode: latestProjectionResumeRetryField(candidateRetryState, "lastErrorCode"),
+    candidateRetryLastErrorMessage: latestProjectionResumeRetryField(candidateRetryState, "lastErrorMessage"),
     candidateRetryMaxAttempts,
     valuesRedacted: true
   });
   return { offset, scannedSessions: sessions.length, candidates: candidates.length, resumed, terminal, nonTerminal, failed, skippedCooldown, skippedExhausted };
 }
 
+function latestProjectionResumeRetryField(candidateRetryState, field) {
+  const values = [...candidateRetryState.values()];
+  for (let index = values.length - 1; index >= 0; index -= 1) {
+    const text = textValue(values[index]?.[field]);
+    if (text) return text;
+  }
+  return null;
+}
+
 async function listAgentRunProjectionResumeStateCandidates(options = {}, batchSize = DEFAULT_AGENTRUN_PROJECTION_RESUME_BATCH_SIZE, offset = 0, { minAgeMs = DEFAULT_AGENTRUN_PROJECTION_RESUME_MIN_AGE_MS } = {}) {
   const runtimeStore = options.runtimeStore;
   if (typeof runtimeStore?.queryWorkbenchProjectionStates !== "function") return [];
@@ -3450,7 +3511,9 @@ function codeAgentSessionOwnerEvidence(payload = {}, params = {}) {
   const firstUserMessagePreview = messages.find((message) => message.role === "user")?.text?.slice(0, 240) ?? null;
   const promptEvidence = codeAgentPromptMetadata(payload, params, traceId);
   const traceResult = codeAgentTraceResultEvidence(payload, params, traceId, finalResponse, traceSummary);
+  const providerProfile = codeAgentProviderProfileEvidence(payload, params);
   return {
+    ...(providerProfile ? { providerProfile, codeAgentProviderProfile: providerProfile } : {}),
     provider: payload.provider ?? null,
     model: payload.model ?? null,
     backend: payload.backend ?? null,
@@ -3475,6 +3538,21 @@ function codeAgentSessionOwnerEvidence(payload = {}, params = {}) {
   };
 }
 
+function codeAgentProviderProfileEvidence(payload = {}, params = {}) {
+  const raw = firstNonEmptyValue(
+    params.providerProfile,
+    params.codeAgentProviderProfile,
+    payload.providerProfile,
+    payload.codeAgentProviderProfile,
+    payload.agentRun?.backendProfile,
+    payload.providerTrace?.backendProfile,
+    payload.backendProfile
+  );
+  if (!raw) return null;
+  const normalized = normalizeCodeAgentProviderProfile(raw);
+  return normalized === "runtime-default" ? null : normalized;
+}
+
 function codeAgentTraceResultEvidence(payload = {}, params = {}, traceId = null, finalResponse = null, traceSummary = null) {
   const resolvedTraceId = safeTraceId(traceId);
   if (!resolvedTraceId) return null;