fix(agents): keep PI telemetry on model provider

Keep PI embedded-run usage metadata and traces attributed to the resolved model provider instead of the PI harness label.

CHANGELOG.md

@@ -21,6 +21,7 @@ Docs: https://docs.openclaw.ai
 - Plugins/runtime-deps: memoize packaged bundled runtime dist-mirror preparation after the first successful pass while keeping source-checkout mirrors refreshable, so constrained Docker/VPS installs avoid repeated root scans before chat turns. Refs #73428, #73421, #73532, and #73477. Thanks @Dimaoggg, @oromeis, @oadiazp, @jmfraga, @bstanbury, @antoniusfelix, and @jkobject.
 - Channels/Discord: treat bare numeric outbound targets that match the effective Discord DM allowlist as user DMs while preserving account-specific legacy `dm.allowFrom` precedence over inherited root `allowFrom`. (#74303) Thanks @Squirbie.
 - Control UI: make the chat sidebar split divider focusable, keyboard-resizable, ARIA-described, and pointer-event based so sidebar resizing works without a mouse. Thanks @BunsDev.
+- Agents/usage: keep PI embedded-run telemetry attributed to the resolved model provider instead of the PI harness label, so OpenRouter and other provider-backed turns report the right provider in session usage and traces. Thanks @vincentkoc.
 - Agents/auth: keep OAuth auth profiles inherited from the main agent read-through instead of copying refresh tokens into secondary agents, and refresh Codex app-server tokens against the owning store so multi-agent swarms avoid reused refresh-token failures. Fixes #74055. Thanks @ClarityInvest.
 - Channels/Telegram: honor `ALL_PROXY` / `all_proxy` and service-level `OPENCLAW_PROXY_URL` when constructing the HTTP/1-only Telegram Bot API transport, so Windows and service installs that rely on those proxy settings no longer fall back to direct egress. Fixes #74014; refs #74086. Thanks @SymbolStar.
 - Channels/Telegram: continue polling when `deleteWebhook` hits a transient network failure but `getWebhookInfo` confirms no webhook is configured, so startup does not retry cleanup forever after the webhook was already removed. Refs #74086; carries forward #47384. Thanks @clovericbot.

src/agents/pi-embedded-runner/run.ts

@@ -104,6 +104,7 @@ import {
   resolveFinalAssistantRawText,
   resolveFinalAssistantVisibleText,
   resolveMaxRunRetryIterations,
+  resolveReportedModelRef,
   resolveOverloadFailoverBackoffMs,
   resolveOverloadProfileRotationLimit,
   resolveRateLimitProfileRotationLimit,
@@ -1878,11 +1879,16 @@ export async function runEmbeddedPiAgent(
             lastRunPromptUsage,
             lastTurnTotal,
           });
+          const reportedModelRef = resolveReportedModelRef({
+            provider,
+            model: model.id,
+            assistant: sessionLastAssistant,
+          });
           const agentMeta: EmbeddedPiAgentMeta = {
             sessionId: sessionIdUsed,
             sessionFile: sessionFileUsed,
-            provider: sessionLastAssistant?.provider ?? provider,
-            model: sessionLastAssistant?.model ?? model.id,
+            provider: reportedModelRef.provider,
+            model: reportedModelRef.model,
             contextTokens: ctxInfo.tokens,
             agentHarnessId: attempt.agentHarnessId,
             usage: usageMeta.usage,
@@ -2403,8 +2409,8 @@ export async function runEmbeddedPiAgent(
                   ]
                 : undefined,
               executionTrace: {
-                winnerProvider: sessionLastAssistant?.provider ?? provider,
-                winnerModel: sessionLastAssistant?.model ?? model.id,
+                winnerProvider: reportedModelRef.provider,
+                winnerModel: reportedModelRef.model,
                 attempts:
                   traceAttempts.length > 0 ||
                   sessionLastAssistant?.provider ||
@@ -2412,8 +2418,8 @@ export async function runEmbeddedPiAgent(
                     ? [
                         ...traceAttempts,
                         {
-                          provider: sessionLastAssistant?.provider ?? provider,
-                          model: sessionLastAssistant?.model ?? model.id,
+                          provider: reportedModelRef.provider,
+                          model: reportedModelRef.model,
                           result: "success",
                           stage: "assistant",
                         },

src/agents/pi-embedded-runner/run/helpers.resolve-error-context.test.ts

@@ -22,4 +22,17 @@ describe("resolveActiveErrorContext", () => {
 
     expect(result).toEqual({ provider: "openai", model: "gpt-5.4-codex" });
   });
+
+  it("ignores the embedded PI harness provider when the model provider is known", () => {
+    const result = resolveActiveErrorContext({
+      provider: "openrouter",
+      model: "openai/gpt-5.4",
+      assistant: {
+        provider: "pi",
+        model: "pi",
+      },
+    });
+
+    expect(result).toEqual({ provider: "openrouter", model: "openai/gpt-5.4" });
+  });
 });

src/agents/pi-embedded-runner/run/helpers.ts

@@ -85,11 +85,38 @@ export function resolveActiveErrorContext(params: {
 }): {
   provider: string;
   model: string;
+} {
+  return resolveReportedModelRef(params);
+}
+
+function isEmbeddedHarnessProvider(provider: string): boolean {
+  return provider.trim().toLowerCase() === "pi";
+}
+
+export function resolveReportedModelRef(params: {
+  provider: string;
+  model: string;
+  assistant?: { provider?: string; model?: string } | null;
+}): {
+  provider: string;
+  model: string;
 } {
   const assistantProvider = params.assistant?.provider?.trim();
   const assistantModel = params.assistant?.model?.trim();
+  if (!assistantProvider) {
+    return {
+      provider: params.provider,
+      model: assistantModel || params.model,
+    };
+  }
+  if (isEmbeddedHarnessProvider(assistantProvider)) {
+    return {
+      provider: params.provider,
+      model: params.model,
+    };
+  }
   return {
-    provider: assistantProvider || params.provider,
+    provider: assistantProvider,
     model: assistantModel || params.model,
   };
 }

src/agents/pi-embedded-runner/usage-reporting.test.ts

@@ -1,9 +1,10 @@
 import type { AssistantMessage } from "@mariozechner/pi-ai";
-import { beforeAll, beforeEach, describe, expect, it } from "vitest";
+import { beforeAll, beforeEach, describe, expect, it, vi } from "vitest";
 import { makeAttemptResult } from "./run.overflow-compaction.fixture.js";
 import {
   loadRunOverflowCompactionHarness,
   mockedEnsureRuntimePluginsLoaded,
+  mockedResolveModelAsync,
   mockedRunEmbeddedAttempt,
 } from "./run.overflow-compaction.harness.js";
 import type { EmbeddedRunAttemptResult } from "./run/types.js";
@@ -191,4 +192,48 @@ describe("runEmbeddedPiAgent usage reporting", () => {
     // If the bug exists, it will likely be 350
     expect(usage?.total).toBe(200);
   });
+
+  it("reports the resolved model provider when PI marks the assistant message as pi", async () => {
+    mockedResolveModelAsync.mockResolvedValueOnce({
+      model: {
+        id: "openai/gpt-5.4",
+        provider: "openrouter",
+        contextWindow: 200000,
+        api: "openai-completions",
+      },
+      error: null,
+      authStorage: {
+        setRuntimeApiKey: vi.fn(),
+      },
+      modelRegistry: {},
+    });
+    mockedRunEmbeddedAttempt.mockResolvedValueOnce(
+      makeAttemptResult({
+        assistantTexts: ["Response 1"],
+        lastAssistant: makeAssistantMessage({
+          provider: "pi",
+          model: "pi",
+          usage: { input: 100, output: 50, total: 150 } as unknown as AssistantMessage["usage"],
+        }),
+        attemptUsage: { input: 100, output: 50, total: 150 },
+      }),
+    );
+
+    const result = await runEmbeddedPiAgent({
+      sessionId: "test-session",
+      sessionKey: "test-key",
+      sessionFile: "/tmp/session.json",
+      workspaceDir: "/tmp/workspace",
+      prompt: "hello",
+      provider: "openrouter",
+      model: "openai/gpt-5.4",
+      timeoutMs: 30000,
+      runId: "run-provider-attribution",
+    });
+
+    expect(result.meta.agentMeta?.provider).toBe("openrouter");
+    expect(result.meta.agentMeta?.model).toBe("openai/gpt-5.4");
+    expect(result.meta.executionTrace?.winnerProvider).toBe("openrouter");
+    expect(result.meta.executionTrace?.winnerModel).toBe("openai/gpt-5.4");
+  });
 });