fix(look-at): use synchronous prompt to fix race condition (#1620 regression)

PR #1620 migrated all prompt calls from session.prompt (blocking) to
session.promptAsync (fire-and-forget HTTP 204). This broke look_at which
needs the multimodal-looker response to be available immediately after
the prompt call returns.

Fix: add promptSyncWithModelSuggestionRetry() that uses session.prompt
(blocking) with model suggestion retry support. look_at now uses this
sync variant while all other callers keep using promptAsync.

- Add promptSyncWithModelSuggestionRetry to model-suggestion-retry.ts
- Switch look_at from promptWithModelSuggestionRetry to sync variant
- Add comprehensive tests for the new sync function
- No changes to other callers (delegate-task, background-agent)

src/shared/model-suggestion-retry.test.ts

@@ -1,5 +1,5 @@
 import { describe, it, expect, mock } from "bun:test"
-import { parseModelSuggestion, promptWithModelSuggestionRetry } from "./model-suggestion-retry"
+import { parseModelSuggestion, promptWithModelSuggestionRetry, promptSyncWithModelSuggestionRetry } from "./model-suggestion-retry"
 
 describe("parseModelSuggestion", () => {
   describe("structured NamedError format", () => {
@@ -377,3 +377,128 @@ describe("promptWithModelSuggestionRetry", () => {
     expect(promptMock).toHaveBeenCalledTimes(1)
   })
 })
+
+describe("promptSyncWithModelSuggestionRetry", () => {
+  it("should use synchronous prompt (not promptAsync)", async () => {
+    // given a client with both prompt and promptAsync
+    const promptMock = mock(() => Promise.resolve())
+    const promptAsyncMock = mock(() => Promise.resolve())
+    const client = { session: { prompt: promptMock, promptAsync: promptAsyncMock } }
+
+    // when calling promptSyncWithModelSuggestionRetry
+    await promptSyncWithModelSuggestionRetry(client as any, {
+      path: { id: "session-1" },
+      body: {
+        parts: [{ type: "text", text: "hello" }],
+        model: { providerID: "anthropic", modelID: "claude-sonnet-4" },
+      },
+    })
+
+    // then should call prompt (sync), NOT promptAsync
+    expect(promptMock).toHaveBeenCalledTimes(1)
+    expect(promptAsyncMock).toHaveBeenCalledTimes(0)
+  })
+
+  it("should retry with suggested model on ProviderModelNotFoundError", async () => {
+    // given a client that fails first with model-not-found, then succeeds
+    const promptMock = mock()
+      .mockRejectedValueOnce({
+        name: "ProviderModelNotFoundError",
+        data: {
+          providerID: "anthropic",
+          modelID: "claude-sonet-4",
+          suggestions: ["claude-sonnet-4"],
+        },
+      })
+      .mockResolvedValueOnce(undefined)
+    const client = { session: { prompt: promptMock } }
+
+    // when calling promptSyncWithModelSuggestionRetry
+    await promptSyncWithModelSuggestionRetry(client as any, {
+      path: { id: "session-1" },
+      body: {
+        parts: [{ type: "text", text: "hello" }],
+        model: { providerID: "anthropic", modelID: "claude-sonet-4" },
+      },
+    })
+
+    // then should call prompt twice (original + retry with suggestion)
+    expect(promptMock).toHaveBeenCalledTimes(2)
+    const retryCall = promptMock.mock.calls[1][0]
+    expect(retryCall.body.model).toEqual({
+      providerID: "anthropic",
+      modelID: "claude-sonnet-4",
+    })
+  })
+
+  it("should throw original error when no suggestion available", async () => {
+    // given a client that fails with a non-model error
+    const originalError = new Error("Connection refused")
+    const promptMock = mock().mockRejectedValueOnce(originalError)
+    const client = { session: { prompt: promptMock } }
+
+    // when calling promptSyncWithModelSuggestionRetry
+    // then should throw the original error
+    await expect(
+      promptSyncWithModelSuggestionRetry(client as any, {
+        path: { id: "session-1" },
+        body: {
+          parts: [{ type: "text", text: "hello" }],
+          model: { providerID: "anthropic", modelID: "claude-sonnet-4" },
+        },
+      })
+    ).rejects.toThrow("Connection refused")
+
+    expect(promptMock).toHaveBeenCalledTimes(1)
+  })
+
+  it("should throw when model-not-found but no model in original request", async () => {
+    // given a client that fails with model error but no model in body
+    const promptMock = mock().mockRejectedValueOnce({
+      name: "ProviderModelNotFoundError",
+      data: {
+        providerID: "anthropic",
+        modelID: "claude-sonet-4",
+        suggestions: ["claude-sonnet-4"],
+      },
+    })
+    const client = { session: { prompt: promptMock } }
+
+    // when calling without model in body
+    // then should throw (cannot retry without original model)
+    await expect(
+      promptSyncWithModelSuggestionRetry(client as any, {
+        path: { id: "session-1" },
+        body: {
+          parts: [{ type: "text", text: "hello" }],
+        },
+      })
+    ).rejects.toThrow()
+
+    expect(promptMock).toHaveBeenCalledTimes(1)
+  })
+
+  it("should pass all body fields through to prompt", async () => {
+    // given a client where prompt succeeds
+    const promptMock = mock().mockResolvedValueOnce(undefined)
+    const client = { session: { prompt: promptMock } }
+
+    // when calling with additional body fields
+    await promptSyncWithModelSuggestionRetry(client as any, {
+      path: { id: "session-1" },
+      body: {
+        agent: "multimodal-looker",
+        tools: { task: false },
+        parts: [{ type: "text", text: "analyze" }],
+        model: { providerID: "google", modelID: "gemini-3-flash" },
+        variant: "max",
+      },
+    })
+
+    // then call should pass all fields through unchanged
+    const call = promptMock.mock.calls[0][0]
+    expect(call.body.agent).toBe("multimodal-looker")
+    expect(call.body.tools).toEqual({ task: false })
+    expect(call.body.variant).toBe("max")
+  })
+})

src/shared/model-suggestion-retry.ts

@@ -88,3 +88,42 @@ export async function promptWithModelSuggestionRetry(
   // model errors happen asynchronously server-side and cannot be caught here
   await client.session.promptAsync(args as Parameters<typeof client.session.promptAsync>[0])
 }
+
+/**
+ * Synchronous variant of promptWithModelSuggestionRetry.
+ *
+ * Uses `session.prompt` (blocking HTTP call that waits for the LLM response)
+ * instead of `promptAsync` (fire-and-forget HTTP 204).
+ *
+ * Required by callers that need the response to be available immediately after
+ * the call returns — e.g. look_at, which reads session messages right away.
+ */
+export async function promptSyncWithModelSuggestionRetry(
+  client: Client,
+  args: PromptArgs,
+): Promise<void> {
+  try {
+    await client.session.prompt(args as Parameters<typeof client.session.prompt>[0])
+  } catch (error) {
+    const suggestion = parseModelSuggestion(error)
+    if (!suggestion || !args.body.model) {
+      throw error
+    }
+
+    log("[model-suggestion-retry] Model not found, retrying with suggestion", {
+      original: `${suggestion.providerID}/${suggestion.modelID}`,
+      suggested: suggestion.suggestion,
+    })
+
+    await client.session.prompt({
+      ...args,
+      body: {
+        ...args.body,
+        model: {
+          providerID: suggestion.providerID,
+          modelID: suggestion.suggestion,
+        },
+      },
+    } as Parameters<typeof client.session.prompt>[0])
+  }
+}

src/tools/look-at/tools.ts

@@ -3,7 +3,7 @@ import { pathToFileURL } from "node:url"
 import { tool, type PluginInput, type ToolDefinition } from "@opencode-ai/plugin"
 import { LOOK_AT_DESCRIPTION, MULTIMODAL_LOOKER_AGENT } from "./constants"
 import type { LookAtArgs } from "./types"
-import { log, promptWithModelSuggestionRetry } from "../../shared"
+import { log, promptSyncWithModelSuggestionRetry } from "../../shared"
 
 interface LookAtArgsWithAlias extends LookAtArgs {
   path?: string
@@ -223,7 +223,7 @@ Original error: ${createResult.error}`
 
       log(`[look_at] Sending prompt with ${isBase64Input ? "base64 image" : "file"} to session ${sessionID}`)
       try {
-        await promptWithModelSuggestionRetry(ctx.client, {
+        await promptSyncWithModelSuggestionRetry(ctx.client, {
           path: { id: sessionID },
           body: {
             agent: MULTIMODAL_LOOKER_AGENT,