From d037244aed3869858acab7eebd48daccc880622a Mon Sep 17 00:00:00 2001
From: hugocasa <hugo@casademont.ch>
Date: Fri, 7 Aug 2026 17:56:34 +0200
Subject: [PATCH] fix: drop web search guidance on the completions api fallback

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 ai_evals/cases/global.yaml                    |  8 ++--
 .../components/copilot/chat/chatLoop.test.ts  | 38 +++++++++++++++++++
 .../lib/components/copilot/chat/chatLoop.ts   |  8 +++-
 3 files changed, 48 insertions(+), 6 deletions(-)

diff --git a/ai_evals/cases/global.yaml b/ai_evals/cases/global.yaml
index 547abf5b60..2369ae23b5 100644
--- a/ai_evals/cases/global.yaml
+++ b/ai_evals/cases/global.yaml
@@ -2011,11 +2011,9 @@
     - authentication uses Holded's own key header rather than a bearer token
     - the result stays an AI draft and is not deployed
 
-# The fixture hub carries two Baremetrics scripts (List Sources, Create Customer) and
-# none for annotations, so no hub script does what is asked. The integration's
-# conventions — a `Baremetrics` resource type whose single field is `apiKey`, bearer
-# auth, the api.baremetrics.com/v1 base — are only knowable by reading one of those
-# other scripts, which is what makes this a test of using a near-miss as an example.
+# The fixture hub carries two Baremetrics scripts, neither for annotations, so nothing
+# does what is asked. Its conventions — an `apiKey` resource, bearer auth, the
+# api.baremetrics.com/v1 base — are only knowable by reading one of those near-misses.
 - id: global-hub3-integration-as-example
   prompt: |-
     Write me a script that adds an annotation to my Baremetrics account.
diff --git a/frontend/src/lib/components/copilot/chat/chatLoop.test.ts b/frontend/src/lib/components/copilot/chat/chatLoop.test.ts
index 005b12f4c5..83f1856d88 100644
--- a/frontend/src/lib/components/copilot/chat/chatLoop.test.ts
+++ b/frontend/src/lib/components/copilot/chat/chatLoop.test.ts
@@ -376,6 +376,44 @@ describe('runChatLoop onBeforeIteration web search sync', () => {
 		})
 	})
 
+	// The Completions API has no web-search tool at all, so the fallback is a third
+	// path (beside the two retries) that must not ship the guidance.
+	it('drops the web search guidance on the Completions API fallback', async () => {
+		let systemMessage: ChatLoopConfig['systemMessage'] = {
+			role: 'system',
+			content: 'search the web'
+		}
+
+		mocks.getOpenAIResponsesCompletion.mockRejectedValue(
+			new Error('Responses API is not enabled for this organization')
+		)
+		mocks.getCompletion.mockResolvedValue({})
+		const onWebSearchUnavailable = vi.fn()
+
+		await runChatLoop(
+			createConfig({
+				workspace: `workspace-${randomUUID()}`,
+				getSystemMessage: () => systemMessage,
+				onWebSearchUnavailable,
+				onBeforeIteration: async (_t, _h, _m, webSearch: boolean) => {
+					systemMessage = {
+						role: 'system',
+						content: webSearch ? 'search the web' : 'no web search'
+					}
+				}
+			})
+		)
+
+		// Falling back is per request; the model itself can still serve web search.
+		expect(onWebSearchUnavailable).not.toHaveBeenCalled()
+
+		expect(mocks.getCompletion).toHaveBeenCalledTimes(1)
+		expect(mocks.getCompletion.mock.calls[0][0][0]).toEqual({
+			role: 'system',
+			content: 'no web search'
+		})
+	})
+
 	it('sends the system message the callback rewrote on this iteration, not the next', async () => {
 		let systemMessage: ChatLoopConfig['systemMessage'] = { role: 'system', content: 'stale' }
 		const config = createConfig({
diff --git a/frontend/src/lib/components/copilot/chat/chatLoop.ts b/frontend/src/lib/components/copilot/chat/chatLoop.ts
index a6cb97733e..e3e407d6dc 100644
--- a/frontend/src/lib/components/copilot/chat/chatLoop.ts
+++ b/frontend/src/lib/components/copilot/chat/chatLoop.ts
@@ -494,12 +494,18 @@ export async function runChatLoop(config: ChatLoopConfig): Promise<ChatLoopResul
 			}
 
 			if (useCompletionsApi) {
+				let completionsParams = messageParams
 				if (webSearch) {
 					console.warn(
 						'Web search is only supported via the OpenAI Responses API; ignoring it for the Completions API fallback.'
 					)
+					// Re-announce the effective value for the request actually being sent —
+					// not onWebSearchUnavailable, which means the model cannot serve web
+					// search at all and is cached as such; this fallback is per request.
+					await onBeforeIteration?.(tools, helpers, modelProvider, false)
+					completionsParams = messageParamsFor(config.systemMessage)
 				}
-				const completion = await getCompletion(messageParams, abortController, toolDefs, {
+				const completion = await getCompletion(completionsParams, abortController, toolDefs, {
 					forceCompletions: true,
 					forceModelProvider: modelProvider,
 					openaiClient: clients.openai,