fix(ai,coding-agent): expose provider timeout/retry controls closes #3627

packages/ai/CHANGELOG.md

@@ -10,6 +10,7 @@
 
 - Fixed DeepSeek V4 session replay 400 errors by adding `thinkingFormat: "deepseek"` (sends `thinking: { type }` + `reasoning_effort`), a `reasoningEffortMap`, and `requiresReasoningContentOnAssistantMessages` compat that injects empty `reasoning_content` on all replayed assistant messages when reasoning is enabled ([#3636](https://github.com/badlogic/pi-mono/issues/3636))
 - Fixed GPT-5.5 generated context window metadata to use the observed 272k limit.
+- Fixed provider request controls to expose `timeoutMs` and `maxRetries` in stream options and forward them through OpenAI/Azure/Anthropic request options, preventing unconfigurable SDK timeout/retry defaults on long-running local inference requests ([#3627](https://github.com/badlogic/pi-mono/issues/3627))
 
 ## [0.70.0] - 2026-04-23
 

packages/ai/src/providers/anthropic.ts

@@ -457,7 +457,14 @@ export const streamAnthropic: StreamFunction<"anthropic-messages", AnthropicOpti
 				params = nextParams as MessageCreateParamsStreaming;
 			}
 			const response = await client.messages
-				.create({ ...params, stream: true }, { signal: options?.signal })
+				.create(
+					{ ...params, stream: true },
+					{
+						signal: options?.signal,
+						timeout: options?.timeoutMs,
+						maxRetries: options?.maxRetries,
+					},
+				)
 				.asResponse();
 			await options?.onResponse?.({ status: response.status, headers: headersToRecord(response.headers) }, model);
 			stream.push({ type: "start", partial: output });

packages/ai/src/providers/azure-openai-responses.ts

@@ -92,7 +92,11 @@ export const streamAzureOpenAIResponses: StreamFunction<"azure-openai-responses"
 				params = nextParams as ResponseCreateParamsStreaming;
 			}
 			const { data: openaiStream, response } = await client.responses
-				.create(params, options?.signal ? { signal: options.signal } : undefined)
+				.create(params, {
+					signal: options?.signal,
+					timeout: options?.timeoutMs,
+					maxRetries: options?.maxRetries,
+				})
 				.withResponse();
 			await options?.onResponse?.({ status: response.status, headers: headersToRecord(response.headers) }, model);
 			stream.push({ type: "start", partial: output });

packages/ai/src/providers/openai-completions.ts

@@ -145,7 +145,11 @@ export const streamOpenAICompletions: StreamFunction<"openai-completions", OpenA
 				params = nextParams as OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming;
 			}
 			const { data: openaiStream, response } = await client.chat.completions
-				.create(params, { signal: options?.signal })
+				.create(params, {
+					signal: options?.signal,
+					timeout: options?.timeoutMs,
+					maxRetries: options?.maxRetries,
+				})
 				.withResponse();
 			await options?.onResponse?.({ status: response.status, headers: headersToRecord(response.headers) }, model);
 			stream.push({ type: "start", partial: output });

packages/ai/src/providers/openai-responses.ts

@@ -99,7 +99,11 @@ export const streamOpenAIResponses: StreamFunction<"openai-responses", OpenAIRes
 				params = nextParams as ResponseCreateParamsStreaming;
 			}
 			const { data: openaiStream, response } = await client.responses
-				.create(params, options?.signal ? { signal: options.signal } : undefined)
+				.create(params, {
+					signal: options?.signal,
+					timeout: options?.timeoutMs,
+					maxRetries: options?.maxRetries,
+				})
 				.withResponse();
 			await options?.onResponse?.({ status: response.status, headers: headersToRecord(response.headers) }, model);
 			stream.push({ type: "start", partial: output });

packages/ai/src/providers/simple-options.ts

@@ -11,6 +11,8 @@ export function buildBaseOptions(model: Model<Api>, options?: SimpleStreamOption
 		headers: options?.headers,
 		onPayload: options?.onPayload,
 		onResponse: options?.onResponse,
+		timeoutMs: options?.timeoutMs,
+		maxRetries: options?.maxRetries,
 		maxRetryDelayMs: options?.maxRetryDelayMs,
 		metadata: options?.metadata,
 	};

packages/ai/src/types.ts

@@ -101,6 +101,16 @@ export interface StreamOptions {
 	 * Not supported by all providers (e.g., AWS Bedrock uses SDK auth).
 	 */
 	headers?: Record<string, string>;
+	/**
+	 * HTTP request timeout in milliseconds for providers/SDKs that support it.
+	 * For example, OpenAI and Anthropic SDK clients default to 10 minutes.
+	 */
+	timeoutMs?: number;
+	/**
+	 * Maximum retry attempts for providers/SDKs that support client-side retries.
+	 * For example, OpenAI and Anthropic SDK clients default to 2.
+	 */
+	maxRetries?: number;
 	/**
 	 * Maximum delay in milliseconds to wait for a retry when the server requests a long wait.
 	 * If the server's requested delay exceeds this value, the request fails immediately

packages/coding-agent/CHANGELOG.md

@@ -10,6 +10,7 @@
 
 - Fixed `/copy` to avoid unbounded OSC 52 writes and clipboard races that could break terminal rendering or panic the native clipboard addon ([#3639](https://github.com/badlogic/pi-mono/issues/3639))
 - Fixed extension flag docs to show `pi.getFlag()` using registered flag names without the CLI `--` prefix ([#3614](https://github.com/badlogic/pi-mono/issues/3614))
+- Fixed provider retry/timeout settings wiring by adding `retry.provider.{timeoutMs,maxRetries,maxRetryDelayMs}`, migrating legacy `retry.maxDelayMs`, and forwarding provider controls into `streamSimple` request options ([#3627](https://github.com/badlogic/pi-mono/issues/3627))
 
 ## [0.70.0] - 2026-04-23
 

packages/coding-agent/docs/settings.md

@@ -77,12 +77,14 @@ Edit directly or use `/settings` for common options.
 
 | Setting | Type | Default | Description |
 |---------|------|---------|-------------|
-| `retry.enabled` | boolean | `true` | Enable automatic retry on transient errors |
-| `retry.maxRetries` | number | `3` | Maximum retry attempts |
-| `retry.baseDelayMs` | number | `2000` | Base delay for exponential backoff (2s, 4s, 8s) |
-| `retry.maxDelayMs` | number | `60000` | Max server-requested delay before failing (60s) |
+| `retry.enabled` | boolean | `true` | Enable automatic agent-level retry on transient errors |
+| `retry.maxRetries` | number | `3` | Maximum agent-level retry attempts |
+| `retry.baseDelayMs` | number | `2000` | Base delay for agent-level exponential backoff (2s, 4s, 8s) |
+| `retry.provider.timeoutMs` | number | SDK default | Provider/SDK request timeout in milliseconds |
+| `retry.provider.maxRetries` | number | SDK default | Provider/SDK retry attempts |
+| `retry.provider.maxRetryDelayMs` | number | `60000` | Max server-requested delay before failing (60s) |
 
-When a provider requests a retry delay longer than `maxDelayMs` (e.g., Google's "quota will reset after 5h"), the request fails immediately with an informative error instead of waiting silently. Set to `0` to disable the cap.
+When a provider requests a retry delay longer than `retry.provider.maxRetryDelayMs` (e.g., Google's "quota will reset after 5h"), the request fails immediately with an informative error instead of waiting silently. Set to `0` to disable the cap.
 
 ```json
 {
@@ -90,7 +92,11 @@ When a provider requests a retry delay longer than `maxDelayMs` (e.g., Google's
     "enabled": true,
     "maxRetries": 3,
     "baseDelayMs": 2000,
-    "maxDelayMs": 60000
+    "provider": {
+      "timeoutMs": 3600000,
+      "maxRetries": 0,
+      "maxRetryDelayMs": 60000
+    }
   }
 }
 ```

packages/coding-agent/src/core/sdk.ts

@@ -315,10 +315,14 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			if (!auth.ok) {
 				throw new Error(auth.error);
 			}
+			const providerRetrySettings = settingsManager.getProviderRetrySettings();
 			const openRouterAttributionHeaders = getOpenRouterAttributionHeaders(model, settingsManager);
 			return streamSimple(model, context, {
 				...options,
 				apiKey: auth.apiKey,
+				timeoutMs: options?.timeoutMs ?? providerRetrySettings.timeoutMs,
+				maxRetries: options?.maxRetries ?? providerRetrySettings.maxRetries,
+				maxRetryDelayMs: options?.maxRetryDelayMs ?? providerRetrySettings.maxRetryDelayMs,
 				headers:
 					openRouterAttributionHeaders || auth.headers || options?.headers
 						? { ...openRouterAttributionHeaders, ...auth.headers, ...options?.headers }
@@ -353,7 +357,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 		followUpMode: settingsManager.getFollowUpMode(),
 		transport: settingsManager.getTransport(),
 		thinkingBudgets: settingsManager.getThinkingBudgets(),
-		maxRetryDelayMs: settingsManager.getRetrySettings().maxDelayMs,
+		maxRetryDelayMs: settingsManager.getProviderRetrySettings().maxRetryDelayMs,
 	});
 
 	// Restore messages if session has existing data

packages/coding-agent/src/core/settings-manager.ts

@@ -16,11 +16,17 @@ export interface BranchSummarySettings {
 	skipPrompt?: boolean; // default: false - when true, skips "Summarize branch?" prompt and defaults to no summary
 }
 
+export interface ProviderRetrySettings {
+	timeoutMs?: number; // SDK/provider request timeout in milliseconds
+	maxRetries?: number; // SDK/provider retry attempts
+	maxRetryDelayMs?: number; // default: 60000 (max server-requested delay before failing)
+}
+
 export interface RetrySettings {
 	enabled?: boolean; // default: true
 	maxRetries?: number; // default: 3
 	baseDelayMs?: number; // default: 2000 (exponential backoff: 2s, 4s, 8s)
-	maxDelayMs?: number; // default: 60000 (max server-requested delay before failing)
+	provider?: ProviderRetrySettings;
 }
 
 export interface TerminalSettings {
@@ -354,6 +360,30 @@ export class SettingsManager {
 			}
 		}
 
+		// Migrate retry.maxDelayMs -> retry.provider.maxRetryDelayMs
+		if (
+			"retry" in settings &&
+			typeof settings.retry === "object" &&
+			settings.retry !== null &&
+			!Array.isArray(settings.retry)
+		) {
+			const retrySettings = settings.retry as Record<string, unknown>;
+			const providerSettings =
+				typeof retrySettings.provider === "object" && retrySettings.provider !== null
+					? (retrySettings.provider as Record<string, unknown>)
+					: undefined;
+			if (
+				typeof retrySettings.maxDelayMs === "number" &&
+				(providerSettings?.maxRetryDelayMs === undefined || providerSettings?.maxRetryDelayMs === null)
+			) {
+				retrySettings.provider = {
+					...(providerSettings ?? {}),
+					maxRetryDelayMs: retrySettings.maxDelayMs,
+				};
+			}
+			delete retrySettings.maxDelayMs;
+		}
+
 		return settings as Settings;
 	}
 
@@ -683,12 +713,19 @@ export class SettingsManager {
 		this.save();
 	}
 
-	getRetrySettings(): { enabled: boolean; maxRetries: number; baseDelayMs: number; maxDelayMs: number } {
+	getRetrySettings(): { enabled: boolean; maxRetries: number; baseDelayMs: number } {
 		return {
 			enabled: this.getRetryEnabled(),
 			maxRetries: this.settings.retry?.maxRetries ?? 3,
 			baseDelayMs: this.settings.retry?.baseDelayMs ?? 2000,
-			maxDelayMs: this.settings.retry?.maxDelayMs ?? 60000,
+		};
+	}
+
+	getProviderRetrySettings(): { timeoutMs?: number; maxRetries?: number; maxRetryDelayMs: number } {
+		return {
+			timeoutMs: this.settings.retry?.provider?.timeoutMs,
+			maxRetries: this.settings.retry?.provider?.maxRetries,
+			maxRetryDelayMs: this.settings.retry?.provider?.maxRetryDelayMs ?? 60000,
 		};
 	}