Merge PR #596: Add supportsUsageInStreaming compat flag

packages/ai/CHANGELOG.md

@@ -5,6 +5,7 @@
 ### Added
 
 - Added `GOOGLE_APPLICATION_CREDENTIALS` env var support for Vertex AI credential detection (standard for CI/production).
+- Added `supportsUsageInStreaming` compatibility flag for OpenAI-compatible providers that reject `stream_options: { include_usage: true }`. Defaults to `true`. Set to `false` in model config for providers like gatewayz.ai. ([#596](https://github.com/badlogic/pi-mono/pull/596) by [@XesGaDeus](https://github.com/XesGaDeus))
 
 ### Fixed
 

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

@@ -370,9 +370,12 @@ function buildParams(model: Model<"openai-completions">, context: Context, optio
 		model: model.id,
 		messages,
 		stream: true,
-		stream_options: { include_usage: true },
 	};
 
+	if (compat.supportsUsageInStreaming !== false) {
+		(params as any).stream_options = { include_usage: true };
+	}
+
 	if (compat.supportsStore) {
 		params.store = false;
 	}
@@ -657,6 +660,7 @@ function detectCompatFromUrl(baseUrl: string): Required<OpenAICompat> {
 		supportsStore: !isNonStandard,
 		supportsDeveloperRole: !isNonStandard,
 		supportsReasoningEffort: !isGrok,
+		supportsUsageInStreaming: true,
 		maxTokensField: useMaxTokens ? "max_tokens" : "max_completion_tokens",
 		requiresToolResultName: isMistral,
 		requiresAssistantAfterToolResult: false, // Mistral no longer requires this as of Dec 2024
@@ -677,6 +681,7 @@ function getCompat(model: Model<"openai-completions">): Required<OpenAICompat> {
 		supportsStore: model.compat.supportsStore ?? detected.supportsStore,
 		supportsDeveloperRole: model.compat.supportsDeveloperRole ?? detected.supportsDeveloperRole,
 		supportsReasoningEffort: model.compat.supportsReasoningEffort ?? detected.supportsReasoningEffort,
+		supportsUsageInStreaming: model.compat.supportsUsageInStreaming ?? detected.supportsUsageInStreaming,
 		maxTokensField: model.compat.maxTokensField ?? detected.maxTokensField,
 		requiresToolResultName: model.compat.requiresToolResultName ?? detected.requiresToolResultName,
 		requiresAssistantAfterToolResult:

packages/ai/src/types.ts

@@ -207,6 +207,8 @@ export interface OpenAICompat {
 	supportsDeveloperRole?: boolean;
 	/** Whether the provider supports `reasoning_effort`. Default: auto-detected from URL. */
 	supportsReasoningEffort?: boolean;
+	/** Whether the provider supports `stream_options: { include_usage: true }` for token usage in streaming responses. Default: true. */
+	supportsUsageInStreaming?: boolean;
 	/** Which field to use for max tokens. Default: auto-detected from URL. */
 	maxTokensField?: "max_completion_tokens" | "max_tokens";
 	/** Whether tool results require the `name` field. Default: auto-detected from URL. */