@sayknow-cli/ai 0.3.9 → 0.3.10

package/src/provider-models/openai-compat.ts

@@ -665,6 +665,7 @@ export function deepseekModelManagerOptions(
 ): ModelManagerOptions<"openai-completions"> {
 	return createSimpleOpenAICompletionsOptions("deepseek", "https://api.deepseek.com", config);
 }
+
 export interface DeepInfraModelManagerConfig {
 	apiKey?: string;
 	baseUrl?: string;
@@ -673,13 +674,8 @@ export interface DeepInfraModelManagerConfig {
 export function deepinfraModelManagerOptions(
 	config?: DeepInfraModelManagerConfig,
 ): ModelManagerOptions<"openai-completions"> {
-	return createSimpleOpenAICompletionsOptions(
-		"deepinfra" as Parameters<typeof getBundledModels>[0],
-		"https://api.deepinfra.com/v1/openai",
-		config,
-	);
+	return createSimpleOpenAICompletionsOptions("deepinfra", "https://api.deepinfra.com/v1/openai", config);
 }
-
 // ---------------------------------------------------------------------------
 // 7.5 Fireworks
 // ---------------------------------------------------------------------------
@@ -2269,8 +2265,6 @@ const MODELS_DEV_PROVIDER_DESCRIPTORS_CORE: readonly ModelsDevProviderDescriptor
 	}),
 	// --- xAI ---
 	openAiCompletionsDescriptor("xai", "xai", "https://api.x.ai/v1"),
-	// --- DeepInfra ---
-	openAiCompletionsDescriptor("deepinfra", "deepinfra", "https://api.deepinfra.com/v1/openai"),
 	// --- DeepSeek ---
 	openAiCompletionsDescriptor("deepseek", "deepseek", "https://api.deepseek.com", {
 		// Only ship the v4 family as built-ins; older deepseek-chat / deepseek-reasoner
@@ -2299,6 +2293,8 @@ const MODELS_DEV_PROVIDER_DESCRIPTORS_CORE: readonly ModelsDevProviderDescriptor
 			requiresAssistantContentForToolCalls: true,
 		},
 	}),
+	// --- DeepInfra ---
+	openAiCompletionsDescriptor("deepinfra", "deepinfra", "https://api.deepinfra.com/v1/openai"),
 ];
 
 const MODELS_DEV_PROVIDER_DESCRIPTORS_CODING_PLANS: readonly ModelsDevProviderDescriptor[] = [

package/src/types.ts

@@ -187,7 +187,7 @@ export type CacheRetention = "none" | "short" | "long";
  *
  * The unscoped values (`"auto"`, `"default"`, `"flex"`, `"scale"`,
  * `"priority"`) are passed through to providers that understand them
- * (OpenAI's `service_tier` field directly; Anthropic translates
+ * (OpenAI and DeepInfra's `service_tier` field directly; Anthropic translates
  * `"priority"` into `speed: "fast"` on supported Opus models).
  *
  * The scoped values target a specific provider family and behave as the
@@ -224,21 +224,18 @@ export function resolveServiceTier(
 }
 
 /**
- * True when the (possibly scoped) tier should be sent as OpenAI's
- * `service_tier` request field for the given provider. OpenAI accepts
- * `flex`, `scale`, and `priority`; DeepInfra accepts `priority`.
- * Unsupported tiers (`"auto"`, `"default"`) and scope mismatches return false.
+ * True when the (possibly scoped) tier should be sent as an OpenAI-compatible
+ * `service_tier` request field for providers that support it. Unsupported tiers
+ * (`"auto"`, `"default"`) and scope mismatches all return false.
  */
 export function shouldSendServiceTier(
 	serviceTier: ServiceTier | null | undefined,
 	provider: Provider | undefined,
 ): boolean {
 	const resolved = resolveServiceTier(serviceTier, provider);
-	if (provider === "openai" || provider === "openai-codex") {
-		return resolved === "flex" || resolved === "scale" || resolved === "priority";
-	}
 	if (provider === "deepinfra") return resolved === "priority";
-	return false;
+	if (provider !== "openai" && provider !== "openai-codex") return false;
+	return resolved === "flex" || resolved === "scale" || resolved === "priority";
 }
 
 /**

package/src/utils/oauth/index.ts

@@ -364,7 +364,6 @@ export async function refreshOAuthToken(
 		case "fireworks":
 		case "firepass":
 		case "fugu":
-		case "deepinfra":
 		case "nvidia":
 		case "nanogpt":
 		case "synthetic":