@gajae-code/ai 0.15.5 → 0.16.0

package/src/provider-models/descriptors.ts

@@ -14,6 +14,7 @@ import {
 	bizrouterModelManagerOptions,
 	cerebrasModelManagerOptions,
 	cloudflareAiGatewayModelManagerOptions,
+	commandCodeModelManagerOptions,
 	deepinfraModelManagerOptions,
 	deepseekModelManagerOptions,
 	firepassModelManagerOptions,
@@ -203,6 +204,7 @@ export const PROVIDER_DESCRIPTORS: readonly ProviderDescriptor[] = [
 	),
 	descriptor("opencode-zen", "claude-sonnet-4-6", config => opencodeZenModelManagerOptions(config)),
 	descriptor("opencode-go", "kimi-k2.5", config => opencodeGoModelManagerOptions(config)),
+	descriptor("commandcode-goat", "zai-org/GLM-5.3", config => commandCodeModelManagerOptions(config)),
 	catalogDescriptor(
 		"openrouter",
 		"openai/gpt-5.4",
@@ -370,7 +372,7 @@ export const PROVIDER_DESCRIPTORS: readonly ProviderDescriptor[] = [
 		config => cursorModelManagerOptions(config),
 		catalog("Cursor", ["CURSOR_API_KEY"], { oauthProvider: "cursor" }),
 	),
-	descriptor("kiro", "CLAUDE_3_7_SONNET_20250219_V1_0", config => kiroModelManagerOptions(config)),
+	descriptor("kiro", "auto", config => kiroModelManagerOptions(config)),
 ] as const;
 
 /** Default model IDs for all known providers, built from descriptors + special providers. */

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

@@ -832,7 +832,7 @@ export interface OpenCodeModelManagerConfig {
 }
 
 function openCodeModelManagerOptions(
-	providerId: "opencode-go" | "opencode-zen",
+	providerId: "opencode-go" | "opencode-zen" | "commandcode-goat",
 	defaultBaseUrl: string,
 	config?: OpenCodeModelManagerConfig,
 ): ModelManagerOptions<"openai-completions"> {
@@ -873,6 +873,46 @@ export function opencodeGoModelManagerOptions(
 	return openCodeModelManagerOptions("opencode-go", "https://opencode.ai/zen/go/v1", config);
 }
 
+export function commandCodeModelManagerOptions(config?: OpenCodeModelManagerConfig): ModelManagerOptions<Api> {
+	const apiKey = config?.apiKey;
+	const openAiBaseUrl = config?.baseUrl ?? "https://api.commandcode.ai/provider/v1";
+	const anthropicBaseUrl = openAiBaseUrl.replace(/\/v1\/?$/u, "");
+	return {
+		providerId: "commandcode-goat",
+		...(apiKey && {
+			fetchDynamicModels: () =>
+				fetchOpenAICompatibleModels<Api>({
+					api: "openai-completions",
+					provider: "commandcode-goat",
+					baseUrl: openAiBaseUrl,
+					apiKey,
+					mapModel: (_entry, defaults) => {
+						const isClaude = /(?:^|\/)claude(?:[\w.-]|$)/iu.test(defaults.id);
+						if (isClaude) {
+							return {
+								...defaults,
+								api: "anthropic-messages",
+								baseUrl: anthropicBaseUrl,
+							};
+						}
+						return {
+							...defaults,
+							api: "openai-completions",
+							compat: {
+								...defaults.compat,
+								supportsStore: false,
+								supportsDeveloperRole: false,
+								supportsReasoningEffort: false,
+								maxTokensField: "max_tokens",
+								reasoningContentField: "reasoning_content",
+							},
+						};
+					},
+				}),
+		}),
+	};
+}
+
 // ---------------------------------------------------------------------------
 // 9. Ollama
 // ---------------------------------------------------------------------------
@@ -2317,11 +2357,15 @@ const OPENCODE_GO_BASE_PATH = "https://opencode.ai/zen/go";
 const OPENCODE_ZEN_API_RESOLUTION = createOpenCodeApiResolution("https://opencode.ai/zen");
 const OPENCODE_GO_CHAT_COMPLETIONS_MODEL_IDS = [
 	"deepseek-v4-flash",
+	"deepseek-v4-flash-vision-exp",
 	"deepseek-v4-pro",
 	"glm-5.1",
 	"glm-5.2",
+	"glm-5.3-flash",
 	"kimi-k2.6",
 	"kimi-k2.7-code",
+	"hy4-preview",
+	"longcat-2.0",
 	"mimo-v2.5",
 	"mimo-v2.5-pro",
 ] as const;
@@ -2332,6 +2376,7 @@ const OPENCODE_GO_MESSAGES_MODEL_IDS = [
 	"qwen3.6-plus",
 	"qwen3.7-max",
 	"qwen3.7-plus",
+	"qwen3.8-flash",
 ] as const;
 const OPENCODE_GO_API_OVERRIDES: Readonly<Record<string, Api>> = {
 	...Object.fromEntries(OPENCODE_GO_CHAT_COMPLETIONS_MODEL_IDS.map(id => [id, "openai-completions"])),
@@ -2756,6 +2801,62 @@ const OPENCODE_GO_OFFICIAL_MODELS: Readonly<Record<string, OpenCodeGoOfficialMod
 		reasoning: true,
 		cost: { input: 0.066, output: 0.26, cacheRead: 0.029, cacheWrite: 0 },
 	},
+	"deepseek-v4-flash-vision-exp": {
+		name: "DeepSeek V4 Flash Vision Exp",
+		contextWindow: 1_000_000,
+		maxTokens: 384_000,
+		input: ["text", "image"],
+		reasoning: true,
+		cost: { input: 0.22, output: 0.66, cacheRead: 0.007, cacheWrite: 0 },
+	},
+	"glm-5.3-flash": {
+		name: "GLM-5.3-Flash (2x usage)",
+		contextWindow: 1_000_000,
+		maxTokens: 131_072,
+		input: ["text", "image"],
+		reasoning: true,
+		cost: { input: 0.075, output: 0.25, cacheRead: 0.015, cacheWrite: 0 },
+	},
+	"grok-4.6": {
+		name: "Grok 4.6",
+		contextWindow: 500_000,
+		maxTokens: 500_000,
+		input: ["text", "image"],
+		reasoning: true,
+		cost: { input: 2, output: 6, cacheRead: 0.5, cacheWrite: 0 },
+	},
+	"hy4-preview": {
+		name: "Hy4 preview",
+		contextWindow: 1_024_000,
+		maxTokens: 64_000,
+		input: ["text"],
+		reasoning: true,
+		cost: { input: 0.834, output: 2.501, cacheRead: 0.042, cacheWrite: 0 },
+	},
+	"longcat-2.0": {
+		name: "LongCat-2.0",
+		contextWindow: 1_000_000,
+		maxTokens: 131_072,
+		input: ["text"],
+		reasoning: true,
+		cost: { input: 0.3, output: 1.2, cacheRead: 0.006, cacheWrite: 0 },
+	},
+	"muse-spark-1.2-contributor": {
+		name: "Muse Spark 1.2 Contributor",
+		contextWindow: 1_048_576,
+		maxTokens: 131_072,
+		input: ["text", "image"],
+		reasoning: true,
+		cost: { input: 0.1, output: 0.2, cacheRead: 0.002, cacheWrite: 0 },
+	},
+	"qwen3.8-flash": {
+		name: "Qwen3.8 Flash",
+		contextWindow: 1_000_000,
+		maxTokens: 131_072,
+		input: ["text", "image"],
+		reasoning: true,
+		cost: { input: 0.15, output: 0.47, cacheRead: 0.016, cacheWrite: 0.2 },
+	},
 };
 
 function applyOpenCodeGoOfficialMetadata<TApi extends Api>(model: Model<TApi>): Model<TApi> {

package/src/providers/amazon-bedrock.ts

@@ -8,6 +8,7 @@
  */
 
 import { $credentialEnv, $env, $flag, extractHttpStatusFromError, fetchWithRetry } from "@gajae-code/utils";
+import { assertAwsRegionLabel } from "../adapter-internals/aws-region";
 import type { Effort } from "../model-thinking";
 import {
 	mapEffortToAnthropicAdaptiveEffort,
@@ -202,9 +203,10 @@ export const streamBedrock: StreamFunction<"bedrock-converse-stream"> = (
 
 		const blocks = output.content as Block[];
 		let rawRequestDump: RawHttpRequestDump | undefined;
-		const region = options.region || $env.AWS_REGION || $env.AWS_DEFAULT_REGION || "us-east-1";
+		const region = options.region ?? $env.AWS_REGION ?? $env.AWS_DEFAULT_REGION ?? "us-east-1";
 
 		try {
+			assertAwsRegionLabel(region);
 			const cacheRetention = resolveCacheRetention(options.cacheRetention);
 			const resolvedToolChoice = resolveToolChoice(model, options.toolChoice);
 			const toolConfig = convertToolConfig(context.tools, resolvedToolChoice.resolvedChoice);
@@ -303,6 +305,7 @@ export const streamBedrock: StreamFunction<"bedrock-converse-stream"> = (
 					method: "POST",
 					headers: await buildRequestHeaders(retryBody),
 					body: retryBody,
+					redirect: "error",
 					signal: options.signal,
 					maxAttempts: 1,
 				});
@@ -312,6 +315,7 @@ export const streamBedrock: StreamFunction<"bedrock-converse-stream"> = (
 				method: "POST",
 				headers: requestHeaders,
 				body,
+				redirect: "error",
 				signal: options.signal,
 				maxAttempts: resolveRetryBudget(options.requestMaxRetries, 4) + 1,
 			});

package/src/providers/anthropic.d.ts

@@ -107,7 +107,7 @@ export interface CpaToolAliasRestoreFailure {
  */
 export declare function parseCpaToolAliasRestoreFailure(error: unknown): CpaToolAliasRestoreFailure | undefined;
 export declare function isCpaToolAliasRestoreFailure(error: unknown): boolean;
-export declare const claudeCodeVersion = "2.1.219";
+export declare const claudeCodeVersion = "2.1.257";
 export declare const claudeCodeEntrypoint = "sdk-cli";
 export declare const claudeToolPrefix: string;
 export declare const claudeCodeSystemInstruction = "You are a Claude agent, built on Anthropic's Claude Agent SDK.";

package/src/providers/anthropic.ts

@@ -828,7 +828,7 @@ function getCacheControl(
 }
 
 // Stealth mode: Mimic Anthropic Code headers and tool prefixing.
-export const claudeCodeVersion = "2.1.219";
+export const claudeCodeVersion = "2.1.257";
 export const claudeCodeEntrypoint = "sdk-cli";
 export const claudeToolPrefix: string = "proxy_";
 export const claudeCodeSystemInstruction = "You are a Claude agent, built on Anthropic's Claude Agent SDK.";
@@ -2143,6 +2143,7 @@ export const streamAnthropic: StreamFunction<"anthropic-messages"> = (
 				const idleTimeoutAbortError = new Error("Anthropic stream stalled while waiting for the next event");
 				const { requestSignal } = activeAbortTracker;
 				setRawRequestDump(params);
+				options?.onStreamCreated?.();
 				const anthropicRequest = client.messages.create(
 					{ ...params, stream: true },
 					{
@@ -2616,6 +2617,7 @@ export const streamAnthropic: StreamFunction<"anthropic-messages"> = (
 					}
 					if (
 						!options?.fallbackManaged &&
+						!options?.disableProviderRetries &&
 						!disableStrictTools &&
 						firstTokenTime === undefined &&
 						hasStrictAnthropicTools(params) &&
@@ -2636,6 +2638,7 @@ export const streamAnthropic: StreamFunction<"anthropic-messages"> = (
 						!droppedForcedToolChoice &&
 						firstTokenTime === undefined &&
 						!options?.fallbackManaged &&
+						!options?.disableProviderRetries &&
 						isSentForcedAnthropicToolChoice(params.tool_choice) &&
 						isForcedToolChoiceUnsupportedError(streamFailure, true)
 					) {
@@ -2667,6 +2670,7 @@ export const streamAnthropic: StreamFunction<"anthropic-messages"> = (
 					const maskedProxyRejection = isAnthropicMaskedProxyRejection(streamFailure);
 					if (
 						!options?.fallbackManaged &&
+						!options?.disableProviderRetries &&
 						thinkingReplayRepairScope === "none" &&
 						thinkingReplayRepairAttempts < ANTHROPIC_MAX_THINKING_REPAIRS &&
 						firstTokenTime === undefined &&
@@ -2804,6 +2808,7 @@ export const streamAnthropic: StreamFunction<"anthropic-messages"> = (
 					}
 					if (
 						!options?.fallbackManaged &&
+						!options?.disableProviderRetries &&
 						!dropFastMode &&
 						resolveServiceTier(options?.serviceTier, model.provider) === "priority" &&
 						firstTokenTime === undefined &&
@@ -2825,6 +2830,7 @@ export const streamAnthropic: StreamFunction<"anthropic-messages"> = (
 					}
 					if (
 						!options?.fallbackManaged &&
+						!options?.disableProviderRetries &&
 						generatedCacheBudget > 0 &&
 						firstTokenTime === undefined &&
 						isAnthropicCacheBreakpointOverflowError(streamFailure)
@@ -2862,7 +2868,7 @@ export const streamAnthropic: StreamFunction<"anthropic-messages"> = (
 					// Anthropic and non-CPA proxies are untouched.
 					const cpaAliasFailure = parseCpaToolAliasRestoreFailure(streamFailure);
 					if (cpaAliasFailure && firstTokenTime === undefined) {
-						if (options?.fallbackManaged) {
+						if (options?.fallbackManaged || options?.disableProviderRetries) {
 							// The managed fallback controller owns retries: never retry
 							// inside the attempt it handed us. Record the corrective
 							// steering against this exact turn and surface the raw error;

package/src/providers/aws-credentials.ts

@@ -22,6 +22,7 @@
 import * as fs from "node:fs";
 import * as path from "node:path";
 import { $env, getTrustedHomeDir, isEnoent, logger } from "@gajae-code/utils";
+import { assertAwsRegionLabel } from "../adapter-internals/aws-region";
 import {
 	type AwsIniFile,
 	classifyAwsProfileCapability,
@@ -155,6 +156,7 @@ async function readSsoCredentials(
 		}
 	}
 	if (!startUrl || !ssoRegion) return undefined;
+	assertAwsRegionLabel(ssoRegion);
 
 	const token = await loadSsoCachedToken(startUrl, sessionName);
 	if (!token?.accessToken) {
@@ -172,6 +174,7 @@ async function readSsoCredentials(
 	const response = await fetch(url, {
 		method: "GET",
 		headers: { "x-amz-sso_bearer_token": token.accessToken },
+		redirect: "error",
 		signal,
 	});
 	if (!response.ok) {
@@ -411,6 +414,7 @@ async function readImdsCredentials(parentSignal: AbortSignal | undefined): Promi
 		const tokenRes = await fetch(`http://${IMDS_HOST}/latest/api/token`, {
 			method: "PUT",
 			headers: { "x-aws-ec2-metadata-token-ttl-seconds": "21600" },
+			redirect: "error",
 			signal,
 		});
 		if (!tokenRes.ok) return undefined;
@@ -418,6 +422,7 @@ async function readImdsCredentials(parentSignal: AbortSignal | undefined): Promi
 
 		const roleRes = await fetch(`http://${IMDS_HOST}/latest/meta-data/iam/security-credentials/`, {
 			headers: { "x-aws-ec2-metadata-token": token },
+			redirect: "error",
 			signal,
 		});
 		if (!roleRes.ok) return undefined;
@@ -428,6 +433,7 @@ async function readImdsCredentials(parentSignal: AbortSignal | undefined): Promi
 			`http://${IMDS_HOST}/latest/meta-data/iam/security-credentials/${encodeURIComponent(role)}`,
 			{
 				headers: { "x-aws-ec2-metadata-token": token },
+				redirect: "error",
 				signal,
 			},
 		);

package/src/providers/azure-openai-responses.ts

@@ -142,11 +142,13 @@ export const streamAzureOpenAIResponses: StreamFunction<"azure-openai-responses"
 			};
 			let openaiStream: Awaited<ReturnType<typeof client.responses.create>>;
 			try {
+				options?.onStreamCreated?.();
 				openaiStream = await client.responses.create(params, { signal: requestSignal });
 			} catch (error) {
 				if (
 					!isForcedToolChoiceUnsupportedError(error, isForcedAzureResponsesToolChoice(params.tool_choice)) ||
-					options?.fallbackManaged
+					options?.fallbackManaged ||
+					options?.disableProviderRetries
 				) {
 					throw error;
 				}
@@ -165,6 +167,7 @@ export const streamAzureOpenAIResponses: StreamFunction<"azure-openai-responses"
 				});
 				delete params.tool_choice;
 				rawRequestDump = { ...rawRequestDump, body: params };
+				options?.onStreamCreated?.();
 				openaiStream = await client.responses.create(params, { signal: requestSignal });
 			}
 			streamConnected = true;