@clinebot/llms 0.0.0 → 0.0.1

package/dist/providers/handlers/openai-responses.d.ts

@@ -17,14 +17,23 @@ import { BaseHandler } from "./base";
 /**
  * Convert messages to Responses API input format
  */
-declare function convertToResponsesInput(messages: Message[]): {
+declare function convertToResponsesInput(messages: Message[]): ({
     type: "message";
     role: "user" | "assistant";
-    content: string | Array<{
+    content: Array<{
         type: "input_text" | "output_text";
         text: string;
     }>;
-}[];
+} | {
+    type: "function_call";
+    call_id: string;
+    name: string;
+    arguments: string;
+} | {
+    type: "function_call_output";
+    call_id: string;
+    output: string;
+})[];
 /**
  * Handler for OpenAI Responses API
  *
@@ -54,7 +63,10 @@ export declare class OpenAIResponsesHandler extends BaseHandler {
     /**
      * Process a single chunk from the Responses API stream
      */
-    protected processResponseChunk(chunk: any, _modelInfo: ModelInfo, responseId: string): Generator<import("../types").ApiStreamChunk>;
+    protected processResponseChunk(chunk: any, _modelInfo: ModelInfo, responseId: string, functionCallMetadataByItemId: Map<string, {
+        callId?: string;
+        name?: string;
+    }>): Generator<import("../types").ApiStreamChunk>;
     private getApiResponseId;
 }
 /**

package/dist/providers/types/config.d.ts

@@ -73,7 +73,7 @@ export interface TokenConfig {
  */
 export interface ReasoningConfig {
     /** Reasoning effort level */
-    reasoningEffort?: "low" | "medium" | "high";
+    reasoningEffort?: "low" | "medium" | "high" | "xhigh";
     /** Extended thinking budget in tokens */
     thinkingBudgetTokens?: number;
     /** Enable thinking with provider/model defaults when supported */

package/dist/providers/types/messages.d.ts

@@ -45,6 +45,8 @@ export interface ToolUseContent {
     type: "tool_use";
     /** Unique ID for this tool call */
     id: string;
+    /** Provider-native call ID for this tool call (if available) */
+    call_id?: string;
     /** Name of the tool being called */
     name: string;
     /** Arguments for the tool call */

package/dist/providers/types/settings.d.ts

@@ -27,6 +27,7 @@ export declare const ReasoningSettingsSchema: z.ZodObject<{
         low: "low";
         high: "high";
         medium: "medium";
+        xhigh: "xhigh";
         none: "none";
     }>>;
     budgetTokens: z.ZodOptional<z.ZodNumber>;
@@ -136,6 +137,7 @@ export declare const ProviderSettingsSchema: z.ZodObject<{
             low: "low";
             high: "high";
             medium: "medium";
+            xhigh: "xhigh";
             none: "none";
         }>>;
         budgetTokens: z.ZodOptional<z.ZodNumber>;

package/package.json

@@ -1,6 +1,7 @@
 {
 	"name": "@clinebot/llms",
-	"version": "0.0.0",
+	"version": "0.0.1",
+	"description": "Config-driven SDK for selecting, extending, and instantiating LLM providers and models",
 	"main": "./dist/index.js",
 	"module": "./dist/index.js",
 	"dependencies": {
@@ -17,7 +18,7 @@
 		"ai-sdk-provider-codex-cli": "^1.1.0",
 		"ai-sdk-provider-opencode-sdk": "^2.1.2",
 		"dify-ai-provider": "^1.1.0",
-		"nanoid": "^5.1.6",
+		"nanoid": "^5.1.7",
 		"openai": "^6.25.0",
 		"zod": "^4.3.6"
 	},
@@ -43,7 +44,6 @@
 			"types": "./dist/index.browser.d.ts"
 		}
 	},
-	"description": "Config-driven SDK for selecting, extending, and instantiating LLM providers and models",
 	"files": [
 		"dist",
 		"src"

package/src/live-providers.test.ts

@@ -25,8 +25,6 @@ const PROVIDER_TIMEOUT_MS = Number(
 	process.env.LLMS_LIVE_PROVIDER_TIMEOUT_MS ?? "90000",
 );
 
-const liveDescribe = LIVE_TEST_ENABLED ? describe : describe.skip;
-
 function requireProvidersFilePath(): string {
 	const filePath = process.env[PROVIDERS_FILE_ENV];
 	if (!filePath) {
@@ -108,8 +106,11 @@ async function runPrompt(target: ProviderTarget): Promise<void> {
 	}
 }
 
-liveDescribe("live provider smoke test", () => {
+describe("live provider smoke test", () => {
 	it("reads configured providers from json and reports providers with failed responses", async () => {
+		if (!LIVE_TEST_ENABLED) {
+			return null;
+		}
 		const filePath = requireProvidersFilePath();
 		const targets = loadProviderTargets(filePath);
 

package/src/models/generated.ts

@@ -14,7 +14,7 @@ export const GENERATED_PROVIDER_MODELS: {
 	version: number;
 	providers: Record<string, Record<string, ModelInfo>>;
 } = {
-	version: 1773798019500,
+	version: 1773862738565,
 	providers: {
 		aihubmix: {
 			"claude-sonnet-4-6": {
@@ -1124,7 +1124,7 @@ export const GENERATED_PROVIDER_MODELS: {
 			"anthropic.claude-sonnet-4-6": {
 				id: "anthropic.claude-sonnet-4-6",
 				name: "Claude Sonnet 4.6",
-				contextWindow: 200000,
+				contextWindow: 1000000,
 				maxTokens: 64000,
 				capabilities: ["images", "files", "tools", "reasoning", "temperature"],
 				pricing: {
@@ -1138,7 +1138,7 @@ export const GENERATED_PROVIDER_MODELS: {
 			"eu.anthropic.claude-sonnet-4-6": {
 				id: "eu.anthropic.claude-sonnet-4-6",
 				name: "Claude Sonnet 4.6 (EU)",
-				contextWindow: 200000,
+				contextWindow: 1000000,
 				maxTokens: 64000,
 				capabilities: ["images", "files", "tools", "reasoning", "temperature"],
 				pricing: {
@@ -1152,7 +1152,7 @@ export const GENERATED_PROVIDER_MODELS: {
 			"global.anthropic.claude-sonnet-4-6": {
 				id: "global.anthropic.claude-sonnet-4-6",
 				name: "Claude Sonnet 4.6 (Global)",
-				contextWindow: 200000,
+				contextWindow: 1000000,
 				maxTokens: 64000,
 				capabilities: ["images", "files", "tools", "reasoning", "temperature"],
 				pricing: {
@@ -1180,7 +1180,7 @@ export const GENERATED_PROVIDER_MODELS: {
 			"us.anthropic.claude-sonnet-4-6": {
 				id: "us.anthropic.claude-sonnet-4-6",
 				name: "Claude Sonnet 4.6 (US)",
-				contextWindow: 200000,
+				contextWindow: 1000000,
 				maxTokens: 64000,
 				capabilities: ["images", "files", "tools", "reasoning", "temperature"],
 				pricing: {
@@ -1222,7 +1222,7 @@ export const GENERATED_PROVIDER_MODELS: {
 			"anthropic.claude-opus-4-6-v1": {
 				id: "anthropic.claude-opus-4-6-v1",
 				name: "Claude Opus 4.6",
-				contextWindow: 200000,
+				contextWindow: 1000000,
 				maxTokens: 128000,
 				capabilities: ["images", "files", "tools", "reasoning", "temperature"],
 				pricing: {
@@ -1236,7 +1236,7 @@ export const GENERATED_PROVIDER_MODELS: {
 			"eu.anthropic.claude-opus-4-6-v1": {
 				id: "eu.anthropic.claude-opus-4-6-v1",
 				name: "Claude Opus 4.6 (EU)",
-				contextWindow: 200000,
+				contextWindow: 1000000,
 				maxTokens: 128000,
 				capabilities: ["images", "files", "tools", "reasoning", "temperature"],
 				pricing: {
@@ -1250,7 +1250,7 @@ export const GENERATED_PROVIDER_MODELS: {
 			"global.anthropic.claude-opus-4-6-v1": {
 				id: "global.anthropic.claude-opus-4-6-v1",
 				name: "Claude Opus 4.6 (Global)",
-				contextWindow: 200000,
+				contextWindow: 1000000,
 				maxTokens: 128000,
 				capabilities: ["images", "files", "tools", "reasoning", "temperature"],
 				pricing: {
@@ -1264,7 +1264,7 @@ export const GENERATED_PROVIDER_MODELS: {
 			"us.anthropic.claude-opus-4-6-v1": {
 				id: "us.anthropic.claude-opus-4-6-v1",
 				name: "Claude Opus 4.6 (US)",
-				contextWindow: 200000,
+				contextWindow: 1000000,
 				maxTokens: 128000,
 				capabilities: ["images", "files", "tools", "reasoning", "temperature"],
 				pricing: {
@@ -3434,6 +3434,34 @@ export const GENERATED_PROVIDER_MODELS: {
 			},
 		},
 		minimax: {
+			"MiniMax-M2.7": {
+				id: "MiniMax-M2.7",
+				name: "MiniMax-M2.7",
+				contextWindow: 204800,
+				maxTokens: 131072,
+				capabilities: ["tools", "reasoning", "temperature"],
+				pricing: {
+					input: 0.3,
+					output: 1.2,
+					cacheRead: 0.06,
+					cacheWrite: 0.375,
+				},
+				releaseDate: "2026-03-18",
+			},
+			"MiniMax-M2.7-highspeed": {
+				id: "MiniMax-M2.7-highspeed",
+				name: "MiniMax-M2.7-highspeed",
+				contextWindow: 204800,
+				maxTokens: 131072,
+				capabilities: ["tools", "reasoning", "temperature"],
+				pricing: {
+					input: 0.6,
+					output: 2.4,
+					cacheRead: 0.06,
+					cacheWrite: 0.375,
+				},
+				releaseDate: "2026-03-18",
+			},
 			"MiniMax-M2.5-highspeed": {
 				id: "MiniMax-M2.5-highspeed",
 				name: "MiniMax-M2.5-highspeed",
@@ -5201,6 +5229,20 @@ export const GENERATED_PROVIDER_MODELS: {
 			},
 		},
 		openrouter: {
+			"minimax/minimax-m2.7": {
+				id: "minimax/minimax-m2.7",
+				name: "MiniMax M2.7",
+				contextWindow: 204800,
+				maxTokens: 131072,
+				capabilities: ["tools", "reasoning", "temperature"],
+				pricing: {
+					input: 0.3,
+					output: 1.2,
+					cacheRead: 0.06,
+					cacheWrite: 0.375,
+				},
+				releaseDate: "2026-03-18",
+			},
 			"x-ai/grok-4.20-beta": {
 				id: "x-ai/grok-4.20-beta",
 				name: "Grok 4.20 Beta",
@@ -9071,6 +9113,34 @@ export const GENERATED_PROVIDER_MODELS: {
 			},
 		},
 		"vercel-ai-gateway": {
+			"minimax/minimax-m2.7": {
+				id: "minimax/minimax-m2.7",
+				name: "Minimax M2.7",
+				contextWindow: 204800,
+				maxTokens: 131000,
+				capabilities: ["images", "files", "tools", "reasoning", "temperature"],
+				pricing: {
+					input: 0.3,
+					output: 1.2,
+					cacheRead: 0.06,
+					cacheWrite: 0.375,
+				},
+				releaseDate: "2026-03-18",
+			},
+			"minimax/minimax-m2.7-highspeed": {
+				id: "minimax/minimax-m2.7-highspeed",
+				name: "MiniMax M2.7 High Speed",
+				contextWindow: 204800,
+				maxTokens: 131100,
+				capabilities: ["images", "tools", "reasoning", "temperature"],
+				pricing: {
+					input: 0.6,
+					output: 2.4,
+					cacheRead: 0.06,
+					cacheWrite: 0.375,
+				},
+				releaseDate: "2026-03-18",
+			},
 			"openai/gpt-5.4-mini": {
 				id: "openai/gpt-5.4-mini",
 				name: "GPT 5.4 Mini",

package/src/models/providers/gemini.ts

@@ -15,7 +15,7 @@ export const GEMINI_PROVIDER: ModelCollection = {
 	provider: {
 		id: "gemini",
 		name: "Google Gemini",
-		description: "Google's multimodal AI models",
+		description: "Google Gemini API",
 		protocol: "gemini",
 		baseUrl: "https://generativelanguage.googleapis.com",
 		defaultModelId: GEMINI_DEFAULT_MODEL,

package/src/providers/handlers/anthropic-base.ts

@@ -145,7 +145,7 @@ export class AnthropicHandler extends BaseHandler {
 				thinking: reasoningOn
 					? { type: "enabled", budget_tokens: budgetTokens }
 					: undefined,
-				max_tokens: model.info.maxTokens || 8192,
+				max_tokens: model.info.maxTokens ?? this.config.maxOutputTokens ?? 8192,
 				temperature: reasoningOn ? undefined : 0,
 				system: supportsPromptCache
 					? [

package/src/providers/handlers/bedrock-base.ts

@@ -143,7 +143,7 @@ export class BedrockHandler extends BaseHandler {
 			model: factory(modelId),
 			messages: this.getMessages(systemPrompt, messages),
 			tools: toAiSdkTools(tools),
-			maxTokens: model.info.maxTokens ?? undefined,
+			maxTokens: model.info.maxTokens ?? this.config.maxOutputTokens ?? 8192,
 			temperature: reasoningEnabled ? undefined : (model.info.temperature ?? 0),
 			providerOptions:
 				Object.keys(providerOptions).length > 0 ? providerOptions : undefined,

package/src/providers/handlers/gemini-base.test.ts

@@ -0,0 +1,221 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+import type { ToolDefinition } from "../types/messages";
+import type { ApiStreamChunk } from "../types/stream";
+
+const generateContentStreamSpy = vi.fn();
+const googleGenAIConstructorSpy = vi.fn();
+
+vi.mock("@google/genai", () => {
+	class GoogleGenAI {
+		models = {
+			generateContentStream: generateContentStreamSpy,
+		};
+
+		constructor(config: unknown) {
+			googleGenAIConstructorSpy(config);
+		}
+	}
+
+	return {
+		GoogleGenAI,
+		FunctionCallingConfigMode: { AUTO: "AUTO" },
+		ThinkingLevel: { HIGH: "HIGH", LOW: "LOW" },
+	};
+});
+
+import { GeminiHandler } from "./gemini-base";
+
+function createAsyncIterable<T>(items: T[]): AsyncIterable<T> {
+	return {
+		async *[Symbol.asyncIterator]() {
+			for (const item of items) {
+				yield item;
+			}
+		},
+	};
+}
+
+async function collectChunks(stream: AsyncIterable<ApiStreamChunk>) {
+	const chunks: ApiStreamChunk[] = [];
+	for await (const chunk of stream) {
+		chunks.push(chunk);
+	}
+	return chunks;
+}
+
+describe("GeminiHandler", () => {
+	beforeEach(() => {
+		vi.clearAllMocks();
+	});
+
+	it("preserves per-call ids for parallel function calls and keeps falsy args", async () => {
+		generateContentStreamSpy.mockResolvedValue(
+			createAsyncIterable([
+				{
+					candidates: [
+						{
+							content: {
+								parts: [
+									{
+										functionCall: {
+											id: "call_a",
+											name: "power_disco_ball",
+											args: { power: false },
+										},
+									},
+									{
+										functionCall: {
+											id: "call_b",
+											name: "dim_lights",
+											args: { brightness: 0 },
+										},
+									},
+								],
+							},
+						},
+					],
+					usageMetadata: {
+						promptTokenCount: 10,
+						candidatesTokenCount: 4,
+					},
+				},
+			]),
+		);
+
+		const handler = new GeminiHandler({
+			providerId: "gemini",
+			modelId: "gemini-2.5-flash",
+			apiKey: "test-key",
+			modelInfo: {
+				id: "gemini-2.5-flash",
+				contextWindow: 1_000_000,
+				maxTokens: 8192,
+				temperature: 1,
+			},
+		});
+
+		const tools: ToolDefinition[] = [
+			{
+				name: "power_disco_ball",
+				description: "toggle disco ball power",
+				inputSchema: {
+					type: "object",
+					properties: { power: { type: "boolean" } },
+					required: ["power"],
+				},
+			},
+			{
+				name: "dim_lights",
+				description: "set light brightness",
+				inputSchema: {
+					type: "object",
+					properties: { brightness: { type: "number" } },
+					required: ["brightness"],
+				},
+			},
+		];
+
+		const chunks = await collectChunks(
+			handler.createMessage(
+				"You are helpful.",
+				[{ role: "user", content: "start" }],
+				tools,
+			),
+		);
+
+		const toolChunks = chunks.filter((chunk) => chunk.type === "tool_calls");
+		expect(toolChunks).toHaveLength(2);
+		expect(toolChunks[0]).toMatchObject({
+			tool_call: {
+				call_id: "call_a",
+				function: {
+					id: "call_a",
+					name: "power_disco_ball",
+					arguments: { power: false },
+				},
+			},
+		});
+		expect(toolChunks[1]).toMatchObject({
+			tool_call: {
+				call_id: "call_b",
+				function: {
+					id: "call_b",
+					name: "dim_lights",
+					arguments: { brightness: 0 },
+				},
+			},
+		});
+	});
+
+	it("generates distinct fallback ids when Gemini omits functionCall.id", async () => {
+		generateContentStreamSpy.mockResolvedValue(
+			createAsyncIterable([
+				{
+					candidates: [
+						{
+							content: {
+								parts: [
+									{
+										functionCall: {
+											name: "read_file",
+											args: { path: "a.ts" },
+										},
+									},
+									{
+										functionCall: {
+											name: "search_files",
+											args: { query: "TODO" },
+										},
+									},
+								],
+							},
+						},
+					],
+					usageMetadata: {
+						promptTokenCount: 5,
+						candidatesTokenCount: 3,
+					},
+				},
+			]),
+		);
+
+		const handler = new GeminiHandler({
+			providerId: "gemini",
+			modelId: "gemini-2.5-flash",
+			apiKey: "test-key",
+		});
+
+		const chunks = await collectChunks(
+			handler.createMessage(
+				"System",
+				[{ role: "user", content: "go" }],
+				[
+					{
+						name: "read_file",
+						description: "read file",
+						inputSchema: {
+							type: "object",
+							properties: { path: { type: "string" } },
+						},
+					},
+					{
+						name: "search_files",
+						description: "search files",
+						inputSchema: {
+							type: "object",
+							properties: { query: { type: "string" } },
+						},
+					},
+				],
+			),
+		);
+
+		const toolChunks = chunks.filter((chunk) => chunk.type === "tool_calls");
+		expect(toolChunks).toHaveLength(2);
+		const firstId = toolChunks[0].tool_call.call_id;
+		const secondId = toolChunks[1].tool_call.call_id;
+		expect(firstId).toBeTruthy();
+		expect(secondId).toBeTruthy();
+		expect(firstId).not.toBe(secondId);
+	});
+});

package/src/providers/handlers/gemini-base.ts

@@ -138,6 +138,7 @@ export class GeminiHandler extends BaseHandler {
 			abortSignal,
 			systemInstruction: systemPrompt,
 			temperature: info.temperature ?? 1,
+			maxOutputTokens: info.maxTokens ?? this.config.maxOutputTokens,
 		};
 
 		// Add thinking config only when explicitly requested and supported.
@@ -176,6 +177,7 @@ export class GeminiHandler extends BaseHandler {
 			let outputTokens = 0;
 			let cacheReadTokens = 0;
 			let thoughtsTokenCount = 0;
+			let syntheticToolCallIndex = 0;
 
 			for await (const chunk of result) {
 				// Handle content parts
@@ -203,18 +205,19 @@ export class GeminiHandler extends BaseHandler {
 					if (part.functionCall) {
 						// Tool call
 						const functionCall = part.functionCall;
-						const args = Object.entries(functionCall.args ?? {}).filter(
-							([, val]) => !!val,
-						);
-
-						if (args.length > 0) {
+						const callId =
+							functionCall.id ??
+							`${responseId}_tool_${syntheticToolCallIndex++}`;
+						if (functionCall.name) {
 							yield {
 								type: "tool_calls",
 								tool_call: {
+									call_id: callId,
 									function: {
-										id: responseId,
+										id: callId,
 										name: functionCall.name,
-										arguments: functionCall.args as Record<string, unknown>,
+										arguments:
+											(functionCall.args as Record<string, unknown>) ?? {},
 									},
 								},
 								id: responseId,

package/src/providers/handlers/openai-base.ts

@@ -150,8 +150,9 @@ export class OpenAIBaseHandler extends BaseHandler {
 		};
 
 		// Add max tokens if configured
-		if (modelInfo.maxTokens) {
-			requestOptions.max_completion_tokens = modelInfo.maxTokens;
+		const maxTokens = modelInfo.maxTokens ?? this.config.maxOutputTokens;
+		if (maxTokens) {
+			requestOptions.max_completion_tokens = maxTokens;
 		}
 
 		// Add temperature if not a reasoning model