update LLM models

2026-01-16 08:21:55 +00:00 · 2025-05-29 22:40:20 +08:00 · 2025-05-29 22:40:20 +08:00 · 120c442274
commit 120c442274
parent 48b95ea416
9 changed files with 1014 additions and 536 deletions
--- a/package.json
+++ b/package.json
@ -56,6 +56,7 @@
 		"@codemirror/lang-markdown": "^6.3.2",
 		"@codemirror/merge": "^6.10.0",
 		"@electric-sql/pglite": "0.2.14",
+		"@google/genai": "^1.2.0",
 		"@google/generative-ai": "^0.21.0",
 		"@langchain/core": "^0.3.26",
 		"@lexical/clipboard": "^0.17.1",
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
--- a/src/constants.ts
+++ b/src/constants.ts
@ -33,7 +33,7 @@ export const OPENROUTER_BASE_URL = 'https://openrouter.ai/api/v1'
 export const GROK_BASE_URL = 'https://api.x.ai/v1'
 export const SILICONFLOW_BASE_URL = 'https://api.siliconflow.cn/v1'
 export const ALIBABA_QWEN_BASE_URL = 'https://dashscope.aliyuncs.com/compatible-mode/v1'
-export const INFIO_BASE_URL = 'https://api.infio.com/api/raw_message'
+export const INFIO_BASE_URL = 'https://api.infio.app'
 export const JINA_BASE_URL = 'https://r.jina.ai'
 export const SERPER_BASE_URL = 'https://serpapi.com/search'
 // Pricing in dollars per million tokens
--- a/src/core/llm/gemini.ts
+++ b/src/core/llm/gemini.ts
@ -1,11 +1,11 @@
 import {
 	Content,
-	EnhancedGenerateContentResponse,
-	GenerateContentResult,
-	GenerateContentStreamResult,
-	GoogleGenerativeAI,
+	GoogleGenAI,
 	Part,
-} from '@google/generative-ai'
+	type GenerateContentConfig,
+	type GenerateContentParameters,
+	type GenerateContentResponse,
+} from "@google/genai"

 import { LLMModel } from '../../types/llm/model'
 import {
@ -18,6 +18,12 @@ import {
 	LLMResponseNonStreaming,
 	LLMResponseStreaming,
 } from '../../types/llm/response'
+import {
+	GeminiModelId,
+	ModelInfo,
+	geminiDefaultModelId,
+	geminiModels
+} from "../../utils/api"
 import { parseImageDataUrl } from '../../utils/image'

 import { BaseLLMProvider } from './base'
@ -34,12 +40,41 @@ import {
 * issues are resolved.
 */
 export class GeminiProvider implements BaseLLMProvider {
-	private client: GoogleGenerativeAI
+	private client: GoogleGenAI
 	private apiKey: string
+	private baseUrl: string

-	constructor(apiKey: string) {
+	constructor(apiKey: string, baseUrl?: string) {
 		this.apiKey = apiKey
-		this.client = new GoogleGenerativeAI(apiKey)
+		this.baseUrl = baseUrl
+		this.client = new GoogleGenAI({ apiKey })
+	}
+
+	getModel(modelId: string) {
+		let id = modelId
+		let info: ModelInfo = geminiModels[id as GeminiModelId]
+
+		if (id?.endsWith(":thinking")) {
+			id = id.slice(0, -":thinking".length)
+
+			if (geminiModels[id as GeminiModelId]) {
+				info = geminiModels[id as GeminiModelId]
+
+				return {
+					id,
+					info,
+					thinkingConfig: undefined,
+					maxOutputTokens: info.maxTokens ?? undefined,
+				}
+			}
+		}
+
+		if (!info) {
+			id = geminiDefaultModelId
+			info = geminiModels[geminiDefaultModelId]
+		}
+
+		return { id, info }
 	}

 	async generateResponse(
@ -53,6 +88,8 @@ export class GeminiProvider implements BaseLLMProvider {
 			)
 		}

+		const { id: modelName, thinkingConfig, maxOutputTokens, info } = this.getModel(model.modelId)
+
 		const systemMessages = request.messages.filter((m) => m.role === 'system')
 		const systemInstruction: string | undefined =
 			systemMessages.length > 0
@ -60,30 +97,26 @@ export class GeminiProvider implements BaseLLMProvider {
 				: undefined

 		try {
-			const model = this.client.getGenerativeModel({
-				model: request.model,
-				generationConfig: {
-					maxOutputTokens: request.max_tokens,
-					temperature: request.temperature,
-					topP: request.top_p,
-					presencePenalty: request.presence_penalty,
-					frequencyPenalty: request.frequency_penalty,
-				},
-				systemInstruction: systemInstruction,
-			})

-			const result = await model.generateContent(
-				{
-					systemInstruction: systemInstruction,
-					contents: request.messages
-						.map((message) => GeminiProvider.parseRequestMessage(message))
-						.filter((m): m is Content => m !== null),
-				},
-				{
-					signal: options?.signal,
-				},
-			)
+			const config: GenerateContentConfig = {
+				systemInstruction,
+				httpOptions: this.baseUrl ? { baseUrl: this.baseUrl } : undefined,
+				thinkingConfig,
+				maxOutputTokens: maxOutputTokens ?? request.max_tokens,
+				temperature: request.temperature ?? 0,
+				topP: request.top_p ?? 1,
+				presencePenalty: request.presence_penalty ?? 0,
+				frequencyPenalty: request.frequency_penalty ?? 0,
+			}
+			const params: GenerateContentParameters = {
+				model: modelName,
+				contents: request.messages
+					.map((message) => GeminiProvider.parseRequestMessage(message))
+					.filter((m): m is Content => m !== null),
+				config,
+			}

+			const result = await this.client.models.generateContent(params)
 			const messageId = crypto.randomUUID() // Gemini does not return a message id
 			return GeminiProvider.parseNonStreamingResponse(
 				result,
@ -115,6 +148,7 @@ export class GeminiProvider implements BaseLLMProvider {
 				`Gemini API key is missing. Please set it in settings menu.`,
 			)
 		}
+		const { id: modelName, thinkingConfig, maxOutputTokens, info } = this.getModel(model.modelId)

 		const systemMessages = request.messages.filter((m) => m.role === 'system')
 		const systemInstruction: string | undefined =
@ -123,30 +157,25 @@ export class GeminiProvider implements BaseLLMProvider {
 				: undefined

 		try {
-			const model = this.client.getGenerativeModel({
-				model: request.model,
-				generationConfig: {
-					maxOutputTokens: request.max_tokens,
-					temperature: request.temperature,
-					topP: request.top_p,
-					presencePenalty: request.presence_penalty,
-					frequencyPenalty: request.frequency_penalty,
-				},
-				systemInstruction: systemInstruction,
-			})
-
-			const stream = await model.generateContentStream(
-				{
-					systemInstruction: systemInstruction,
-					contents: request.messages
-						.map((message) => GeminiProvider.parseRequestMessage(message))
-						.filter((m): m is Content => m !== null),
-				},
-				{
-					signal: options?.signal,
-				},
-			)
+			const config: GenerateContentConfig = {
+				systemInstruction,
+				httpOptions: this.baseUrl ? { baseUrl: this.baseUrl } : undefined,
+				thinkingConfig,
+				maxOutputTokens: maxOutputTokens ?? request.max_tokens,
+				temperature: request.temperature ?? 0,
+				topP: request.top_p ?? 1,
+				presencePenalty: request.presence_penalty ?? 0,
+				frequencyPenalty: request.frequency_penalty ?? 0,
+			}
+			const params: GenerateContentParameters = {
+				model: modelName,
+				contents: request.messages
+					.map((message) => GeminiProvider.parseRequestMessage(message))
+					.filter((m): m is Content => m !== null),
+				config,
+			}

+			const stream = await this.client.models.generateContentStream(params)
 			const messageId = crypto.randomUUID() // Gemini does not return a message id
 			return this.streamResponseGenerator(stream, request.model, messageId)
 		} catch (error) {
@ -165,11 +194,11 @@ export class GeminiProvider implements BaseLLMProvider {
 	}

 	private async *streamResponseGenerator(
-		stream: GenerateContentStreamResult,
+		stream: AsyncGenerator<GenerateContentResponse>,
 		model: string,
 		messageId: string,
 	): AsyncIterable<LLMResponseStreaming> {
-		for await (const chunk of stream.stream) {
+		for await (const chunk of stream) {
 			yield GeminiProvider.parseStreamingResponseChunk(chunk, model, messageId)
 		}
 	}
@ -215,7 +244,7 @@ export class GeminiProvider implements BaseLLMProvider {
 	}

 	static parseNonStreamingResponse(
-		response: GenerateContentResult,
+		response: GenerateContentResponse,
 		model: string,
 		messageId: string,
 	): LLMResponseNonStreaming {
@ -224,9 +253,9 @@ export class GeminiProvider implements BaseLLMProvider {
 			choices: [
 				{
 					finish_reason:
-						response.response.candidates?.[0]?.finishReason ?? null,
+						response.candidates?.[0]?.finishReason ?? null,
 					message: {
-						content: response.response.text(),
+						content: response.candidates?.[0]?.content?.parts?.[0]?.text ?? '',
 						role: 'assistant',
 					},
 				},
@ -234,29 +263,32 @@ export class GeminiProvider implements BaseLLMProvider {
 			created: Date.now(),
 			model: model,
 			object: 'chat.completion',
-			usage: response.response.usageMetadata
+			usage: response.usageMetadata
 				? {
-					prompt_tokens: response.response.usageMetadata.promptTokenCount,
+					prompt_tokens: response.usageMetadata.promptTokenCount,
 					completion_tokens:
-						response.response.usageMetadata.candidatesTokenCount,
-					total_tokens: response.response.usageMetadata.totalTokenCount,
+						response.usageMetadata.candidatesTokenCount,
+					total_tokens: response.usageMetadata.totalTokenCount,
 				}
 				: undefined,
 		}
 	}

 	static parseStreamingResponseChunk(
-		chunk: EnhancedGenerateContentResponse,
+		chunk: GenerateContentResponse,
 		model: string,
 		messageId: string,
 	): LLMResponseStreaming {
+		const firstCandidate = chunk.candidates?.[0]
+		const textContent = firstCandidate?.content?.parts?.[0]?.text || ''
+		
 		return {
 			id: messageId,
 			choices: [
 				{
-					finish_reason: chunk.candidates?.[0]?.finishReason ?? null,
+					finish_reason: firstCandidate?.finishReason ?? null,
 					delta: {
-						content: chunk.text(),
+						content: textContent,
 					},
 				},
 			],
--- a/src/core/llm/infio.ts
+++ b/src/core/llm/infio.ts
@ -1,251 +0,0 @@
-import OpenAI from 'openai'
-import {
-	ChatCompletion,
-	ChatCompletionChunk,
-} from 'openai/resources/chat/completions'
-
-import { INFIO_BASE_URL } from '../../constants'
-import { LLMModel } from '../../types/llm/model'
-import {
-	LLMRequestNonStreaming,
-	LLMRequestStreaming,
-	RequestMessage
-} from '../../types/llm/request'
-import {
-	LLMResponseNonStreaming,
-	LLMResponseStreaming,
-} from '../../types/llm/response'
-
-import { BaseLLMProvider } from './base'
-import {
-	LLMAPIKeyInvalidException,
-	LLMAPIKeyNotSetException,
-} from './exception'
-
-export type RangeFilter = {
-	gte?: number;
-	lte?: number;
-}
-
-export type ChunkFilter = {
-	field: string;
-	match_all?: string[];
-	range?: RangeFilter;
-}
-
-/**
- * Interface for making requests to the Infio API
- */
-export type InfioRequest = {
-	/** Required: The content of the user message to attach to the topic and then generate an assistant message in response to */
-	messages: RequestMessage[];
-	// /** Required: The ID of the topic to attach the message to */
-	// topic_id: string;
-	/** Optional: URLs to include */
-	links?: string[];
-	/** Optional: Files to include */
-	files?: string[];
-	/** Optional: Whether to highlight results in chunk_html. Default is true */
-	highlight_results?: boolean;
-	/** Optional: Delimiters for highlighting citations. Default is [".", "!", "?", "\n", "\t", ","] */
-	highlight_delimiters?: string[];
-	/** Optional: Search type - "semantic", "fulltext", or "hybrid". Default is "hybrid" */
-	search_type?: string;
-	/** Optional: Filters for chunk filtering */
-	filters?: ChunkFilter;
-	/** Optional: Whether to use web search API. Default is false */
-	use_web_search?: boolean;
-	/** Optional: LLM model to use */
-	llm_model?: string;
-	/** Optional: Force source */
-	force_source?: string;
-	/** Optional: Whether completion should come before chunks in stream. Default is false */
-	completion_first?: boolean;
-	/** Optional: Whether to stream the response. Default is true */
-	stream_response?: boolean;
-	/** Optional: Sampling temperature between 0 and 2. Default is 0.5 */
-	temperature?: number;
-	/** Optional: Frequency penalty between -2.0 and 2.0. Default is 0.7 */
-	frequency_penalty?: number;
-	/** Optional: Presence penalty between -2.0 and 2.0. Default is 0.7 */
-	presence_penalty?: number;
-	/** Optional: Maximum tokens to generate */
-	max_tokens?: number;
-	/** Optional: Stop tokens (up to 4 sequences) */
-	stop_tokens?: string[];
-}
-
-export class InfioProvider implements BaseLLMProvider {
-	//   private adapter: OpenAIMessageAdapter
-	//   private client: OpenAI
-	private apiKey: string
-	private baseUrl: string
-
-	constructor(apiKey: string) {
-		// this.client = new OpenAI({ apiKey, dangerouslyAllowBrowser: true })
-		// this.adapter = new OpenAIMessageAdapter()
-		this.apiKey = apiKey
-		this.baseUrl = INFIO_BASE_URL
-	}
-
-	async generateResponse(
-		model: LLMModel,
-		request: LLMRequestNonStreaming,
-		// options?: LLMOptions,
-	): Promise<LLMResponseNonStreaming> {
-		if (!this.apiKey) {
-			throw new LLMAPIKeyNotSetException(
-				'OpenAI API key is missing. Please set it in settings menu.',
-			)
-		}
-		try {
-			const req: InfioRequest = {
-				messages: request.messages,
-				stream_response: false,
-				temperature: request.temperature,
-				frequency_penalty: request.frequency_penalty,
-				presence_penalty: request.presence_penalty,
-				max_tokens: request.max_tokens,
-			}
-			const req_options = {
-				method: 'POST',
-				headers: {
-					Authorization: this.apiKey,
-					"TR-Dataset": "74aaec22-0cf0-4cba-80a5-ae5c0518344e",
-					'Content-Type': 'application/json'
-				},
-				body: JSON.stringify(req)
-			};
-
-			const response = await fetch(this.baseUrl, req_options);
-			if (!response.ok) {
-				throw new Error(`HTTP error! status: ${response.status}`);
-			}
-			const data = await response.json() as ChatCompletion;
-			return InfioProvider.parseNonStreamingResponse(data);
-		} catch (error) {
-			if (error instanceof OpenAI.AuthenticationError) {
-				throw new LLMAPIKeyInvalidException(
-					'OpenAI API key is invalid. Please update it in settings menu.',
-				)
-			}
-			throw error
-		}
-	}
-
-	async streamResponse(
-		model: LLMModel,
-		request: LLMRequestStreaming,
-	): Promise<AsyncIterable<LLMResponseStreaming>> {
-		if (!this.apiKey) {
-			throw new LLMAPIKeyNotSetException(
-				'OpenAI API key is missing. Please set it in settings menu.',
-			)
-		}
-
-		try {
-			const req: InfioRequest = {
-				llm_model: request.model,
-				messages: request.messages,
-				stream_response: true,
-				temperature: request.temperature,
-				frequency_penalty: request.frequency_penalty,
-				presence_penalty: request.presence_penalty,
-				max_tokens: request.max_tokens,
-			}
-			const req_options = {
-				method: 'POST',
-				headers: {
-					Authorization: this.apiKey,
-					"TR-Dataset": "74aaec22-0cf0-4cba-80a5-ae5c0518344e",
-					"Content-Type": "application/json"
-				},
-				body: JSON.stringify(req)
-			};
-
-			const response = await fetch(this.baseUrl, req_options);
-			if (!response.ok) {
-				throw new Error(`HTTP error! status: ${response.status}`);
-			}
-			if (!response.body) {
-				throw new Error('Response body is null');
-			}
-
-			const reader = response.body.getReader();
-			const decoder = new TextDecoder();
-
-			return {
-				[Symbol.asyncIterator]: async function* () {
-					try {
-						while (true) {
-							const { done, value } = await reader.read();
-							if (done) break;
-
-							const chunk = decoder.decode(value);
-							const lines = chunk.split('\n').filter(line => line.trim());
-
-							for (const line of lines) {
-								if (line.startsWith('data: ')) {
-									const jsonData = JSON.parse(line.slice(6)) as ChatCompletionChunk;
-									if (!jsonData || typeof jsonData !== 'object' || !('choices' in jsonData)) {
-										throw new Error('Invalid chunk format received');
-									}
-									yield InfioProvider.parseStreamingResponseChunk(jsonData);
-								}
-							}
-						}
-					} finally {
-						reader.releaseLock();
-					}
-				}
-			};
-		} catch (error) {
-			if (error instanceof OpenAI.AuthenticationError) {
-				throw new LLMAPIKeyInvalidException(
-					'OpenAI API key is invalid. Please update it in settings menu.',
-				)
-			}
-			throw error
-		}
-	}
-
-	static parseNonStreamingResponse(
-		response: ChatCompletion,
-	): LLMResponseNonStreaming {
-		return {
-			id: response.id,
-			choices: response.choices.map((choice) => ({
-				finish_reason: choice.finish_reason,
-				message: {
-					content: choice.message.content,
-					role: choice.message.role,
-				},
-			})),
-			created: response.created,
-			model: response.model,
-			object: 'chat.completion',
-			system_fingerprint: response.system_fingerprint,
-			usage: response.usage,
-		}
-	}
-
-	static parseStreamingResponseChunk(
-		chunk: ChatCompletionChunk,
-	): LLMResponseStreaming {
-		return {
-			id: chunk.id,
-			choices: chunk.choices.map((choice) => ({
-				finish_reason: choice.finish_reason ?? null,
-				delta: {
-					content: choice.delta.content ?? null,
-					role: choice.delta.role,
-				},
-			})),
-			created: chunk.created,
-			model: chunk.model,
-			object: 'chat.completion.chunk',
-			system_fingerprint: chunk.system_fingerprint,
-			usage: chunk.usage ?? undefined,
-		}
-	}
-}
--- a/src/core/llm/manager.ts
+++ b/src/core/llm/manager.ts
@ -1,4 +1,4 @@
-import { ALIBABA_QWEN_BASE_URL, DEEPSEEK_BASE_URL, GROK_BASE_URL, OPENROUTER_BASE_URL, SILICONFLOW_BASE_URL } from '../../constants'
+import { ALIBABA_QWEN_BASE_URL, DEEPSEEK_BASE_URL, GROK_BASE_URL, INFIO_BASE_URL, OPENROUTER_BASE_URL, SILICONFLOW_BASE_URL } from '../../constants'
 import { ApiProvider, LLMModel } from '../../types/llm/model'
 import {
 	LLMOptions,
@ -14,7 +14,6 @@ import { InfioSettings } from '../../types/settings'
 import { AnthropicProvider } from './anthropic'
 import { GeminiProvider } from './gemini'
 import { GroqProvider } from './groq'
-import { InfioProvider } from './infio'
 import { OllamaProvider } from './ollama'
 import { OpenAIAuthenticatedProvider } from './openai'
 import { OpenAICompatibleProvider } from './openai-compatible'
@ -40,7 +39,7 @@ class LLMManager implements LLMManagerInterface {
 	private googleProvider: GeminiProvider
 	private groqProvider: GroqProvider
 	private grokProvider: OpenAICompatibleProvider
-	private infioProvider: InfioProvider
+	private infioProvider: OpenAICompatibleProvider
 	private openrouterProvider: OpenAICompatibleProvider
 	private siliconflowProvider: OpenAICompatibleProvider
 	private alibabaQwenProvider: OpenAICompatibleProvider
@ -49,7 +48,10 @@ class LLMManager implements LLMManagerInterface {
 	private isInfioEnabled: boolean

 	constructor(settings: InfioSettings) {
-		this.infioProvider = new InfioProvider(settings.infioProvider.apiKey)
+		this.infioProvider = new OpenAICompatibleProvider(
+			settings.infioProvider.apiKey,
+			INFIO_BASE_URL
+		)
 		this.openrouterProvider = new OpenAICompatibleProvider(
 			settings.openrouterProvider.apiKey,
 			settings.openrouterProvider.baseUrl && settings.openrouterProvider.useCustomUrl ?
@ -93,14 +95,14 @@ class LLMManager implements LLMManagerInterface {
 		request: LLMRequestNonStreaming,
 		options?: LLMOptions,
 	): Promise<LLMResponseNonStreaming> {
-		if (this.isInfioEnabled) {
-			return await this.infioProvider.generateResponse(
-				model,
-				request,
-			)
-		}
-		// use custom provider
+		console.log("model", model)
 		switch (model.provider) {
+			case ApiProvider.Infio:
+				return await this.infioProvider.generateResponse(
+					model,
+					request,
+					options,
+				)
 			case ApiProvider.OpenRouter:
 				return await this.openrouterProvider.generateResponse(
 					model,
@ -169,11 +171,9 @@ class LLMManager implements LLMManagerInterface {
 		request: LLMRequestStreaming,
 		options?: LLMOptions,
 	): Promise<AsyncIterable<LLMResponseStreaming>> {
-		if (this.isInfioEnabled) {
-			return await this.infioProvider.streamResponse(model, request)
-		}
-		// use custom provider
 		switch (model.provider) {
+			case ApiProvider.Infio:
+				return await this.infioProvider.streamResponse(model, request, options)
 			case ApiProvider.OpenRouter:
 				return await this.openrouterProvider.streamResponse(model, request, options)
 			case ApiProvider.SiliconFlow:
--- a/src/settings/components/ModelProviderSettings.tsx
+++ b/src/settings/components/ModelProviderSettings.tsx
@ -171,6 +171,7 @@ const CustomProviderSettings: React.FC<CustomProviderSettingsProps> = ({ plugin,
 			<div className="infio-llm-setting-divider"></div>
 			<ComboBoxComponent
 				name={t("settings.Models.chatModel")}
+				settings={settings}
 				provider={settings.chatModelProvider || currProvider}
 				modelId={settings.chatModelId}
 				updateModel={updateChatModelId}
@ -178,6 +179,7 @@ const CustomProviderSettings: React.FC<CustomProviderSettingsProps> = ({ plugin,
 			<div className="infio-llm-setting-divider"></div>
 			<ComboBoxComponent
 				name={t("settings.Models.autocompleteModel")}
+				settings={settings}
 				provider={settings.applyModelProvider || currProvider}
 				modelId={settings.applyModelId}
 				updateModel={updateApplyModelId}
@ -185,6 +187,7 @@ const CustomProviderSettings: React.FC<CustomProviderSettingsProps> = ({ plugin,
 			<div className="infio-llm-setting-divider"></div>
 			<ComboBoxComponent
 				name={t("settings.Models.embeddingModel")}
+				settings={settings}
 				provider={settings.embeddingModelProvider || ApiProvider.Google}
 				modelId={settings.embeddingModelId}
 				isEmbedding={true}
--- a/src/settings/components/ProviderModelsPicker.tsx
+++ b/src/settings/components/ProviderModelsPicker.tsx
@ -3,6 +3,7 @@ import Fuse, { FuseResult } from "fuse.js";
 import React, { useEffect, useMemo, useRef, useState } from "react";

 import { ApiProvider } from "../../types/llm/model";
+import { InfioSettings } from "../../types/settings";
 // import { PROVIDERS } from '../constants';
 import { GetAllProviders, GetEmbeddingProviderModelIds, GetEmbeddingProviders, GetProviderModelIds } from "../../utils/api";

@ -149,6 +150,7 @@ export type ComboBoxComponentProps = {
 	name: string;
 	provider: ApiProvider;
 	modelId: string;
+	settings?: InfioSettings | null;
 	isEmbedding?: boolean,
 	updateModel: (provider: ApiProvider, modelId: string) => void;
 };
@ -157,6 +159,7 @@ export const ComboBoxComponent: React.FC<ComboBoxComponentProps> = ({
 	name,
 	provider,
 	modelId,
+	settings = null,
 	isEmbedding = false,
 	updateModel,
 }) => {
@ -177,7 +180,7 @@ export const ComboBoxComponent: React.FC<ComboBoxComponentProps> = ({
 		const fetchModelIds = async () => {
 			const ids = isEmbedding
 				? GetEmbeddingProviderModelIds(modelProvider)
-				: await GetProviderModelIds(modelProvider);
+				: await GetProviderModelIds(modelProvider, settings);
 			setModelIds(ids);
 		};

--- a/src/utils/api.ts
+++ b/src/utils/api.ts
@ -1,5 +1,6 @@
-import { OPENROUTER_BASE_URL } from '../constants'
+import { INFIO_BASE_URL, OPENROUTER_BASE_URL } from '../constants'
 import { ApiProvider } from '../types/llm/model'
+import { InfioSettings } from '../types/settings'

 export interface ModelInfo {
 	maxTokens?: number
@ -14,6 +15,15 @@ export interface ModelInfo {
 	description?: string
 	reasoningEffort?: string,
 	thinking?: boolean
+	maxThinkingTokens?: number
+	supportsReasoningBudget?: boolean
+	requiredReasoningBudget?: boolean
+	tiers?: readonly {
+		readonly contextWindow: number,
+		readonly inputPrice: number,
+		readonly outputPrice: number,
+		readonly cacheReadsPrice: number,
+	}[]
 }

 export interface EmbeddingModelInfo {
@ -21,89 +31,13 @@ export interface EmbeddingModelInfo {
 	description?: string
 }

-// Infio
-// https://infio.app/pricing
-export type InfioModelId = keyof typeof infioModels
-export const infioDefaultModelId: InfioModelId = "deepseek-chat"
-export const infioModels = {
-	"deepseek-chat": {
-		maxTokens: 8_000,
-		contextWindow: 64_000,
-		supportsImages: false,
-		supportsPromptCache: true, // supports context caching, but not in the way anthropic does it (deepseek reports input tokens and reads/writes in the same usage report) FIXME: we need to show users cache stats how deepseek does it
-		inputPrice: 0, // technically there is no input price, it's all either a cache hit or miss (ApiOptions will not show this)
-		outputPrice: 0.28,
-		cacheWritesPrice: 0.14,
-		cacheReadsPrice: 0.014,
-	},
-	"deepseek-reasoner": {
-		maxTokens: 8_000,
-		contextWindow: 64_000,
-		supportsImages: false,
-		supportsPromptCache: true, // supports context caching, but not in the way anthropic does it (deepseek reports input tokens and reads/writes in the same usage report) FIXME: we need to show users cache stats how deepseek does it
-		inputPrice: 0, // technically there is no input price, it's all either a cache hit or miss (ApiOptions will not show this)
-		outputPrice: 2.19,
-		cacheWritesPrice: 0.55,
-		cacheReadsPrice: 0.14,
-	},
-	"o3-mini": {
-		maxTokens: 100_000,
-		contextWindow: 200_000,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 1.1,
-		outputPrice: 4.4,
-	},
-	// don't support tool use yet
-	o1: {
-		maxTokens: 100_000,
-		contextWindow: 200_000,
-		supportsImages: true,
-		supportsPromptCache: false,
-		inputPrice: 15,
-		outputPrice: 60,
-	},
-	"o1-preview": {
-		maxTokens: 32_768,
-		contextWindow: 128_000,
-		supportsImages: true,
-		supportsPromptCache: false,
-		inputPrice: 15,
-		outputPrice: 60,
-	},
-	"o1-mini": {
-		maxTokens: 65_536,
-		contextWindow: 128_000,
-		supportsImages: true,
-		supportsPromptCache: false,
-		inputPrice: 1.1,
-		outputPrice: 4.4,
-	},
-	"gpt-4o": {
-		maxTokens: 4_096,
-		contextWindow: 128_000,
-		supportsImages: true,
-		supportsPromptCache: false,
-		inputPrice: 2.5,
-		outputPrice: 10,
-	},
-	"gpt-4o-mini": {
-		maxTokens: 16_384,
-		contextWindow: 128_000,
-		supportsImages: true,
-		supportsPromptCache: false,
-		inputPrice: 0.15,
-		outputPrice: 0.6,
-	},
-} as const satisfies Record<string, ModelInfo>
-
 // Anthropic
 // https://docs.anthropic.com/en/docs/about-claude/models
 export type AnthropicModelId = keyof typeof anthropicModels
-export const anthropicDefaultModelId: AnthropicModelId = "claude-3-7-sonnet-20250219"
+export const anthropicDefaultModelId: AnthropicModelId = "claude-sonnet-4-20250514"
 export const anthropicModels = {
-	"claude-3-7-sonnet-20250219:thinking": {
-		maxTokens: 128_000,
+	"claude-sonnet-4-20250514": {
+		maxTokens: 64_000, // Overridden to 8k if `enableReasoningEffort` is false.
 		contextWindow: 200_000,
 		supportsImages: true,
 		supportsComputerUse: true,
@ -112,10 +46,35 @@ export const anthropicModels = {
 		outputPrice: 15.0, // $15 per million output tokens
 		cacheWritesPrice: 3.75, // $3.75 per million tokens
 		cacheReadsPrice: 0.3, // $0.30 per million tokens
-		thinking: true,
+		supportsReasoningBudget: true,
+	},
+	"claude-opus-4-20250514": {
+		maxTokens: 32_000, // Overridden to 8k if `enableReasoningEffort` is false.
+		contextWindow: 200_000,
+		supportsImages: true,
+		supportsComputerUse: true,
+		supportsPromptCache: true,
+		inputPrice: 15.0, // $15 per million input tokens
+		outputPrice: 75.0, // $75 per million output tokens
+		cacheWritesPrice: 18.75, // $18.75 per million tokens
+		cacheReadsPrice: 1.5, // $1.50 per million tokens
+		supportsReasoningBudget: true,
+	},
+	"claude-3-7-sonnet-20250219:thinking": {
+		maxTokens: 128_000, // Unlocked by passing `beta` flag to the model. Otherwise, it's 64k.
+		contextWindow: 200_000,
+		supportsImages: true,
+		supportsComputerUse: true,
+		supportsPromptCache: true,
+		inputPrice: 3.0, // $3 per million input tokens
+		outputPrice: 15.0, // $15 per million output tokens
+		cacheWritesPrice: 3.75, // $3.75 per million tokens
+		cacheReadsPrice: 0.3, // $0.30 per million tokens
+		supportsReasoningBudget: true,
+		requiredReasoningBudget: true,
 	},
 	"claude-3-7-sonnet-20250219": {
-		maxTokens: 8192,
+		maxTokens: 8192, // Since we already have a `:thinking` virtual model we aren't setting `supportsReasoningBudget: true` here.
 		contextWindow: 200_000,
 		supportsImages: true,
 		supportsComputerUse: true,
@ -124,7 +83,6 @@ export const anthropicModels = {
 		outputPrice: 15.0, // $15 per million output tokens
 		cacheWritesPrice: 3.75, // $3.75 per million tokens
 		cacheReadsPrice: 0.3, // $0.30 per million tokens
-		thinking: false,
 	},
 	"claude-3-5-sonnet-20241022": {
 		maxTokens: 8192,
@ -167,10 +125,71 @@ export const anthropicModels = {
 		cacheWritesPrice: 0.3,
 		cacheReadsPrice: 0.03,
 	},
-} as const satisfies Record<string, ModelInfo>
+} as const satisfies Record<string, ModelInfo> // as const assertion makes the object 
+
+// Infio
+export const infioDefaultModelId = "deepseek/deepseek-v3" // will always exist 
+export const infioDefaultModelInfo: ModelInfo = {
+	maxTokens: 8192,
+	contextWindow: 65_536,
+	supportsImages: false,
+	supportsComputerUse: true,
+	supportsPromptCache: true,
+	inputPrice: 0.272,
+	outputPrice: 1.088,
+	cacheWritesPrice: 0.14,
+	cacheReadsPrice: 0.014,
+}
+let infioModelsCache: Record<string, ModelInfo> | null = null;
+
+async function fetchInfioModels(apiKey?: string): Promise<Record<string, ModelInfo>> {
+	if (infioModelsCache) {
+		return infioModelsCache;
+	}
+
+	try {
+		const headers: Record<string, string> = {
+			'Content-Type': 'application/json'
+		};
+
+		// 添加Authorization请求头，使用Bearer格式，如果有API密钥的话
+		if (apiKey) {
+			headers['Authorization'] = `Bearer ${apiKey}`;
+		}
+
+		const response = await fetch(INFIO_BASE_URL + "/model_group/info", {
+			method: 'GET',
+			headers: headers
+		});
+		const data = await response.json();
+		const models: Record<string, ModelInfo> = {};
+		if (data?.data) {
+			for (const model of data.data) {
+				models[model.model_group] = {
+					maxTokens: model.max_output_tokens,
+					contextWindow: model.max_input_tokens,
+					supportsImages: false,
+					supportsPromptCache: false,
+					inputPrice: model.input_cost_per_token ? model.input_cost_per_token * 1000000 : 0,
+					outputPrice: model.output_cost_per_token ? model.output_cost_per_token * 1000000 : 0,
+				};
+			}
+		}
+
+		infioModelsCache = models;
+		return models;
+	} catch (error) {
+		console.error('Failed to fetch Infio models:', error);
+		// 如果出错，返回默认模型
+		return {
+			[infioDefaultModelId]: infioDefaultModelInfo
+		};
+	}
+}
+
 // OpenRouter
 // https://openrouter.ai/models?order=newest&supported_parameters=tools
-export const openRouterDefaultModelId = "anthropic/claude-3.5-sonnet" // will always exist in openRouterModels
+export const openRouterDefaultModelId = "anthropic/claude-sonnet-4" // will always exist in openRouterModels
 export const openRouterDefaultModelInfo: ModelInfo = {
 	maxTokens: 8192,
 	contextWindow: 200_000,
@ -222,8 +241,31 @@ async function fetchOpenRouterModels(): Promise<Record<string, ModelInfo>> {
 // Gemini
 // https://ai.google.dev/gemini-api/docs/models/gemini
 export type GeminiModelId = keyof typeof geminiModels
-export const geminiDefaultModelId: GeminiModelId = "gemini-2.5-flash-preview-04-17"
+export const geminiDefaultModelId: GeminiModelId = "gemini-2.5-flash-preview-05-20"
 export const geminiModels = {
+	"gemini-2.5-flash-preview-05-20:thinking": {
+		maxTokens: 65_535,
+		contextWindow: 1_048_576,
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 0.15,
+		outputPrice: 3.5,
+		cacheReadsPrice: 0.0375,
+		cacheWritesPrice: 1.0,
+		maxThinkingTokens: 24_576,
+		supportsReasoningBudget: true,
+		requiredReasoningBudget: true,
+	},
+	"gemini-2.5-flash-preview-05-20": {
+		maxTokens: 65_535,
+		contextWindow: 1_048_576,
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 0.15,
+		outputPrice: 0.6,
+		cacheReadsPrice: 0.0375,
+		cacheWritesPrice: 1.0,
+	},
 	"gemini-2.5-flash-preview-04-17:thinking": {
 		maxTokens: 65_535,
 		contextWindow: 1_048_576,
@ -231,8 +273,9 @@ export const geminiModels = {
 		supportsPromptCache: false,
 		inputPrice: 0.15,
 		outputPrice: 3.5,
-		thinking: true,
-		// maxThinkingTokens: 24_576,
+		maxThinkingTokens: 24_576,
+		supportsReasoningBudget: true,
+		requiredReasoningBudget: true,
 	},
 	"gemini-2.5-flash-preview-04-17": {
 		maxTokens: 65_535,
@ -241,7 +284,6 @@ export const geminiModels = {
 		supportsPromptCache: false,
 		inputPrice: 0.15,
 		outputPrice: 0.6,
-		thinking: false,
 	},
 	"gemini-2.5-pro-exp-03-25": {
 		maxTokens: 65_535,
@ -260,6 +302,44 @@ export const geminiModels = {
 		outputPrice: 15,
 		cacheReadsPrice: 0.625,
 		cacheWritesPrice: 4.5,
+		tiers: [
+			{
+				contextWindow: 200_000,
+				inputPrice: 1.25,
+				outputPrice: 10,
+				cacheReadsPrice: 0.31,
+			},
+			{
+				contextWindow: Infinity,
+				inputPrice: 2.5,
+				outputPrice: 15,
+				cacheReadsPrice: 0.625,
+			},
+		],
+	},
+	"gemini-2.5-pro-preview-05-06": {
+		maxTokens: 65_535,
+		contextWindow: 1_048_576,
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 2.5, // This is the pricing for prompts above 200k tokens.
+		outputPrice: 15,
+		cacheReadsPrice: 0.625,
+		cacheWritesPrice: 4.5,
+		tiers: [
+			{
+				contextWindow: 200_000,
+				inputPrice: 1.25,
+				outputPrice: 10,
+				cacheReadsPrice: 0.31,
+			},
+			{
+				contextWindow: Infinity,
+				inputPrice: 2.5,
+				outputPrice: 15,
+				cacheReadsPrice: 0.625,
+			},
+		],
 	},
 	"gemini-2.0-flash-001": {
 		maxTokens: 8192,
@ -315,9 +395,25 @@ export const geminiModels = {
 		maxTokens: 8192,
 		contextWindow: 1_048_576,
 		supportsImages: true,
-		supportsPromptCache: false,
-		inputPrice: 0,
-		outputPrice: 0,
+		supportsPromptCache: true,
+		inputPrice: 0.15, // This is the pricing for prompts above 128k tokens.
+		outputPrice: 0.6,
+		cacheReadsPrice: 0.0375,
+		cacheWritesPrice: 1.0,
+		tiers: [
+			{
+				contextWindow: 128_000,
+				inputPrice: 0.075,
+				outputPrice: 0.3,
+				cacheReadsPrice: 0.01875,
+			},
+			{
+				contextWindow: Infinity,
+				inputPrice: 0.15,
+				outputPrice: 0.6,
+				cacheReadsPrice: 0.0375,
+			},
+		],
 	},
 	"gemini-1.5-flash-exp-0827": {
 		maxTokens: 8192,
@ -360,6 +456,7 @@ export const geminiModels = {
 		outputPrice: 0,
 	},
 } as const satisfies Record<string, ModelInfo>
+
 export const geminiEmbeddingModels = {
 	"text-embedding-004": {
 		dimensions: 768,
@ -482,8 +579,8 @@ export const deepSeekModels = {
 		contextWindow: 64_000,
 		supportsImages: false,
 		supportsPromptCache: true, // supports context caching, but not in the way anthropic does it (deepseek reports input tokens and reads/writes in the same usage report) FIXME: we need to show users cache stats how deepseek does it
-		inputPrice: 0, // technically there is no input price, it's all either a cache hit or miss (ApiOptions will not show this)
-		outputPrice: 0.28,
+		inputPrice: 0.272, // technically there is no input price, it's all either a cache hit or miss (ApiOptions will not show this)
+		outputPrice: 1.088,
 		cacheWritesPrice: 0.14,
 		cacheReadsPrice: 0.014,
 	},
@ -1505,10 +1602,46 @@ export const GetEmbeddingProviders = (): ApiProvider[] => {
 }

 // Get all models for a provider
-export const GetProviderModels = async (provider: ApiProvider): Promise<Record<string, ModelInfo>> => {
+export const GetProviderModels = async (provider: ApiProvider, settings?: InfioSettings): Promise<Record<string, ModelInfo>> => {
 	switch (provider) {
-		case ApiProvider.Infio:
-			return infioModels
+		case ApiProvider.Infio: {
+			const apiKey = settings?.infioProvider?.apiKey
+			return await fetchInfioModels(apiKey)
+		}
+		case ApiProvider.OpenRouter:
+			return await fetchOpenRouterModels()
+		case ApiProvider.OpenAI:
+			return openAiNativeModels
+		case ApiProvider.AlibabaQwen:
+			return qwenModels
+		case ApiProvider.SiliconFlow:
+			return siliconFlowModels
+		case ApiProvider.Anthropic:
+			return anthropicModels
+		case ApiProvider.Deepseek:
+			return deepSeekModels
+		case ApiProvider.Google:
+			return geminiModels
+		case ApiProvider.Groq:
+			return groqModels
+		case ApiProvider.Grok:
+			return grokModels
+		case ApiProvider.Ollama:
+			return {}
+		case ApiProvider.OpenAICompatible:
+			return {}
+		default:
+			return {}
+	}
+}
+
+// Get all models for a provider with settings (needed for providers that require API keys)
+export const GetProviderModelsWithSettings = async (provider: ApiProvider, settings?: InfioSettings): Promise<Record<string, ModelInfo>> => {
+	switch (provider) {
+		case ApiProvider.Infio: {
+			const apiKey = settings?.infioProvider?.apiKey
+			return await fetchInfioModels(apiKey)
+		}
 		case ApiProvider.OpenRouter:
 			return await fetchOpenRouterModels()
 		case ApiProvider.OpenAI:
@ -1537,8 +1670,8 @@ export const GetProviderModels = async (provider: ApiProvider): Promise<Record<s
 }

 // Get all model ids for a provider
-export const GetProviderModelIds = async (provider: ApiProvider): Promise<string[]> => {
-	const models = await GetProviderModels(provider)
+export const GetProviderModelIds = async (provider: ApiProvider, settings?: InfioSettings): Promise<string[]> => {
+	const models = await GetProviderModels(provider, settings)
 	return Object.keys(models)
 }