feat: add HuggingFace provider support

packages/types/src/provider-settings.ts

@@ -32,6 +32,7 @@ export const providerNames = [
 	"groq",
 	"chutes",
 	"litellm",
+	"huggingface",
 ] as const
 
 export const providerNamesSchema = z.enum(providerNames)
@@ -219,6 +220,12 @@ const groqSchema = apiModelIdProviderModelSchema.extend({
 	groqApiKey: z.string().optional(),
 })
 
+const huggingFaceSchema = baseProviderSettingsSchema.extend({
+	huggingFaceApiKey: z.string().optional(),
+	huggingFaceModelId: z.string().optional(),
+	huggingFaceInferenceProvider: z.string().optional(),
+})
+
 const chutesSchema = apiModelIdProviderModelSchema.extend({
 	chutesApiKey: z.string().optional(),
 })
@@ -256,6 +263,7 @@ export const providerSettingsSchemaDiscriminated = z.discriminatedUnion("apiProv
 	fakeAiSchema.merge(z.object({ apiProvider: z.literal("fake-ai") })),
 	xaiSchema.merge(z.object({ apiProvider: z.literal("xai") })),
 	groqSchema.merge(z.object({ apiProvider: z.literal("groq") })),
+	huggingFaceSchema.merge(z.object({ apiProvider: z.literal("huggingface") })),
 	chutesSchema.merge(z.object({ apiProvider: z.literal("chutes") })),
 	litellmSchema.merge(z.object({ apiProvider: z.literal("litellm") })),
 	defaultSchema,
@@ -285,6 +293,7 @@ export const providerSettingsSchema = z.object({
 	...fakeAiSchema.shape,
 	...xaiSchema.shape,
 	...groqSchema.shape,
+	...huggingFaceSchema.shape,
 	...chutesSchema.shape,
 	...litellmSchema.shape,
 	...codebaseIndexProviderSchema.shape,
@@ -304,6 +313,7 @@ export const MODEL_ID_KEYS: Partial<keyof ProviderSettings>[] = [
 	"unboundModelId",
 	"requestyModelId",
 	"litellmModelId",
+	"huggingFaceModelId",
 ]
 
 export const getModelId = (settings: ProviderSettings): string | undefined => {

packages/types/src/providers/huggingface.ts

@@ -0,0 +1,61 @@
+import { z } from "zod"
+import { modelInfoSchema } from "../model.js"
+
+export const huggingFaceDefaultModelId = "meta-llama/Llama-3.3-70B-Instruct"
+
+export const huggingFaceModels = {
+	"meta-llama/Llama-3.3-70B-Instruct": {
+		maxTokens: 8192,
+		contextWindow: 131072,
+		supportsImages: false,
+		supportsPromptCache: false,
+	},
+	"meta-llama/Llama-3.2-11B-Vision-Instruct": {
+		maxTokens: 4096,
+		contextWindow: 131072,
+		supportsImages: true,
+		supportsPromptCache: false,
+	},
+	"Qwen/Qwen2.5-72B-Instruct": {
+		maxTokens: 8192,
+		contextWindow: 131072,
+		supportsImages: false,
+		supportsPromptCache: false,
+	},
+	"mistralai/Mistral-7B-Instruct-v0.3": {
+		maxTokens: 8192,
+		contextWindow: 32768,
+		supportsImages: false,
+		supportsPromptCache: false,
+	},
+} as const
+
+export type HuggingFaceModelId = keyof typeof huggingFaceModels
+
+export const huggingFaceModelSchema = z.enum(
+	Object.keys(huggingFaceModels) as [HuggingFaceModelId, ...HuggingFaceModelId[]],
+)
+
+export const huggingFaceModelInfoSchema = z
+	.discriminatedUnion("id", [
+		z.object({
+			id: z.literal("meta-llama/Llama-3.3-70B-Instruct"),
+			info: modelInfoSchema.optional(),
+		}),
+		z.object({
+			id: z.literal("meta-llama/Llama-3.2-11B-Vision-Instruct"),
+			info: modelInfoSchema.optional(),
+		}),
+		z.object({
+			id: z.literal("Qwen/Qwen2.5-72B-Instruct"),
+			info: modelInfoSchema.optional(),
+		}),
+		z.object({
+			id: z.literal("mistralai/Mistral-7B-Instruct-v0.3"),
+			info: modelInfoSchema.optional(),
+		}),
+	])
+	.transform(({ id, info }) => ({
+		id,
+		info: { ...huggingFaceModels[id], ...info },
+	}))

packages/types/src/providers/index.ts

@@ -6,6 +6,7 @@ export * from "./deepseek.js"
 export * from "./gemini.js"
 export * from "./glama.js"
 export * from "./groq.js"
+export * from "./huggingface.js"
 export * from "./lite-llm.js"
 export * from "./lm-studio.js"
 export * from "./mistral.js"

src/api/huggingface-models.ts

@@ -0,0 +1,17 @@
+import { fetchHuggingFaceModels, type HuggingFaceModel } from "../services/huggingface-models"
+
+export interface HuggingFaceModelsResponse {
+	models: HuggingFaceModel[]
+	cached: boolean
+	timestamp: number
+}
+
+export async function getHuggingFaceModels(): Promise<HuggingFaceModelsResponse> {
+	const models = await fetchHuggingFaceModels()
+
+	return {
+		models,
+		cached: false, // We could enhance this to track if data came from cache
+		timestamp: Date.now(),
+	}
+}

src/api/index.ts

@@ -26,6 +26,7 @@ import {
 	FakeAIHandler,
 	XAIHandler,
 	GroqHandler,
+	HuggingFaceHandler,
 	ChutesHandler,
 	LiteLLMHandler,
 	ClaudeCodeHandler,
@@ -108,6 +109,8 @@ export function buildApiHandler(configuration: ProviderSettings): ApiHandler {
 			return new XAIHandler(options)
 		case "groq":
 			return new GroqHandler(options)
+		case "huggingface":
+			return new HuggingFaceHandler(options)
 		case "chutes":
 			return new ChutesHandler(options)
 		case "litellm":

src/api/providers/huggingface.ts

@@ -0,0 +1,99 @@
+import OpenAI from "openai"
+import { Anthropic } from "@anthropic-ai/sdk"
+
+import type { ApiHandlerOptions } from "../../shared/api"
+import { ApiStream } from "../transform/stream"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import type { SingleCompletionHandler, ApiHandlerCreateMessageMetadata } from "../index"
+import { DEFAULT_HEADERS } from "./constants"
+import { BaseProvider } from "./base-provider"
+
+export class HuggingFaceHandler extends BaseProvider implements SingleCompletionHandler {
+	private client: OpenAI
+	private options: ApiHandlerOptions
+
+	constructor(options: ApiHandlerOptions) {
+		super()
+		this.options = options
+
+		if (!this.options.huggingFaceApiKey) {
+			throw new Error("Hugging Face API key is required")
+		}
+
+		this.client = new OpenAI({
+			baseURL: "https://router.huggingface.co/v1",
+			apiKey: this.options.huggingFaceApiKey,
+			defaultHeaders: DEFAULT_HEADERS,
+		})
+	}
+
+	override async *createMessage(
+		systemPrompt: string,
+		messages: Anthropic.Messages.MessageParam[],
+		metadata?: ApiHandlerCreateMessageMetadata,
+	): ApiStream {
+		const modelId = this.options.huggingFaceModelId || "meta-llama/Llama-3.3-70B-Instruct"
+		const temperature = this.options.modelTemperature ?? 0.7
+
+		const params: OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming = {
+			model: modelId,
+			temperature,
+			messages: [{ role: "system", content: systemPrompt }, ...convertToOpenAiMessages(messages)],
+			stream: true,
+			stream_options: { include_usage: true },
+		}
+
+		const stream = await this.client.chat.completions.create(params)
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices[0]?.delta
+
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+				}
+			}
+		}
+	}
+
+	async completePrompt(prompt: string): Promise<string> {
+		const modelId = this.options.huggingFaceModelId || "meta-llama/Llama-3.3-70B-Instruct"
+
+		try {
+			const response = await this.client.chat.completions.create({
+				model: modelId,
+				messages: [{ role: "user", content: prompt }],
+			})
+
+			return response.choices[0]?.message.content || ""
+		} catch (error) {
+			if (error instanceof Error) {
+				throw new Error(`Hugging Face completion error: ${error.message}`)
+			}
+
+			throw error
+		}
+	}
+
+	override getModel() {
+		const modelId = this.options.huggingFaceModelId || "meta-llama/Llama-3.3-70B-Instruct"
+		return {
+			id: modelId,
+			info: {
+				maxTokens: 8192,
+				contextWindow: 131072,
+				supportsImages: false,
+				supportsPromptCache: false,
+			},
+		}
+	}
+}

src/api/providers/index.ts

@@ -9,6 +9,7 @@ export { FakeAIHandler } from "./fake-ai"
 export { GeminiHandler } from "./gemini"
 export { GlamaHandler } from "./glama"
 export { GroqHandler } from "./groq"
+export { HuggingFaceHandler } from "./huggingface"
 export { HumanRelayHandler } from "./human-relay"
 export { LiteLLMHandler } from "./lite-llm"
 export { LmStudioHandler } from "./lm-studio"

src/core/webview/webviewMessageHandler.ts

@@ -674,6 +674,22 @@ export const webviewMessageHandler = async (
 			// TODO: Cache like we do for OpenRouter, etc?
 			provider.postMessageToWebview({ type: "vsCodeLmModels", vsCodeLmModels })
 			break
+		case "requestHuggingFaceModels":
+			try {
+				const { getHuggingFaceModels } = await import("../../api/huggingface-models")
+				const huggingFaceModelsResponse = await getHuggingFaceModels()
+				provider.postMessageToWebview({
+					type: "huggingFaceModels",
+					huggingFaceModels: huggingFaceModelsResponse.models,
+				})
+			} catch (error) {
+				console.error("Failed to fetch Hugging Face models:", error)
+				provider.postMessageToWebview({
+					type: "huggingFaceModels",
+					huggingFaceModels: [],
+				})
+			}
+			break
 		case "openImage":
 			openImage(message.text!, { values: message.values })
 			break