Merge branch 'main' into feature/assistants

Author: nsarrazin

.env

@@ -121,6 +121,7 @@ PUBLIC_APP_DATA_SHARING=#set to 1 to enable options & text regarding data sharin
 PUBLIC_APP_DISCLAIMER=#set to 1 to show a disclaimer on login page
 LLM_SUMMERIZATION=true
 
+EXPOSE_API=true
 # PUBLIC_APP_NAME=HuggingChat
 # PUBLIC_APP_ASSETS=huggingchat
 # PUBLIC_APP_COLOR=yellow

.env.template

@@ -60,6 +60,33 @@ MODELS=`[
         "max_new_tokens": 1024
       }
     },
+    {
+      "name" : "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
+      "description" : "Nous Hermes 2 Mixtral 8x7B DPO is the new flagship Nous Research model trained over the Mixtral 8x7B MoE LLM.",
+      "websiteUrl" : "https://nousresearch.com/",
+      "chatPromptTemplate" : "<|im_start|>system\n{{#if @root.preprompt}}{{@root.preprompt}}<|im_end|>\n{{/if}}{{#each messages}}{{#ifUser}}<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n{{/ifUser}}{{#ifAssistant}}{{content}}<|im_end|>\n{{/ifAssistant}}{{/each}}",
+      "promptExamples": [
+        {
+          "title": "Write an email from bullet list",
+          "prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
+        }, {
+          "title": "Code a snake game",
+          "prompt": "Code a basic snake game in python, give explanations for each step."
+        }, {
+          "title": "Assist in a task",
+          "prompt": "How do I make a delicious lemon cheesecake?"
+        }
+      ],
+      "parameters": {
+        "temperature": 0.7,
+        "top_p": 0.95,
+        "repetition_penalty": 1,
+        "top_k": 50,
+        "truncate": 24576,
+        "max_new_tokens": 2048,
+        "stop": ["<|im_end|>"]
+      }
+    },
     {
       "name": "codellama/CodeLlama-34b-Instruct-hf",
       "displayName": "codellama/CodeLlama-34b-Instruct-hf",
@@ -226,4 +253,5 @@ PUBLIC_GOOGLE_ANALYTICS_ID=G-8Q63TH4CSL
 # ADDRESS_HEADER=X-Forwarded-For
 # XFF_DEPTH=2
 
-ENABLE_ASSISTANTS=true
+ENABLE_ASSISTANTS=true
+EXPOSE_API=false

PROMPTS.md

@@ -49,3 +49,9 @@ System: {{preprompt}}\nUser:{{#each messages}}{{#ifUser}}{{content}}\nFalcon:{{/
 ```env
 <s> {{#each messages}}{{#ifUser}}[INST]{{#if @first}}{{#if @root.preprompt}}{{@root.preprompt}}\n{{/if}}{{/if}} {{content}} [/INST]{{/ifUser}}{{#ifAssistant}} {{content}}</s> {{/ifAssistant}}{{/each}}
 ```
+
+## ChatML
+
+```env
+<|im_start|>system\n{{#if @root.preprompt}}{{@root.preprompt}}<|im_end|>\n{{/if}}{{#each messages}}{{#ifUser}}<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n{{/ifUser}}{{#ifAssistant}}{{content}}<|im_end|>\n{{/ifAssistant}}{{/each}}
+```

src/hooks.server.ts

@@ -1,4 +1,4 @@
-import { COOKIE_NAME, MESSAGES_BEFORE_LOGIN } from "$env/static/private";
+import { COOKIE_NAME, EXPOSE_API, MESSAGES_BEFORE_LOGIN } from "$env/static/private";
 import type { Handle } from "@sveltejs/kit";
 import {
 	PUBLIC_GOOGLE_ANALYTICS_ID,
@@ -13,6 +13,10 @@ import { sha256 } from "$lib/utils/sha256";
 import { addWeeks } from "date-fns";
 
 export const handle: Handle = async ({ event, resolve }) => {
+	if (event.url.pathname.startsWith(`${base}/api/`) && EXPOSE_API !== "true") {
+		return new Response("API is disabled", { status: 403 });
+	}
+
 	function errorResponse(status: number, message: string) {
 		const sendJson =
 			event.request.headers.get("accept")?.includes("application/json") ||

src/lib/server/generateFromDefaultEndpoint.ts

@@ -21,6 +21,7 @@ export async function generateFromDefaultEndpoint({
 					generated_text = generated_text.slice(0, -stop.length).trimEnd();
 				}
 			}
+			tokenStream.return();
 			return generated_text;
 		}
 	}

src/routes/api/conversation/[id]/+server.ts

@@ -0,0 +1,38 @@
+import { collections } from "$lib/server/database";
+import { authCondition } from "$lib/server/auth";
+import { z } from "zod";
+import { ObjectId } from "mongodb";
+
+export async function GET({ locals, params }) {
+	const id = z.string().parse(params.id);
+	const convId = new ObjectId(id);
+
+	if (locals.user?._id || locals.sessionId) {
+		const conv = await collections.conversations.findOne({
+			_id: convId,
+			...authCondition(locals),
+		});
+
+		if (conv) {
+			const res = {
+				id: conv._id,
+				title: conv.title,
+				updatedAt: conv.updatedAt,
+				modelId: conv.model,
+				messages: conv.messages.map((message) => ({
+					content: message.content,
+					from: message.from,
+					id: message.id,
+					createdAt: message.createdAt,
+					updatedAt: message.updatedAt,
+					webSearch: message.webSearch,
+				})),
+			};
+			return Response.json(res);
+		} else {
+			return Response.json({ message: "Conversation not found" }, { status: 404 });
+		}
+	} else {
+		return Response.json({ message: "Must have session cookie" }, { status: 401 });
+	}
+}

src/routes/api/conversations/+server.ts

@@ -0,0 +1,30 @@
+import { collections } from "$lib/server/database";
+import { authCondition } from "$lib/server/auth";
+import type { Conversation } from "$lib/types/Conversation";
+
+export async function GET({ locals }) {
+	if (locals.user?._id || locals.sessionId) {
+		const convs = await collections.conversations
+			.find({
+				...authCondition(locals),
+			})
+			.project<Pick<Conversation, "_id" | "title" | "updatedAt" | "model">>({
+				title: 1,
+				updatedAt: 1,
+				model: 1,
+			})
+			.sort({ updatedAt: -1 })
+			.toArray();
+
+		const res = convs.map((conv) => ({
+			id: conv._id,
+			title: conv.title,
+			updatedAt: conv.updatedAt,
+			modelId: conv.model,
+		}));
+
+		return Response.json(res);
+	} else {
+		return Response.json({ message: "Must have session cookie" }, { status: 401 });
+	}
+}

src/routes/api/models/+server.ts

@@ -0,0 +1,19 @@
+import { models } from "$lib/server/models";
+
+export async function GET() {
+	const res = models.map((model) => ({
+		id: model.id,
+		name: model.name,
+		websiteUrl: model.websiteUrl,
+		modelUrl: model.modelUrl,
+		datasetName: model.datasetName,
+		datasetUrl: model.datasetUrl,
+		displayName: model.displayName,
+		description: model.description,
+		promptExamples: model.promptExamples,
+		preprompt: model.preprompt,
+		multimodal: model.multimodal,
+		unlisted: model.unlisted,
+	}));
+	return Response.json(res);
+}

src/routes/conversation/[id]/+page.svelte

@@ -111,14 +111,12 @@
 
 			files = [];
 
-			const responseId = randomUUID();
 			const response = await fetch(`${base}/conversation/${$page.params.id}`, {
 				method: "POST",
 				headers: { "Content-Type": "application/json" },
 				body: JSON.stringify({
 					inputs: message,
 					id: messageId,
-					response_id: responseId,
 					is_retry: isRetry,
 					web_search: $webSearchParameters.useSearch,
 					files: isRetry ? undefined : resizedImages,

src/routes/conversation/[id]/+server.ts

@@ -90,7 +90,6 @@ export async function POST({ request, locals, params, getClientAddress }) {
 
 	const {
 		inputs: newPrompt,
-		response_id: responseId,
 		id: messageId,
 		is_retry,
 		web_search: webSearch,
@@ -99,7 +98,6 @@ export async function POST({ request, locals, params, getClientAddress }) {
 		.object({
 			inputs: z.string().trim().min(1),
 			id: z.optional(z.string().uuid()),
-			response_id: z.optional(z.string().uuid()),
 			is_retry: z.optional(z.boolean()),
 			web_search: z.optional(z.boolean()),
 			files: z.optional(z.array(z.string())),
@@ -268,7 +266,7 @@ export async function POST({ request, locals, params, getClientAddress }) {
 										content: output.token.text.trimStart(),
 										webSearch: webSearchResults,
 										updates: updates,
-										id: (responseId as Message["id"]) || crypto.randomUUID(),
+										id: crypto.randomUUID(),
 										createdAt: new Date(),
 										updatedAt: new Date(),
 									},