add lang chain as AI layer

.env.example

@@ -14,4 +14,9 @@ NEXTAUTH_SECRET=
 NEXTAUTH_URL=
 
 AZURE_COSMOSEDB_URI=
-AZURE_COSMOSEDB_KEY=
+AZURE_COSMOSEDB_KEY=
+
+AZURE_SEARCH_API_KEY=
+AZURE_SEARCH_NAME=
+AZURE_SEARCH_INDEX_NAME=
+AZURE_SEARCH_API_VERSION=

README.md

@@ -33,7 +33,7 @@ AzureChatGPT is built with the following technologies.
 
 # 💙 One click Azure deployment
 
-[![Deploy to Azure](https://aka.ms/deploytoazurebutton)](https://portal.azure.com/#create/Microsoft.Template/uri/https%3A%2F%2Fraw.githubusercontent.com%2Fmicrosoft%2Fazurechatgpt%2Fmain%2Finfra%2Fmain.json)
+[![Deploy to Azure](https://aka.ms/deploytoazurebutton)](https://portal.azure.com/#create/Microsoft.Template/uri/https%3A%2F%2Fraw.githubusercontent.com%2Fthivy%2Fazure-chatgpt%2Fmain%2Finfra%2Fmain.json)
 
 Click on the Deploy to Azure button and configure your settings in the Azure Portal as described in the [Environment variables](#🔑-environment-variables) section.
 

app/api/chat/route.ts

@@ -0,0 +1,6 @@
+import { PromptGPT } from "@/features/chat/chat-api";
+
+export async function POST(req: Request) {
+  const body = await req.json();
+  return await PromptGPT(body);
+}

app/chat/[id]/page.tsx

@@ -3,8 +3,6 @@ import { FindChatThreadByID } from "@/features/chat/chat-thread-service";
 import { ChatUI } from "@/features/chat/chat-ui";
 import { notFound } from "next/navigation";
 
-export const revalidate = 0;
-
 export default async function Home({ params }: { params: { id: string } }) {
   const items = await FindAllChats(params.id);
   const thread = await FindChatThreadByID(params.id);

app/chat/layout.tsx

@@ -2,8 +2,6 @@ import { ProtectedPage } from "@/features/auth/protected-page";
 import { ChatMenu } from "@/features/chat/chat-menu/chat-menu";
 import { MainMenu } from "@/features/menu/menu";
 
-export const revalidate = 0;
-
 export const metadata = {
   title: "AzureChatGPT",
   description: "AzureChatGPT",

app/page.tsx

@@ -3,8 +3,6 @@ import { Card } from "@/components/ui/card";
 import { userSession } from "@/features/auth/helpers";
 import { redirect } from "next/navigation";
 
-export const revalidate = 0;
-
 export default async function Home() {
   const user = await userSession();
   if (user) {

app/reporting/layout.tsx

@@ -1,8 +1,6 @@
 import { ProtectedPage } from "@/features/auth/protected-page";
 import { MainMenu } from "@/features/menu/menu";
 
-export const revalidate = 0;
-
 export const metadata = {
   title: "AzureChatGPT",
   description: "AzureChatGPT",

components/chat/chat-row.tsx

@@ -1,3 +1,4 @@
+import { chatRole } from "@/features/chat/chat-service";
 import { cn } from "@/lib/utils";
 import { FC } from "react";
 import remarkGfm from "remark-gfm";
@@ -10,7 +11,7 @@ interface ChatRowProps {
   name: string;
   profilePicture: string;
   message: string;
-  type: "user" | "assistant" | "system";
+  type: chatRole;
 }
 
 const ChatRow: FC<ChatRowProps> = (props) => {

components/hooks/use-chat-scroll-anchor.tsx

@@ -1,10 +1,10 @@
 "use client";
 
-import { ChatMessageOutputModel } from "@/features/chat/chat-service";
+import { Message } from "ai";
 import { RefObject, useEffect } from "react";
 
 export const useChatScrollAnchor = (
-  chats: ChatMessageOutputModel[],
+  chats: Message[],
   ref: RefObject<HTMLDivElement>
 ) => {
   useEffect(() => {

features/azure-cog-search/azure-cog-vector-store.ts

@@ -0,0 +1,196 @@
+import { Callbacks } from "langchain/callbacks";
+import { Document } from "langchain/document";
+import { Embeddings } from "langchain/embeddings/base";
+import { VectorStore } from "langchain/vectorstores/base";
+import { nanoid } from "nanoid";
+
+// example index model below
+// export interface AzureCogDocumentIndex extends Record<string, unknown> {
+//     id: string;
+//     content: string;
+//     user: string;
+//     embedding?: number[];
+//     pageContent: string;
+//     metadata: any;
+//   }
+
+interface AzureSearchConfig {
+  name: string;
+  indexName: string;
+  apiKey: string;
+  apiVersion: string;
+  vectorFieldName: string;
+}
+
+interface DocumentSearchResponseModel<TModel> {
+  value: TModel[];
+}
+
+type DocumentSearchModel = {
+  "@search.score": number;
+};
+
+export interface AzureCogDocument extends Record<string, unknown> {}
+
+type AzureCogVectorField = {
+  value: number[];
+  fields: string;
+  k: number;
+};
+
+type AzureCogFilter = {
+  search?: string;
+  facets?: string[];
+  filter?: string;
+  top?: number;
+  vectorFields: string;
+};
+
+type AzureCogRequestObject = {
+  search: string;
+  facets: string[];
+  filter: string;
+  vectors: AzureCogVectorField[];
+  top: number;
+};
+
+export class AzureCogSearch<
+  TModel extends Record<string, unknown>
+> extends VectorStore {
+  private _config: AzureSearchConfig;
+
+  constructor(embeddings: Embeddings, dbConfig: AzureSearchConfig) {
+    super(embeddings, dbConfig);
+    this._config = dbConfig;
+  }
+
+  _vectorstoreType(): string {
+    return "azure-cog-search";
+  }
+
+  get config(): AzureSearchConfig {
+    return this._config;
+  }
+
+  get baseUrl(): string {
+    return `https://${this._config.name}.search.windows.net/indexes/${this._config.indexName}/docs`;
+  }
+
+  async addDocuments(documents: Document<TModel>[]): Promise<string[]> {
+    const texts = documents.map(({ pageContent }) => pageContent);
+    return this.addVectors(
+      await this.embeddings.embedDocuments(texts),
+      documents
+    );
+  }
+
+  /**
+   * Search for the most similar documents to a query
+   */
+  async similaritySearch(
+    query: string,
+    k?: number,
+    filter?: AzureCogFilter
+  ): Promise<Document<TModel>[]> {
+    const results = await this.similaritySearchVectorWithScore(
+      await this.embeddings.embedQuery(query),
+      k || 4,
+      filter
+    );
+
+    return results.map(([doc, _score]) => doc);
+  }
+
+  /**
+   * Search for the most similar documents to a query,
+   * and return their similarity score
+   */
+  async similaritySearchWithScore(
+    query: string,
+    k?: number,
+    filter?: AzureCogFilter,
+    _callbacks: Callbacks | undefined = undefined
+  ): Promise<[Document, number][]> {
+    const embeddings = await this.embeddings.embedQuery(query);
+    return this.similaritySearchVectorWithScore(embeddings, k || 5, filter);
+  }
+
+  /**
+   * Advanced: Add more documents to an existing VectorStore,
+   * when you already have their embeddings
+   */
+  async addVectors(
+    vectors: number[][],
+    documents: Document<TModel>[]
+  ): Promise<string[]> {
+    const indexes: Array<any> = [];
+
+    documents.forEach((document, i) => {
+      indexes.push({
+        id: nanoid(),
+        content: document.pageContent,
+        ...document,
+        [this._config.vectorFieldName]: vectors[i],
+      });
+    });
+
+    const documentIndexRequest: DocumentSearchResponseModel<TModel> = {
+      value: indexes,
+    };
+
+    const url = `${this.baseUrl}/index?api-version=${this._config.apiVersion}`;
+    const responseObj = await fetcher(
+      url,
+      documentIndexRequest,
+      this._config.apiKey
+    );
+    return responseObj.value.map((doc: any) => doc.key);
+  }
+
+  /**
+   * Advanced: Search for the most similar documents to a query,
+   * when you already have the embedding of the query
+   */
+  async similaritySearchVectorWithScore(
+    query: number[],
+    k: number,
+    filter?: AzureCogFilter
+  ): Promise<[Document<TModel>, number][]> {
+    const url = `${this.baseUrl}/search?api-version=${this._config.apiVersion}`;
+
+    const searchBody: AzureCogRequestObject = {
+      search: filter?.search || "*",
+      facets: filter?.facets || [],
+      filter: filter?.filter || "",
+      vectors: [{ value: query, fields: filter?.vectorFields || "", k: k }],
+      top: filter?.top || k,
+    };
+
+    const resultDocuments = (await fetcher(
+      url,
+      searchBody,
+      this._config.apiKey
+    )) as DocumentSearchResponseModel<Document<TModel> & DocumentSearchModel>;
+
+    return resultDocuments.value.map((doc) => [doc, doc["@search.score"] || 0]);
+  }
+}
+
+const fetcher = async (url: string, body: any, apiKey: string) => {
+  const response = await fetch(url, {
+    method: "POST",
+    body: JSON.stringify(body),
+    headers: {
+      "Content-Type": "application/json",
+      "api-key": apiKey,
+    },
+  });
+
+  if (!response.ok) {
+    const err = await response.json();
+    console.log(err);
+    throw new Error(err);
+  }
+
+  return await response.json();
+};