feat(app): add image & video search functionality

ui/lib/chains/imageSearchAgent.ts

@@ -0,0 +1,90 @@
+import {
+  RunnableSequence,
+  RunnableMap,
+  RunnableLambda,
+} from '@langchain/core/runnables';
+import { PromptTemplate } from '@langchain/core/prompts';
+import formatChatHistoryAsString from '../utils/formatHistory';
+import { BaseMessage } from '@langchain/core/messages';
+import { StringOutputParser } from '@langchain/core/output_parsers';
+import { searchSearxng } from '../searxng';
+import type { BaseChatModel } from '@langchain/core/language_models/chat_models';
+
+const imageSearchChainPrompt = `
+You will be given a conversation below and a follow up question. You need to rephrase the follow-up question so it is a standalone question that can be used by the LLM to search the web for images.
+You need to make sure the rephrased question agrees with the conversation and is relevant to the conversation.
+
+Example:
+1. Follow up question: What is a cat?
+Rephrased: A cat
+
+2. Follow up question: What is a car? How does it works?
+Rephrased: Car working
+
+3. Follow up question: How does an AC work?
+Rephrased: AC working
+
+Conversation:
+{chat_history}
+
+Follow up question: {query}
+Rephrased question:
+`;
+
+type ImageSearchChainInput = {
+  chat_history: BaseMessage[];
+  query: string;
+};
+
+interface ImageSearchResult {
+  img_src: string;
+  url: string;
+  title: string;
+}
+
+const strParser = new StringOutputParser();
+
+const createImageSearchChain = (llm: BaseChatModel) => {
+  return RunnableSequence.from([
+    RunnableMap.from({
+      chat_history: (input: ImageSearchChainInput) => {
+        return formatChatHistoryAsString(input.chat_history);
+      },
+      query: (input: ImageSearchChainInput) => {
+        return input.query;
+      },
+    }),
+    PromptTemplate.fromTemplate(imageSearchChainPrompt),
+    llm,
+    strParser,
+    RunnableLambda.from(async (input: string) => {
+      const res = await searchSearxng(input, {
+        engines: ['bing images', 'google images'],
+      });
+
+      const images: ImageSearchResult[] = [];
+
+      res.results.forEach((result) => {
+        if (result.img_src && result.url && result.title) {
+          images.push({
+            img_src: result.img_src,
+            url: result.url,
+            title: result.title,
+          });
+        }
+      });
+
+      return images.slice(0, 10);
+    }),
+  ]);
+};
+
+const handleImageSearch = (
+  input: ImageSearchChainInput,
+  llm: BaseChatModel,
+) => {
+  const imageSearchChain = createImageSearchChain(llm);
+  return imageSearchChain.invoke(input);
+};
+
+export default handleImageSearch;

ui/lib/chains/videoSearchAgent.ts

@@ -0,0 +1,97 @@
+import {
+  RunnableSequence,
+  RunnableMap,
+  RunnableLambda,
+} from '@langchain/core/runnables';
+import { PromptTemplate } from '@langchain/core/prompts';
+import formatChatHistoryAsString from '../utils/formatHistory';
+import { BaseMessage } from '@langchain/core/messages';
+import { StringOutputParser } from '@langchain/core/output_parsers';
+import { searchSearxng } from '../searxng';
+import type { BaseChatModel } from '@langchain/core/language_models/chat_models';
+
+const VideoSearchChainPrompt = `
+  You will be given a conversation below and a follow up question. You need to rephrase the follow-up question so it is a standalone question that can be used by the LLM to search Youtube for videos.
+  You need to make sure the rephrased question agrees with the conversation and is relevant to the conversation.
+  
+  Example:
+  1. Follow up question: How does a car work?
+  Rephrased: How does a car work?
+  
+  2. Follow up question: What is the theory of relativity?
+  Rephrased: What is theory of relativity
+  
+  3. Follow up question: How does an AC work?
+  Rephrased: How does an AC work
+  
+  Conversation:
+  {chat_history}
+  
+  Follow up question: {query}
+  Rephrased question:
+  `;
+
+type VideoSearchChainInput = {
+  chat_history: BaseMessage[];
+  query: string;
+};
+
+interface VideoSearchResult {
+  img_src: string;
+  url: string;
+  title: string;
+  iframe_src: string;
+}
+
+const strParser = new StringOutputParser();
+
+const createVideoSearchChain = (llm: BaseChatModel) => {
+  return RunnableSequence.from([
+    RunnableMap.from({
+      chat_history: (input: VideoSearchChainInput) => {
+        return formatChatHistoryAsString(input.chat_history);
+      },
+      query: (input: VideoSearchChainInput) => {
+        return input.query;
+      },
+    }),
+    PromptTemplate.fromTemplate(VideoSearchChainPrompt),
+    llm,
+    strParser,
+    RunnableLambda.from(async (input: string) => {
+      const res = await searchSearxng(input, {
+        engines: ['youtube'],
+      });
+
+      const videos: VideoSearchResult[] = [];
+
+      res.results.forEach((result) => {
+        if (
+          result.thumbnail &&
+          result.url &&
+          result.title &&
+          result.iframe_src
+        ) {
+          videos.push({
+            img_src: result.thumbnail,
+            url: result.url,
+            title: result.title,
+            iframe_src: result.iframe_src,
+          });
+        }
+      });
+
+      return videos.slice(0, 10);
+    }),
+  ]);
+};
+
+const handleVideoSearch = (
+  input: VideoSearchChainInput,
+  llm: BaseChatModel,
+) => {
+  const VideoSearchChain = createVideoSearchChain(llm);
+  return VideoSearchChain.invoke(input);
+};
+
+export default handleVideoSearch;