Merge pull request #736 from ItzCrazyKns/master

Merge master into feat/deep-research

docs/API/SEARCH.md

@@ -33,6 +33,7 @@ The API accepts a JSON object in the request body, where you define the focus mo
     ["human", "Hi, how are you?"],
     ["assistant", "I am doing well, how can I help you today?"]
   ],
+  "systemInstructions": "Focus on providing technical details about Perplexica's architecture.",
   "stream": false
 }
 ```
@@ -63,6 +64,8 @@ The API accepts a JSON object in the request body, where you define the focus mo
 
 - **`query`** (string, required): The search query or question.
 
+- **`systemInstructions`** (string, optional): Custom instructions provided by the user to guide the AI's response. These instructions are treated as user preferences and have lower priority than the system's core instructions. For example, you can specify a particular writing style, format, or focus area.
+
 - **`history`** (array, optional): An array of message pairs representing the conversation history. Each pair consists of a role (either 'human' or 'assistant') and the message content. This allows the system to use the context of the conversation to refine results. Example:
 
   ```json

package.json

@@ -1,6 +1,6 @@
 {
   "name": "perplexica-frontend",
-  "version": "1.10.1",
+  "version": "1.10.2",
   "license": "MIT",
   "author": "ItzCrazyKns",
   "scripts": {

sample.config.toml

@@ -22,5 +22,8 @@ MODEL_NAME = ""
 [MODELS.OLLAMA]
 API_URL = "" # Ollama API URL - http://host.docker.internal:11434
 
+[MODELS.DEEPSEEK]
+API_KEY = ""
+
 [API_ENDPOINTS]
 SEARXNG = "" # SearxNG API URL - http://localhost:32768

src/app/api/chat/route.ts

@@ -49,6 +49,7 @@ type Body = {
   files: Array<string>;
   chatModel: ChatModel;
   embeddingModel: EmbeddingModel;
+  systemInstructions: string;
 };
 
 const handleEmitterEvents = async (
@@ -278,6 +279,7 @@ export const POST = async (req: Request) => {
       embedding,
       body.optimizationMode,
       body.files,
+      body.systemInstructions,
     );
 
     const responseStream = new TransformStream();

src/app/api/config/route.ts

@@ -7,6 +7,7 @@ import {
   getGroqApiKey,
   getOllamaApiEndpoint,
   getOpenaiApiKey,
+  getDeepseekApiKey,
   updateConfig,
 } from '@/lib/config';
 import {
@@ -53,6 +54,7 @@ export const GET = async (req: Request) => {
     config['anthropicApiKey'] = getAnthropicApiKey();
     config['groqApiKey'] = getGroqApiKey();
     config['geminiApiKey'] = getGeminiApiKey();
+    config['deepseekApiKey'] = getDeepseekApiKey();
     config['customOpenaiApiUrl'] = getCustomOpenaiApiUrl();
     config['customOpenaiApiKey'] = getCustomOpenaiApiKey();
     config['customOpenaiModelName'] = getCustomOpenaiModelName();
@@ -88,6 +90,9 @@ export const POST = async (req: Request) => {
         OLLAMA: {
           API_URL: config.ollamaApiUrl,
         },
+        DEEPSEEK: {
+          API_KEY: config.deepseekApiKey,
+        },
         CUSTOM_OPENAI: {
           API_URL: config.customOpenaiApiUrl,
           API_KEY: config.customOpenaiApiKey,

src/app/api/search/route.ts

@@ -34,6 +34,7 @@ interface ChatRequestBody {
   query: string;
   history: Array<[string, string]>;
   stream?: boolean;
+  systemInstructions?: string;
 }
 
 export const POST = async (req: Request) => {
@@ -125,6 +126,7 @@ export const POST = async (req: Request) => {
       embeddings,
       body.optimizationMode,
       [],
+      body.systemInstructions || '',
     );
 
     if (!body.stream) {

src/app/settings/page.tsx

@@ -20,6 +20,7 @@ interface SettingsType {
   anthropicApiKey: string;
   geminiApiKey: string;
   ollamaApiUrl: string;
+  deepseekApiKey: string;
   customOpenaiApiKey: string;
   customOpenaiApiUrl: string;
   customOpenaiModelName: string;
@@ -54,6 +55,38 @@ const Input = ({ className, isSaving, onSave, ...restProps }: InputProps) => {
   );
 };
 
+interface TextareaProps extends React.InputHTMLAttributes<HTMLTextAreaElement> {
+  isSaving?: boolean;
+  onSave?: (value: string) => void;
+}
+
+const Textarea = ({
+  className,
+  isSaving,
+  onSave,
+  ...restProps
+}: TextareaProps) => {
+  return (
+    <div className="relative">
+      <textarea
+        placeholder="Any special instructions for the LLM"
+        className="placeholder:text-sm text-sm w-full flex items-center justify-between p-3 bg-light-secondary dark:bg-dark-secondary rounded-lg hover:bg-light-200 dark:hover:bg-dark-200 transition-colors"
+        rows={4}
+        onBlur={(e) => onSave?.(e.target.value)}
+        {...restProps}
+      />
+      {isSaving && (
+        <div className="absolute right-3 top-3">
+          <Loader2
+            size={16}
+            className="animate-spin text-black/70 dark:text-white/70"
+          />
+        </div>
+      )}
+    </div>
+  );
+};
+
 const Select = ({
   className,
   options,
@@ -111,6 +144,7 @@ const Page = () => {
   const [isLoading, setIsLoading] = useState(false);
   const [automaticImageSearch, setAutomaticImageSearch] = useState(false);
   const [automaticVideoSearch, setAutomaticVideoSearch] = useState(false);
+  const [systemInstructions, setSystemInstructions] = useState<string>('');
   const [savingStates, setSavingStates] = useState<Record<string, boolean>>({});
 
   useEffect(() => {
@@ -172,6 +206,8 @@ const Page = () => {
         localStorage.getItem('autoVideoSearch') === 'true',
       );
 
+      setSystemInstructions(localStorage.getItem('systemInstructions')!);
+
       setIsLoading(false);
     };
 
@@ -328,6 +364,8 @@ const Page = () => {
         localStorage.setItem('embeddingModelProvider', value);
       } else if (key === 'embeddingModel') {
         localStorage.setItem('embeddingModel', value);
+      } else if (key === 'systemInstructions') {
+        localStorage.setItem('systemInstructions', value);
       }
     } catch (err) {
       console.error('Failed to save:', err);
@@ -473,6 +511,19 @@ const Page = () => {
               </div>
             </SettingsSection>
 
+            <SettingsSection title="System Instructions">
+              <div className="flex flex-col space-y-4">
+                <Textarea
+                  value={systemInstructions}
+                  isSaving={savingStates['systemInstructions']}
+                  onChange={(e) => {
+                    setSystemInstructions(e.target.value);
+                  }}
+                  onSave={(value) => saveConfig('systemInstructions', value)}
+                />
+              </div>
+            </SettingsSection>
+
             <SettingsSection title="Model Settings">
               {config.chatModelProviders && (
                 <div className="flex flex-col space-y-4">
@@ -788,6 +839,25 @@ const Page = () => {
                     onSave={(value) => saveConfig('geminiApiKey', value)}
                   />
                 </div>
+
+                <div className="flex flex-col space-y-1">
+                  <p className="text-black/70 dark:text-white/70 text-sm">
+                    Deepseek API Key
+                  </p>
+                  <Input
+                    type="text"
+                    placeholder="Deepseek API Key"
+                    value={config.deepseekApiKey}
+                    isSaving={savingStates['deepseekApiKey']}
+                    onChange={(e) => {
+                      setConfig((prev) => ({
+                        ...prev!,
+                        deepseekApiKey: e.target.value,
+                      }));
+                    }}
+                    onSave={(value) => saveConfig('deepseekApiKey', value)}
+                  />
+                </div>
               </div>
             </SettingsSection>
           </div>

src/components/ChatWindow.tsx

@@ -478,6 +478,7 @@ const ChatWindow = ({ id }: { id?: string }) => {
           name: embeddingModelProvider.name,
           provider: embeddingModelProvider.provider,
         },
+        systemInstructions: localStorage.getItem('systemInstructions'),
       }),
     });
 

src/components/MessageBox.tsx

@@ -48,6 +48,7 @@ const MessageBox = ({
   const [speechMessage, setSpeechMessage] = useState(message.content);
 
   useEffect(() => {
+    const citationRegex = /\[([^\]]+)\]/g;
     const regex = /\[(\d+)\]/g;
     let processedMessage = message.content;
 
@@ -67,11 +68,33 @@ const MessageBox = ({
     ) {
       setParsedMessage(
         processedMessage.replace(
-          regex,
-          (_, number) =>
-            `<a href="${
-              message.sources?.[number - 1]?.metadata?.url
-            }" target="_blank" className="bg-light-secondary dark:bg-dark-secondary px-1 rounded ml-1 no-underline text-xs text-black/70 dark:text-white/70 relative">${number}</a>`,
+          citationRegex,
+          (_, capturedContent: string) => {
+            const numbers = capturedContent
+              .split(',')
+              .map((numStr) => numStr.trim());
+
+            const linksHtml = numbers
+              .map((numStr) => {
+                const number = parseInt(numStr);
+
+                if (isNaN(number) || number <= 0) {
+                  return `[${numStr}]`;
+                }
+
+                const source = message.sources?.[number - 1];
+                const url = source?.metadata?.url;
+
+                if (url) {
+                  return `<a href="${url}" target="_blank" className="bg-light-secondary dark:bg-dark-secondary px-1 rounded ml-1 no-underline text-xs text-black/70 dark:text-white/70 relative">${numStr}</a>`;
+                } else {
+                  return `[${numStr}]`;
+                }
+              })
+              .join('');
+
+            return linksHtml;
+          },
         ),
       );
       return;

src/components/MessageInputActions/Optimization.tsx

@@ -76,13 +76,11 @@ const Optimization = ({
               <PopoverButton
                 onClick={() => setOptimizationMode(mode.key)}
                 key={i}
-                disabled={mode.key === 'quality'}
                 className={cn(
                   'p-2 rounded-lg flex flex-col items-start justify-start text-start space-y-1 duration-200 cursor-pointer transition',
                   optimizationMode === mode.key
                     ? 'bg-light-secondary dark:bg-dark-secondary'
                     : 'hover:bg-light-secondary dark:hover:bg-dark-secondary',
-                  mode.key === 'quality' && 'opacity-50 cursor-not-allowed',
                 )}
               >
                 <div className="flex flex-row items-center space-x-1 text-black dark:text-white">

src/lib/config.ts

@@ -25,6 +25,9 @@ interface Config {
     OLLAMA: {
       API_URL: string;
     };
+    DEEPSEEK: {
+      API_KEY: string;
+    };
     CUSTOM_OPENAI: {
       API_URL: string;
       API_KEY: string;
@@ -63,6 +66,8 @@ export const getSearxngApiEndpoint = () =>
 
 export const getOllamaApiEndpoint = () => loadConfig().MODELS.OLLAMA.API_URL;
 
+export const getDeepseekApiKey = () => loadConfig().MODELS.DEEPSEEK.API_KEY;
+
 export const getCustomOpenaiApiKey = () =>
   loadConfig().MODELS.CUSTOM_OPENAI.API_KEY;
 

src/lib/prompts/academicSearch.ts

@@ -51,6 +51,10 @@ export const academicSearchResponsePrompt = `
     - If no relevant information is found, say: "Hmm, sorry I could not find any relevant information on this topic. Would you like me to search again or ask something else?" Be transparent about limitations and suggest alternatives or ways to reframe the query.
     - You are set on focus mode 'Academic', this means you will be searching for academic papers and articles on the web.
     
+    ### User instructions
+    These instructions are shared to you by the user and not by the system. You will have to follow them but give them less priority than the above instructions. If the user has provided specific instructions or preferences, incorporate them into your response while adhering to the overall guidelines.
+    {systemInstructions}
+
     ### Example Output
     - Begin with a brief introduction summarizing the event or query topic.
     - Follow with detailed sections under clear headings, covering all aspects of the query if possible.

src/lib/prompts/redditSearch.ts

@@ -51,6 +51,10 @@ export const redditSearchResponsePrompt = `
     - If no relevant information is found, say: "Hmm, sorry I could not find any relevant information on this topic. Would you like me to search again or ask something else?" Be transparent about limitations and suggest alternatives or ways to reframe the query.
     - You are set on focus mode 'Reddit', this means you will be searching for information, opinions and discussions on the web using Reddit.
     
+    ### User instructions
+    These instructions are shared to you by the user and not by the system. You will have to follow them but give them less priority than the above instructions. If the user has provided specific instructions or preferences, incorporate them into your response while adhering to the overall guidelines.
+    {systemInstructions}
+
     ### Example Output
     - Begin with a brief introduction summarizing the event or query topic.
     - Follow with detailed sections under clear headings, covering all aspects of the query if possible.

src/lib/prompts/webSearch.ts

@@ -1,6 +1,6 @@
 export const webSearchRetrieverPrompt = `
 You are an AI question rephraser. You will be given a conversation and a follow-up question,  you will have to rephrase the follow up question so it is a standalone question and can be used by another LLM to search the web for information to answer it.
-If it is a smple writing task or a greeting (unless the greeting contains a question after it) like Hi, Hello, How are you, etc. than a question then you need to return \`not_needed\` as the response (This is because the LLM won't need to search the web for finding information on this topic).
+If it is a simple writing task or a greeting (unless the greeting contains a question after it) like Hi, Hello, How are you, etc. than a question then you need to return \`not_needed\` as the response (This is because the LLM won't need to search the web for finding information on this topic).
 If the user asks some question from some URL or wants you to summarize a PDF or a webpage (via URL) you need to return the links inside the \`links\` XML block and the question inside the \`question\` XML block. If the user wants to you to summarize the webpage or the PDF you need to return \`summarize\` inside the \`question\` XML block in place of a question and the link to summarize in the \`links\` XML block.
 You must always return the rephrased question inside the \`question\` XML block, if there are no links in the follow-up question then don't insert a \`links\` XML block in your response.
 
@@ -92,6 +92,10 @@ export const webSearchResponsePrompt = `
     - If the user provides vague input or if relevant information is missing, explain what additional details might help refine the search.
     - If no relevant information is found, say: "Hmm, sorry I could not find any relevant information on this topic. Would you like me to search again or ask something else?" Be transparent about limitations and suggest alternatives or ways to reframe the query.
 
+    ### User instructions
+    These instructions are shared to you by the user and not by the system. You will have to follow them but give them less priority than the above instructions. If the user has provided specific instructions or preferences, incorporate them into your response while adhering to the overall guidelines.
+    {systemInstructions}
+
     ### Example Output
     - Begin with a brief introduction summarizing the event or query topic.
     - Follow with detailed sections under clear headings, covering all aspects of the query if possible.

src/lib/prompts/wolframAlpha.ts

@@ -51,6 +51,10 @@ export const wolframAlphaSearchResponsePrompt = `
     - If no relevant information is found, say: "Hmm, sorry I could not find any relevant information on this topic. Would you like me to search again or ask something else?" Be transparent about limitations and suggest alternatives or ways to reframe the query.
     - You are set on focus mode 'Wolfram Alpha', this means you will be searching for information on the web using Wolfram Alpha. It is a computational knowledge engine that can answer factual queries and perform computations.
     
+    ### User instructions
+    These instructions are shared to you by the user and not by the system. You will have to follow them but give them less priority than the above instructions. If the user has provided specific instructions or preferences, incorporate them into your response while adhering to the overall guidelines.
+    {systemInstructions}
+
     ### Example Output
     - Begin with a brief introduction summarizing the event or query topic.
     - Follow with detailed sections under clear headings, covering all aspects of the query if possible.

src/lib/prompts/writingAssistant.ts

@@ -7,6 +7,10 @@ You have to cite the answer using [number] notation. You must cite the sentences
 Place these citations at the end of that particular sentence. You can cite the same sentence multiple times if it is relevant to the user's query like [number1][number2].
 However you do not need to cite it using the same number. You can use different numbers to cite the same sentence multiple times. The number refers to the number of the search result (passed in the context) used to generate that part of the answer.
 
+### User instructions
+These instructions are shared to you by the user and not by the system. You will have to follow them but give them less priority than the above instructions. If the user has provided specific instructions or preferences, incorporate them into your response while adhering to the overall guidelines.
+{systemInstructions}
+
 <context>
 {context}
 </context>

src/lib/prompts/youtubeSearch.ts

@@ -51,6 +51,10 @@ export const youtubeSearchResponsePrompt = `
     - If no relevant information is found, say: "Hmm, sorry I could not find any relevant information on this topic. Would you like me to search again or ask something else?" Be transparent about limitations and suggest alternatives or ways to reframe the query.
     - You are set on focus mode 'Youtube', this means you will be searching for videos on the web using Youtube and providing information based on the video's transcrip
     
+    ### User instructions
+    These instructions are shared to you by the user and not by the system. You will have to follow them but give them less priority than the above instructions. If the user has provided specific instructions or preferences, incorporate them into your response while adhering to the overall guidelines.
+    {systemInstructions}
+
     ### Example Output
     - Begin with a brief introduction summarizing the event or query topic.
     - Follow with detailed sections under clear headings, covering all aspects of the query if possible.

src/lib/providers/deepseek.ts

@@ -0,0 +1,44 @@
+import { ChatOpenAI } from '@langchain/openai';
+import { getDeepseekApiKey } from '../config';
+import { ChatModel } from '.';
+import { BaseChatModel } from '@langchain/core/language_models/chat_models';
+
+const deepseekChatModels: Record<string, string>[] = [
+  {
+    displayName: 'Deepseek Chat (Deepseek V3)',
+    key: 'deepseek-chat',
+  },
+  {
+    displayName: 'Deepseek Reasoner (Deepseek R1)',
+    key: 'deepseek-reasoner',
+  },
+];
+
+export const loadDeepseekChatModels = async () => {
+  const deepseekApiKey = getDeepseekApiKey();
+
+  if (!deepseekApiKey) return {};
+
+  try {
+    const chatModels: Record<string, ChatModel> = {};
+
+    deepseekChatModels.forEach((model) => {
+      chatModels[model.key] = {
+        displayName: model.displayName,
+        model: new ChatOpenAI({
+          openAIApiKey: deepseekApiKey,
+          modelName: model.key,
+          temperature: 0.7,
+          configuration: {
+            baseURL: 'https://api.deepseek.com',
+          },
+        }) as unknown as BaseChatModel,
+      };
+    });
+
+    return chatModels;
+  } catch (err) {
+    console.error(`Error loading Deepseek models: ${err}`);
+    return {};
+  }
+};

src/lib/providers/gemini.ts

@@ -40,8 +40,12 @@ const geminiChatModels: Record<string, string>[] = [
 
 const geminiEmbeddingModels: Record<string, string>[] = [
   {
-    displayName: 'Gemini Embedding',
-    key: 'gemini-embedding-exp',
+    displayName: 'Text Embedding 004',
+    key: 'models/text-embedding-004',
+  },
+  {
+    displayName: 'Embedding 001',
+    key: 'models/embedding-001',
   },
 ];
 

src/lib/providers/groq.ts

@@ -72,6 +72,14 @@ const groqChatModels: Record<string, string>[] = [
     displayName: 'Llama 3.2 90B Vision Preview (Preview)',
     key: 'llama-3.2-90b-vision-preview',
   },
+  /* {
+    displayName: 'Llama 4 Maverick 17B 128E Instruct (Preview)',
+    key: 'meta-llama/llama-4-maverick-17b-128e-instruct',
+  }, */
+  {
+    displayName: 'Llama 4 Scout 17B 16E Instruct (Preview)',
+    key: 'meta-llama/llama-4-scout-17b-16e-instruct',
+  },
 ];
 
 export const loadGroqChatModels = async () => {

src/lib/providers/index.ts

@@ -12,6 +12,7 @@ import { loadGroqChatModels } from './groq';
 import { loadAnthropicChatModels } from './anthropic';
 import { loadGeminiChatModels, loadGeminiEmbeddingModels } from './gemini';
 import { loadTransformersEmbeddingsModels } from './transformers';
+import { loadDeepseekChatModels } from './deepseek';
 
 export interface ChatModel {
   displayName: string;
@@ -32,6 +33,7 @@ export const chatModelProviders: Record<
   groq: loadGroqChatModels,
   anthropic: loadAnthropicChatModels,
   gemini: loadGeminiChatModels,
+  deepseek: loadDeepseekChatModels,
 };
 
 export const embeddingModelProviders: Record<

src/lib/search/metaSearchAgent.ts

@@ -12,7 +12,7 @@ import LineListOutputParser from '../outputParsers/listLineOutputParser';
 import LineOutputParser from '../outputParsers/lineOutputParser';
 import { getDocumentsFromLinks } from '../utils/documents';
 import { Document } from 'langchain/document';
-import { searchSearxng } from '../searxng';
+import { searchSearxng, SearxngSearchResult } from '../searxng';
 import path from 'node:path';
 import fs from 'node:fs';
 import computeSimilarity from '../utils/computeSimilarity';
@@ -29,6 +29,7 @@ export interface MetaSearchAgentType {
     embeddings: Embeddings,
     optimizationMode: 'speed' | 'balanced' | 'quality',
     fileIds: string[],
+    systemInstructions: string,
   ) => Promise<eventEmitter>;
 }
 
@@ -234,11 +235,140 @@ class MetaSearchAgent implements MetaSearchAgentType {
     }
   }
 
+  private async performDeepResearch(
+    llm: BaseChatModel,
+    input: SearchInput,
+    emitter: EventEmitter,
+  ) {
+    (llm as unknown as ChatOpenAI).temperature = 0;
+
+    const queryGenPrompt = PromptTemplate.fromTemplate(
+      this.config.queryGeneratorPrompt,
+    );
+
+    const formattedChatPrompt = await queryGenPrompt.invoke({
+      chat_history: formatChatHistoryAsString(input.chat_history),
+      query: input.query,
+    });
+
+    let i = 0;
+    let currentQuery = await this.strParser.invoke(
+      await llm.invoke(formattedChatPrompt),
+    );
+    const originalQuery = currentQuery;
+    const pastQueries: string[] = [];
+    const results: SearxngSearchResult[] = [];
+
+    while (i < 10) {
+      const res = await searchSearxng(currentQuery, {
+        language: 'en',
+        engines: this.config.activeEngines,
+      });
+
+      results.push(...res.results);
+
+      const reflectorPrompt = PromptTemplate.fromTemplate(`
+          You are an LLM that is tasked with reflecting on the results of a search query.
+
+          ## Goal
+          You will be given question of the user, a list of search results collected from the web to answer that question along with past queries made to collect those results. You have to analyze the results based on user's question and do the following:
+          
+          1. Identify unexplored areas or areas with less detailed information in the results and generate a new query that focuses on those areas. The new queries should be more specific and a similar query shall not exist in past queries which will be provided to you. Make sure to include keywords that you're looking for because the new query will be used to search the web for information on that topic. Make sure the query contains only 1 question and is not too long to ensure it is Search Engine friendly.
+          2. You'll have to generate a description explaining what you are doing for example "I am looking for more information about X" or "Understanding how X works" etc. The description should be short and concise.
+          
+          ## Output format
+
+          You need to output in XML format and do not generate any other text. ake sure to not include any other text in the output or start a conversation in the output. The output should be in the following format:
+
+          <query>(query)</query>
+          <description>(description)</description>
+
+          ## Example
+          Say the user asked "What is Llama 4 by Meta?" and let search results contain information about Llama 4 being an LLM and very little information about its features. You can output:
+
+          <query>Llama 4 features</query> // Generate queries that capture keywords for SEO and not making words like "How", "What", "Why" etc.
+          <description>Looking for new features in Llama 4</description>
+
+          or something like
+
+          <query>How is Llama 4 better than its previous generation models</query>
+          <description>Understanding the difference between Llama 4 and previous generation models.</description>
+          
+          ## BELOW IS THE ACTUAL DATA YOU WILL BE WORKING WITH. IT IS NOT A PART OF EXAMPLES. YOU'LL HAVE TO GENERATE YOUR ANSWER BASED ON THIS DATA.
+          <user_question>\n{question}\n</user_question>
+          <search_results>\n{search_results}\n</search_results>
+          <past_queries>\n{past_queries}\n</past_queries>
+
+          Response:
+      `);
+
+      const formattedReflectorPrompt = await reflectorPrompt.invoke({
+        question: originalQuery,
+        search_results: results
+          .map(
+            (result) => `<result>${result.title} - ${result.content}</result>`,
+          )
+          .join('\n'),
+        past_queries: pastQueries.map((q) => `<query>${q}</query>`).join('\n'),
+      });
+
+      const feedback = await this.strParser.invoke(
+        await llm.invoke(formattedReflectorPrompt),
+      );
+
+      console.log(`Feedback: ${feedback}`);
+
+      const queryOutputParser = new LineOutputParser({
+        key: 'query',
+      });
+
+      const descriptionOutputParser = new LineOutputParser({
+        key: 'description',
+      });
+
+      currentQuery = await queryOutputParser.parse(feedback);
+      const description = await descriptionOutputParser.parse(feedback);
+      console.log(`Query: ${currentQuery}`);
+      console.log(`Description: ${description}`);
+
+      pastQueries.push(currentQuery);
+      ++i;
+    }
+
+    const uniqueResults: SearxngSearchResult[] = [];
+
+    results.forEach((res) => {
+      const exists = uniqueResults.find((r) => r.url === res.url);
+
+      if (!exists) {
+        uniqueResults.push(res);
+      } else {
+        exists.content += `\n\n` + res.content;
+      }
+    });
+
+    const documents = uniqueResults /* .slice(0, 50) */
+      .map(
+        (r) =>
+          new Document({
+            pageContent: r.content || '',
+            metadata: {
+              title: r.title,
+              url: r.url,
+              ...(r.img_src && { img_src: r.img_src }),
+            },
+          }),
+      );
+
+    return documents;
+  }
+
   private async streamAnswer(
     llm: BaseChatModel,
     fileIds: string[],
     embeddings: Embeddings,
     optimizationMode: 'speed' | 'balanced' | 'quality',
+    systemInstructions: string,
     input: SearchInput,
     emitter: EventEmitter,
   ) {
@@ -248,33 +378,49 @@ class MetaSearchAgent implements MetaSearchAgentType {
       ['user', '{query}'],
     ]);
 
-    let docs: Document[] | null = null;
-    let query = input.query;
+    let context = '';
 
-    if (this.config.searchWeb) {
-      const searchResults = await this.searchSources(llm, input, emitter);
+    if (optimizationMode === 'speed' || optimizationMode === 'balanced') {
+      let docs: Document[] | null = null;
+      let query = input.query;
 
-      query = searchResults.query;
-      docs = searchResults.docs;
+      if (this.config.searchWeb) {
+        const searchResults = await this.searchSources(llm, input, emitter);
+
+        query = searchResults.query;
+        docs = searchResults.docs;
+      }
+
+      const sortedDocs = await this.rerankDocs(
+        query,
+        docs ?? [],
+        fileIds,
+        embeddings,
+        optimizationMode,
+      );
+
+      emitter.emit(
+        'data',
+        JSON.stringify({ type: 'sources', data: sortedDocs }),
+      );
+
+      context = this.processDocs(sortedDocs);
+    } else if (optimizationMode === 'quality') {
+      let docs: Document[] = [];
+
+      docs = await this.performDeepResearch(llm, input, emitter);
+
+      emitter.emit('data', JSON.stringify({ type: 'sources', data: docs }));
+
+      context = this.processDocs(docs);
     }
 
-    const sortedDocs = await this.rerankDocs(
-      query,
-      docs ?? [],
-      fileIds,
-      embeddings,
-      optimizationMode,
-    );
-
-    emitter.emit('data', JSON.stringify({ type: 'sources', data: sortedDocs }));
-
-    const context = this.processDocs(sortedDocs);
-
     const formattedChatPrompt = await chatPrompt.invoke({
       query: input.query,
       chat_history: input.chat_history,
       date: new Date().toISOString(),
       context: context,
+      systemInstructions: systemInstructions,
     });
 
     const llmRes = await llm.stream(formattedChatPrompt);
@@ -424,7 +570,9 @@ class MetaSearchAgent implements MetaSearchAgentType {
     return docs
       .map(
         (_, index) =>
-          `${index + 1}. ${docs[index].metadata.title} ${docs[index].pageContent}`,
+          `${index + 1}. ${docs[index].metadata.title} ${
+            docs[index].pageContent
+          }`,
       )
       .join('\n');
   }
@@ -436,6 +584,7 @@ class MetaSearchAgent implements MetaSearchAgentType {
     embeddings: Embeddings,
     optimizationMode: 'speed' | 'balanced' | 'quality',
     fileIds: string[],
+    systemInstructions: string,
   ) {
     const emitter = new eventEmitter();
 
@@ -444,6 +593,7 @@ class MetaSearchAgent implements MetaSearchAgentType {
       fileIds,
       embeddings,
       optimizationMode,
+      systemInstructions,
       {
         chat_history: history,
         query: message,

src/lib/searxng.ts

@@ -8,7 +8,7 @@ interface SearxngSearchOptions {
   pageno?: number;
 }
 
-interface SearxngSearchResult {
+export interface SearxngSearchResult {
   title: string;
   url: string;
   img_src?: string;