feat(types): add llm types

Merge master into canary

package.json

@@ -16,13 +16,14 @@
     "@huggingface/transformers": "^3.7.5",
     "@iarna/toml": "^2.2.5",
     "@icons-pack/react-simple-icons": "^12.3.0",
-    "@langchain/anthropic": "^1.0.0",
-    "@langchain/community": "^1.0.0",
-    "@langchain/core": "^1.0.1",
-    "@langchain/google-genai": "^1.0.0",
-    "@langchain/groq": "^1.0.0",
-    "@langchain/ollama": "^1.0.0",
-    "@langchain/openai": "^1.0.0",
+    "@langchain/anthropic": "^1.0.1",
+    "@langchain/community": "^1.0.3",
+    "@langchain/core": "^1.0.5",
+    "@langchain/google-genai": "^1.0.1",
+    "@langchain/groq": "^1.0.1",
+    "@langchain/langgraph": "^1.0.1",
+    "@langchain/ollama": "^1.0.1",
+    "@langchain/openai": "^1.1.1",
     "@langchain/textsplitters": "^1.0.0",
     "@tailwindcss/typography": "^0.5.12",
     "axios": "^1.8.3",
@@ -33,7 +34,7 @@
     "framer-motion": "^12.23.24",
     "html-to-text": "^9.0.5",
     "jspdf": "^3.0.1",
-    "langchain": "^1.0.1",
+    "langchain": "^1.0.4",
     "lucide-react": "^0.363.0",
     "mammoth": "^1.9.1",
     "markdown-to-jsx": "^7.7.2",
@@ -48,7 +49,7 @@
     "tailwind-merge": "^2.2.2",
     "winston": "^3.17.0",
     "yet-another-react-lightbox": "^3.17.2",
-    "zod": "^3.22.4"
+    "zod": "^4.1.12"
   },
   "devDependencies": {
     "@types/better-sqlite3": "^7.6.12",

src/app/api/images/route.ts

@@ -1,4 +1,4 @@
-import handleImageSearch from '@/lib/chains/imageSearchAgent';
+import searchImages from '@/lib/agents/media/image';
 import ModelRegistry from '@/lib/models/registry';
 import { ModelWithProvider } from '@/lib/models/types';
 import { AIMessage, BaseMessage, HumanMessage } from '@langchain/core/messages';
@@ -13,6 +13,13 @@ export const POST = async (req: Request) => {
   try {
     const body: ImageSearchBody = await req.json();
 
+    const registry = new ModelRegistry();
+
+    const llm = await registry.loadChatModel(
+      body.chatModel.providerId,
+      body.chatModel.key,
+    );
+
     const chatHistory = body.chatHistory
       .map((msg: any) => {
         if (msg.role === 'user') {
@@ -23,16 +30,9 @@ export const POST = async (req: Request) => {
       })
       .filter((msg) => msg !== undefined) as BaseMessage[];
 
-    const registry = new ModelRegistry();
-
-    const llm = await registry.loadChatModel(
-      body.chatModel.providerId,
-      body.chatModel.key,
-    );
-
-    const images = await handleImageSearch(
+    const images = await searchImages(
       {
-        chat_history: chatHistory,
+        chatHistory: chatHistory,
         query: body.query,
       },
       llm,

src/app/api/search/route.ts

@@ -30,12 +30,6 @@ export const POST = async (req: Request) => {
     body.optimizationMode = body.optimizationMode || 'balanced';
     body.stream = body.stream || false;
 
-    const history: BaseMessage[] = body.history.map((msg) => {
-      return msg[0] === 'human'
-        ? new HumanMessage({ content: msg[1] })
-        : new AIMessage({ content: msg[1] });
-    });
-
     const registry = new ModelRegistry();
 
     const [llm, embeddings] = await Promise.all([
@@ -46,6 +40,12 @@ export const POST = async (req: Request) => {
       ),
     ]);
 
+    const history: BaseMessage[] = body.history.map((msg) => {
+      return msg[0] === 'human'
+        ? new HumanMessage({ content: msg[1] })
+        : new AIMessage({ content: msg[1] });
+    });
+
     const searchHandler: MetaSearchAgentType = searchHandlers[body.focusMode];
 
     if (!searchHandler) {
@@ -128,7 +128,7 @@ export const POST = async (req: Request) => {
 
           try {
             controller.close();
-          } catch (error) {}
+          } catch (error) { }
         });
 
         emitter.on('data', (data: string) => {

src/app/api/suggestions/route.ts

@@ -1,7 +1,6 @@
-import generateSuggestions from '@/lib/chains/suggestionGeneratorAgent';
+import generateSuggestions from '@/lib/agents/suggestions';
 import ModelRegistry from '@/lib/models/registry';
 import { ModelWithProvider } from '@/lib/models/types';
-import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 import { AIMessage, BaseMessage, HumanMessage } from '@langchain/core/messages';
 
 interface SuggestionsGenerationBody {
@@ -13,6 +12,13 @@ export const POST = async (req: Request) => {
   try {
     const body: SuggestionsGenerationBody = await req.json();
 
+    const registry = new ModelRegistry();
+
+    const llm = await registry.loadChatModel(
+      body.chatModel.providerId,
+      body.chatModel.key,
+    );
+
     const chatHistory = body.chatHistory
       .map((msg: any) => {
         if (msg.role === 'user') {
@@ -23,16 +29,9 @@ export const POST = async (req: Request) => {
       })
       .filter((msg) => msg !== undefined) as BaseMessage[];
 
-    const registry = new ModelRegistry();
-
-    const llm = await registry.loadChatModel(
-      body.chatModel.providerId,
-      body.chatModel.key,
-    );
-
     const suggestions = await generateSuggestions(
       {
-        chat_history: chatHistory,
+        chatHistory,
       },
       llm,
     );

src/app/api/videos/route.ts

@@ -1,4 +1,4 @@
-import handleVideoSearch from '@/lib/chains/videoSearchAgent';
+import handleVideoSearch from '@/lib/agents/media/video';
 import ModelRegistry from '@/lib/models/registry';
 import { ModelWithProvider } from '@/lib/models/types';
 import { AIMessage, BaseMessage, HumanMessage } from '@langchain/core/messages';
@@ -13,6 +13,13 @@ export const POST = async (req: Request) => {
   try {
     const body: VideoSearchBody = await req.json();
 
+    const registry = new ModelRegistry();
+
+    const llm = await registry.loadChatModel(
+      body.chatModel.providerId,
+      body.chatModel.key,
+    );
+
     const chatHistory = body.chatHistory
       .map((msg: any) => {
         if (msg.role === 'user') {
@@ -23,16 +30,9 @@ export const POST = async (req: Request) => {
       })
       .filter((msg) => msg !== undefined) as BaseMessage[];
 
-    const registry = new ModelRegistry();
-
-    const llm = await registry.loadChatModel(
-      body.chatModel.providerId,
-      body.chatModel.key,
-    );
-
     const videos = await handleVideoSearch(
       {
-        chat_history: chatHistory,
+        chatHistory: chatHistory,
         query: body.query,
       },
       llm,

src/components/ChatWindow.tsx

@@ -8,7 +8,6 @@ import { Settings } from 'lucide-react';
 import Link from 'next/link';
 import NextError from 'next/error';
 import { useChat } from '@/lib/hooks/useChat';
-import Loader from './ui/Loader';
 import SettingsButtonMobile from './Settings/SettingsButtonMobile';
 
 export interface BaseMessage {
@@ -52,7 +51,7 @@ export interface File {
 }
 
 const ChatWindow = () => {
-  const { hasError, isReady, notFound, messages } = useChat();
+  const { hasError, notFound, messages } = useChat();
   if (hasError) {
     return (
       <div className="relative">
@@ -68,24 +67,18 @@ const ChatWindow = () => {
     );
   }
 
-  return isReady ? (
-    notFound ? (
-      <NextError statusCode={404} />
-    ) : (
-      <div>
-        {messages.length > 0 ? (
-          <>
-            <Navbar />
-            <Chat />
-          </>
-        ) : (
-          <EmptyChat />
-        )}
-      </div>
-    )
+  return notFound ? (
+    <NextError statusCode={404} />
   ) : (
-    <div className="flex flex-row items-center justify-center min-h-screen">
-      <Loader />
+    <div>
+      {messages.length > 0 ? (
+        <>
+          <Navbar />
+          <Chat />
+        </>
+      ) : (
+        <EmptyChat />
+      )}
     </div>
   );
 };

src/components/MessageActions/Copy.tsx

@@ -20,9 +20,9 @@ const Copy = ({
         setCopied(true);
         setTimeout(() => setCopied(false), 1000);
       }}
-      className="p-2 text-black/70 dark:text-white/70 rounded-xl hover:bg-light-secondary dark:hover:bg-dark-secondary transition duration-200 hover:text-black dark:hover:text-white"
+      className="p-2 text-black/70 dark:text-white/70 rounded-full hover:bg-light-secondary dark:hover:bg-dark-secondary transition duration-200 hover:text-black dark:hover:text-white"
     >
-      {copied ? <Check size={18} /> : <ClipboardList size={18} />}
+      {copied ? <Check size={16} /> : <ClipboardList size={16} />}
     </button>
   );
 };

src/components/MessageActions/Rewrite.tsx

@@ -1,4 +1,4 @@
-import { ArrowLeftRight } from 'lucide-react';
+import { ArrowLeftRight, Repeat } from 'lucide-react';
 
 const Rewrite = ({
   rewrite,
@@ -10,12 +10,11 @@ const Rewrite = ({
   return (
     <button
       onClick={() => rewrite(messageId)}
-      className="py-2 px-3 text-black/70 dark:text-white/70 rounded-xl hover:bg-light-secondary dark:hover:bg-dark-secondary transition duration-200 hover:text-black dark:hover:text-white flex flex-row items-center space-x-1"
+      className="p-2 text-black/70 dark:text-white/70 rounded-full hover:bg-light-secondary dark:hover:bg-dark-secondary transition duration-200 hover:text-black dark:hover:text-white flex flex-row items-center space-x-1"
     >
-      <ArrowLeftRight size={18} />
-      <p className="text-xs font-medium">Rewrite</p>
+      <Repeat size={16} />
     </button>
   );
 };
-
+1;
 export default Rewrite;

src/components/MessageBox.tsx

@@ -10,6 +10,7 @@ import {
   StopCircle,
   Layers3,
   Plus,
+  CornerDownRight,
 } from 'lucide-react';
 import Markdown, { MarkdownToJSX } from 'markdown-to-jsx';
 import Copy from './MessageActions/Copy';
@@ -122,14 +123,14 @@ const MessageBox = ({
                 </Markdown>
 
                 {loading && isLast ? null : (
-                  <div className="flex flex-row items-center justify-between w-full text-black dark:text-white py-4 -mx-2">
-                    <div className="flex flex-row items-center space-x-1">
+                  <div className="flex flex-row items-center justify-between w-full text-black dark:text-white py-4">
+                    <div className="flex flex-row items-center -ml-2">
                       <Rewrite
                         rewrite={rewrite}
                         messageId={section.assistantMessage.messageId}
                       />
                     </div>
-                    <div className="flex flex-row items-center space-x-1">
+                    <div className="flex flex-row items-center -mr-2">
                       <Copy
                         initialMessage={section.assistantMessage.content}
                         section={section}
@@ -142,12 +143,12 @@ const MessageBox = ({
                             start();
                           }
                         }}
-                        className="p-2 text-black/70 dark:text-white/70 rounded-xl hover:bg-light-secondary dark:hover:bg-dark-secondary transition duration-200 hover:text-black dark:hover:text-white"
+                        className="p-2 text-black/70 dark:text-white/70 rounded-full hover:bg-light-secondary dark:hover:bg-dark-secondary transition duration-200 hover:text-black dark:hover:text-white"
                       >
                         {speechStatus === 'started' ? (
-                          <StopCircle size={18} />
+                          <StopCircle size={16} />
                         ) : (
-                          <Volume2 size={18} />
+                          <Volume2 size={16} />
                         )}
                       </button>
                     </div>
@@ -159,7 +160,7 @@ const MessageBox = ({
                   section.suggestions.length > 0 &&
                   section.assistantMessage &&
                   !loading && (
-                    <div className="mt-8 pt-6 border-t border-light-200/50 dark:border-dark-200/50">
+                    <div className="mt-6">
                       <div className="flex flex-row items-center space-x-2 mb-4">
                         <Layers3
                           className="text-black dark:text-white"
@@ -173,20 +174,24 @@ const MessageBox = ({
                         {section.suggestions.map(
                           (suggestion: string, i: number) => (
                             <div key={i}>
-                              {i > 0 && (
-                                <div className="h-px bg-light-200/40 dark:bg-dark-200/40 mx-3" />
-                              )}
+                              <div className="h-px bg-light-200/40 dark:bg-dark-200/40" />
                               <button
                                 onClick={() => sendMessage(suggestion)}
-                                className="group w-full px-3 py-4 text-left transition-colors duration-200"
+                                className="group w-full py-4 text-left transition-colors duration-200"
                               >
                                 <div className="flex items-center justify-between gap-3">
-                                  <p className="text-sm text-black/70 dark:text-white/70 group-hover:text-[#24A0ED] transition-colors duration-200 leading-relaxed">
-                                    {suggestion}
-                                  </p>
+                                  <div className="flex flex-row space-x-3 items-center ">
+                                    <CornerDownRight
+                                      size={17}
+                                      className="group-hover:text-sky-400 transition-colors duration-200"
+                                    />
+                                    <p className="text-sm text-black/70 dark:text-white/70 group-hover:text-sky-400 transition-colors duration-200 leading-relaxed">
+                                      {suggestion}
+                                    </p>
+                                  </div>
                                   <Plus
                                     size={16}
-                                    className="text-black/40 dark:text-white/40 group-hover:text-[#24A0ED] transition-colors duration-200 flex-shrink-0"
+                                    className="text-black/40 dark:text-white/40 group-hover:text-sky-400 transition-colors duration-200 flex-shrink-0"
                                   />
                                 </div>
                               </button>
@@ -205,11 +210,11 @@ const MessageBox = ({
           <div className="lg:sticky lg:top-20 flex flex-col items-center space-y-3 w-full lg:w-3/12 z-30 h-full pb-4">
             <SearchImages
               query={section.userMessage.content}
-              chatHistory={chatTurns.slice(0, sectionIndex * 2)}
+              chatHistory={chatTurns}
               messageId={section.assistantMessage.messageId}
             />
             <SearchVideos
-              chatHistory={chatTurns.slice(0, sectionIndex * 2)}
+              chatHistory={chatTurns}
               query={section.userMessage.content}
               messageId={section.assistantMessage.messageId}
             />

src/lib/agents/media/image.ts

@@ -0,0 +1,65 @@
+/* I don't think can be classified as agents but to keep the structure consistent i guess ill keep it here */
+
+import {
+  RunnableSequence,
+  RunnableMap,
+  RunnableLambda,
+} from '@langchain/core/runnables';
+import { ChatPromptTemplate } from '@langchain/core/prompts';
+import formatChatHistoryAsString from '@/lib/utils/formatHistory';
+import { BaseMessage, HumanMessage, SystemMessage } from '@langchain/core/messages';
+import { StringOutputParser } from '@langchain/core/output_parsers';
+import { searchSearxng } from '@/lib/searxng';
+import type { BaseChatModel } from '@langchain/core/language_models/chat_models';
+import LineOutputParser from '@/lib/outputParsers/lineOutputParser';
+import { imageSearchFewShots, imageSearchPrompt } from '@/lib/prompts/media/image';
+
+type ImageSearchChainInput = {
+  chatHistory: BaseMessage[];
+  query: string;
+};
+
+type ImageSearchResult = {
+  img_src: string;
+  url: string;
+  title: string;
+}
+
+const outputParser = new LineOutputParser({
+  key: 'query',
+})
+
+const searchImages = async (
+  input: ImageSearchChainInput,
+  llm: BaseChatModel,
+) => {
+  const chatPrompt = await ChatPromptTemplate.fromMessages([
+    new SystemMessage(imageSearchPrompt),
+    ...imageSearchFewShots,
+    new HumanMessage(`<conversation>\n${formatChatHistoryAsString(input.chatHistory)}\n</conversation>\n<follow_up>\n${input.query}\n</follow_up>`)
+  ]).formatMessages({})
+
+  const res = await llm.invoke(chatPrompt)
+
+  const query = await outputParser.invoke(res)
+
+  const searchRes = await searchSearxng(query!, {
+    engines: ['bing images', 'google images'],
+  });
+
+  const images: ImageSearchResult[] = [];
+
+  searchRes.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);
+};
+
+export default searchImages;

src/lib/agents/media/video.ts

@@ -0,0 +1,65 @@
+import { ChatPromptTemplate } from '@langchain/core/prompts';
+import formatChatHistoryAsString from '@/lib/utils/formatHistory';
+import { BaseMessage, HumanMessage, SystemMessage } from '@langchain/core/messages';
+import { searchSearxng } from '@/lib/searxng';
+import type { BaseChatModel } from '@langchain/core/language_models/chat_models';
+import LineOutputParser from '@/lib/outputParsers/lineOutputParser';
+import { videoSearchFewShots, videoSearchPrompt } from '@/lib/prompts/media/videos';
+
+type VideoSearchChainInput = {
+  chatHistory: BaseMessage[];
+  query: string;
+};
+
+type VideoSearchResult = {
+  img_src: string;
+  url: string;
+  title: string;
+  iframe_src: string;
+}
+
+const outputParser = new LineOutputParser({
+  key: 'query',
+});
+
+const searchVideos = async (
+  input: VideoSearchChainInput,
+  llm: BaseChatModel,
+) => {
+  const chatPrompt = await ChatPromptTemplate.fromMessages([
+    new SystemMessage(videoSearchPrompt),
+    ...videoSearchFewShots,
+    new HumanMessage(`<conversation>${formatChatHistoryAsString(input.chatHistory)}\n</conversation>\n<follow_up>\n${input.query}\n</follow_up>`)
+  ]).formatMessages({})
+
+  const res = await llm.invoke(chatPrompt)
+
+  const query = await outputParser.invoke(res)
+
+  const searchRes = await searchSearxng(query!, {
+    engines: ['youtube'],
+  });
+
+  const videos: VideoSearchResult[] = [];
+
+  searchRes.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);
+
+};
+
+export default searchVideos;

src/lib/agents/suggestions/index.ts

@@ -0,0 +1,32 @@
+import ListLineOutputParser from '@/lib/outputParsers/listLineOutputParser';
+import { ChatPromptTemplate, PromptTemplate } from '@langchain/core/prompts';
+import formatChatHistoryAsString from '@/lib/utils/formatHistory';
+import { BaseMessage, HumanMessage, SystemMessage } from '@langchain/core/messages';
+import { BaseChatModel } from '@langchain/core/language_models/chat_models';
+import { suggestionGeneratorPrompt } from '@/lib/prompts/suggestions';
+
+type SuggestionGeneratorInput = {
+  chatHistory: BaseMessage[];
+};
+
+const outputParser = new ListLineOutputParser({
+  key: 'suggestions',
+});
+
+const generateSuggestions = async (
+  input: SuggestionGeneratorInput,
+  llm: BaseChatModel,
+) => {
+  const chatPrompt = await ChatPromptTemplate.fromMessages([
+    new SystemMessage(suggestionGeneratorPrompt),
+    new HumanMessage(`<conversation>${formatChatHistoryAsString(input.chatHistory)}</conversation>`)
+  ]).formatMessages({})
+
+  const res = await llm.invoke(chatPrompt)
+
+  const suggestions = await outputParser.invoke(res)
+
+  return suggestions
+};
+
+export default generateSuggestions;

src/lib/chains/imageSearchAgent.ts

@@ -1,105 +0,0 @@
-import {
-  RunnableSequence,
-  RunnableMap,
-  RunnableLambda,
-} from '@langchain/core/runnables';
-import { ChatPromptTemplate } 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';
-import LineOutputParser from '../outputParsers/lineOutputParser';
-
-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.
-Output only the rephrased query wrapped in an XML <query> element. Do not include any explanation or additional text.
-`;
-
-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;
-      },
-    }),
-    ChatPromptTemplate.fromMessages([
-      ['system', imageSearchChainPrompt],
-      [
-        'user',
-        '<conversation>\n</conversation>\n<follow_up>\nWhat is a cat?\n</follow_up>',
-      ],
-      ['assistant', '<query>A cat</query>'],
-
-      [
-        'user',
-        '<conversation>\n</conversation>\n<follow_up>\nWhat is a car? How does it work?\n</follow_up>',
-      ],
-      ['assistant', '<query>Car working</query>'],
-      [
-        'user',
-        '<conversation>\n</conversation>\n<follow_up>\nHow does an AC work?\n</follow_up>',
-      ],
-      ['assistant', '<query>AC working</query>'],
-      [
-        'user',
-        '<conversation>{chat_history}</conversation>\n<follow_up>\n{query}\n</follow_up>',
-      ],
-    ]),
-    llm,
-    strParser,
-    RunnableLambda.from(async (input: string) => {
-      const queryParser = new LineOutputParser({
-        key: 'query',
-      });
-
-      return await queryParser.parse(input);
-    }),
-    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;

src/lib/chains/suggestionGeneratorAgent.ts

@@ -1,55 +0,0 @@
-import { RunnableSequence, RunnableMap } from '@langchain/core/runnables';
-import ListLineOutputParser from '../outputParsers/listLineOutputParser';
-import { PromptTemplate } from '@langchain/core/prompts';
-import formatChatHistoryAsString from '../utils/formatHistory';
-import { BaseMessage } from '@langchain/core/messages';
-import { BaseChatModel } from '@langchain/core/language_models/chat_models';
-import { ChatOpenAI } from '@langchain/openai';
-
-const suggestionGeneratorPrompt = `
-You are an AI suggestion generator for an AI powered search engine. You will be given a conversation below. You need to generate 4-5 suggestions based on the conversation. The suggestion should be relevant to the conversation that can be used by the user to ask the chat model for more information.
-You need to make sure the suggestions are relevant to the conversation and are helpful to the user. Keep a note that the user might use these suggestions to ask a chat model for more information. 
-Make sure the suggestions are medium in length and are informative and relevant to the conversation.
-
-Provide these suggestions separated by newlines between the XML tags <suggestions> and </suggestions>. For example:
-
-<suggestions>
-Tell me more about SpaceX and their recent projects
-What is the latest news on SpaceX?
-Who is the CEO of SpaceX?
-</suggestions>
-
-Conversation:
-{chat_history}
-`;
-
-type SuggestionGeneratorInput = {
-  chat_history: BaseMessage[];
-};
-
-const outputParser = new ListLineOutputParser({
-  key: 'suggestions',
-});
-
-const createSuggestionGeneratorChain = (llm: BaseChatModel) => {
-  return RunnableSequence.from([
-    RunnableMap.from({
-      chat_history: (input: SuggestionGeneratorInput) =>
-        formatChatHistoryAsString(input.chat_history),
-    }),
-    PromptTemplate.fromTemplate(suggestionGeneratorPrompt),
-    llm,
-    outputParser,
-  ]);
-};
-
-const generateSuggestions = (
-  input: SuggestionGeneratorInput,
-  llm: BaseChatModel,
-) => {
-  (llm as unknown as ChatOpenAI).temperature = 0;
-  const suggestionGeneratorChain = createSuggestionGeneratorChain(llm);
-  return suggestionGeneratorChain.invoke(input);
-};
-
-export default generateSuggestions;

src/lib/chains/videoSearchAgent.ts

@@ -1,110 +0,0 @@
-import {
-  RunnableSequence,
-  RunnableMap,
-  RunnableLambda,
-} from '@langchain/core/runnables';
-import { ChatPromptTemplate } 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';
-import LineOutputParser from '../outputParsers/lineOutputParser';
-
-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.
-Output only the rephrased query wrapped in an XML <query> element. Do not include any explanation or additional text.
-`;
-
-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;
-      },
-    }),
-    ChatPromptTemplate.fromMessages([
-      ['system', videoSearchChainPrompt],
-      [
-        'user',
-        '<conversation>\n</conversation>\n<follow_up>\nHow does a car work?\n</follow_up>',
-      ],
-      ['assistant', '<query>How does a car work?</query>'],
-      [
-        'user',
-        '<conversation>\n</conversation>\n<follow_up>\nWhat is the theory of relativity?\n</follow_up>',
-      ],
-      ['assistant', '<query>Theory of relativity</query>'],
-      [
-        'user',
-        '<conversation>\n</conversation>\n<follow_up>\nHow does an AC work?\n</follow_up>',
-      ],
-      ['assistant', '<query>AC working</query>'],
-      [
-        'user',
-        '<conversation>{chat_history}</conversation>\n<follow_up>\n{query}\n</follow_up>',
-      ],
-    ]),
-    llm,
-    strParser,
-    RunnableLambda.from(async (input: string) => {
-      const queryParser = new LineOutputParser({
-        key: 'query',
-      });
-      return await queryParser.parse(input);
-    }),
-    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;

src/lib/models/base/embedding.ts

@@ -0,0 +1,7 @@
+abstract class BaseEmbedding<CONFIG> {
+  constructor(protected config: CONFIG) {}
+  abstract embedText(texts: string[]): Promise<number[][]>;
+  abstract embedChunks(chunks: Chunk[]): Promise<number[][]>;
+}
+
+export default BaseEmbedding;

src/lib/models/base/llm.ts

@@ -0,0 +1,26 @@
+import {
+  GenerateObjectInput,
+  GenerateObjectOutput,
+  GenerateOptions,
+  GenerateTextInput,
+  GenerateTextOutput,
+  StreamObjectOutput,
+  StreamTextOutput,
+} from '../types';
+
+abstract class BaseLLM<CONFIG> {
+  constructor(protected config: CONFIG) {}
+  abstract withOptions(options: GenerateOptions): this;
+  abstract generateText(input: GenerateTextInput): Promise<GenerateTextOutput>;
+  abstract streamText(
+    input: GenerateTextInput,
+  ): AsyncGenerator<StreamTextOutput>;
+  abstract generateObject<T>(
+    input: GenerateObjectInput,
+  ): Promise<GenerateObjectOutput<T>>;
+  abstract streamObject<T>(
+    input: GenerateObjectInput,
+  ): AsyncGenerator<StreamObjectOutput<T>>;
+}
+
+export default BaseLLM;

src/lib/models/base/provider.ts

@@ -1,7 +1,9 @@
 import { Embeddings } from '@langchain/core/embeddings';
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
-import { Model, ModelList, ProviderMetadata } from '../types';
+import { ModelList, ProviderMetadata } from '../types';
 import { UIConfigField } from '@/lib/config/types';
+import BaseLLM from './llm';
+import BaseEmbedding from './embedding';
 
 abstract class BaseModelProvider<CONFIG> {
   constructor(
@@ -11,8 +13,8 @@ abstract class BaseModelProvider<CONFIG> {
   ) {}
   abstract getDefaultModels(): Promise<ModelList>;
   abstract getModelList(): Promise<ModelList>;
-  abstract loadChatModel(modelName: string): Promise<BaseChatModel>;
-  abstract loadEmbeddingModel(modelName: string): Promise<Embeddings>;
+  abstract loadChatModel(modelName: string): Promise<BaseLLM<any>>;
+  abstract loadEmbeddingModel(modelName: string): Promise<BaseEmbedding<any>>;
   static getProviderConfigFields(): UIConfigField[] {
     throw new Error('Method not implemented.');
   }

src/lib/models/providers/index.ts

@@ -1,27 +1,11 @@
 import { ModelProviderUISection } from '@/lib/config/types';
-import { ProviderConstructor } from './baseProvider';
+import { ProviderConstructor } from '../base/provider';
 import OpenAIProvider from './openai';
 import OllamaProvider from './ollama';
-import TransformersProvider from './transformers';
-import AnthropicProvider from './anthropic';
-import GeminiProvider from './gemini';
-import GroqProvider from './groq';
-import DeepSeekProvider from './deepseek';
-import LMStudioProvider from './lmstudio';
-import LemonadeProvider from './lemonade';
-import AimlProvider from '@/lib/models/providers/aiml';
 
 export const providers: Record<string, ProviderConstructor<any>> = {
   openai: OpenAIProvider,
   ollama: OllamaProvider,
-  transformers: TransformersProvider,
-  anthropic: AnthropicProvider,
-  gemini: GeminiProvider,
-  groq: GroqProvider,
-  deepseek: DeepSeekProvider,
-  aiml: AimlProvider,
-  lmstudio: LMStudioProvider,
-  lemonade: LemonadeProvider,
 };
 
 export const getModelProvidersUIConfigSection =

src/lib/models/providers/ollama/index.ts

@@ -1,10 +1,11 @@
-import { BaseChatModel } from '@langchain/core/language_models/chat_models';
-import { Model, ModelList, ProviderMetadata } from '../types';
-import BaseModelProvider from './baseProvider';
-import { ChatOllama, OllamaEmbeddings } from '@langchain/ollama';
-import { Embeddings } from '@langchain/core/embeddings';
 import { UIConfigField } from '@/lib/config/types';
 import { getConfiguredModelProviderById } from '@/lib/config/serverRegistry';
+import BaseModelProvider from '../../base/provider';
+import { Model, ModelList, ProviderMetadata } from '../../types';
+import BaseLLM from '../../base/llm';
+import BaseEmbedding from '../../base/embedding';
+import OllamaLLM from './ollamaLLM';
+import OllamaEmbedding from './ollamaEmbedding';
 
 interface OllamaConfig {
   baseURL: string;
@@ -76,7 +77,7 @@ class OllamaProvider extends BaseModelProvider<OllamaConfig> {
     };
   }
 
-  async loadChatModel(key: string): Promise<BaseChatModel> {
+  async loadChatModel(key: string): Promise<BaseLLM<any>> {
     const modelList = await this.getModelList();
 
     const exists = modelList.chat.find((m) => m.key === key);
@@ -87,14 +88,13 @@ class OllamaProvider extends BaseModelProvider<OllamaConfig> {
       );
     }
 
-    return new ChatOllama({
-      temperature: 0.7,
+    return new OllamaLLM({
+      baseURL: this.config.baseURL,
       model: key,
-      baseUrl: this.config.baseURL,
     });
   }
 
-  async loadEmbeddingModel(key: string): Promise<Embeddings> {
+  async loadEmbeddingModel(key: string): Promise<BaseEmbedding<any>> {
     const modelList = await this.getModelList();
     const exists = modelList.embedding.find((m) => m.key === key);
 
@@ -104,9 +104,9 @@ class OllamaProvider extends BaseModelProvider<OllamaConfig> {
       );
     }
 
-    return new OllamaEmbeddings({
+    return new OllamaEmbedding({
       model: key,
-      baseUrl: this.config.baseURL,
+      baseURL: this.config.baseURL,
     });
   }
 

src/lib/models/providers/ollama/ollamaEmbedding.ts

@@ -0,0 +1,39 @@
+import { Ollama } from 'ollama';
+import BaseEmbedding from '../../base/embedding';
+
+type OllamaConfig = {
+  model: string;
+  baseURL?: string;
+};
+
+class OllamaEmbedding extends BaseEmbedding<OllamaConfig> {
+  ollamaClient: Ollama;
+
+  constructor(protected config: OllamaConfig) {
+    super(config);
+
+    this.ollamaClient = new Ollama({
+      host: this.config.baseURL || 'http://localhost:11434',
+    });
+  }
+
+  async embedText(texts: string[]): Promise<number[][]> {
+    const response = await this.ollamaClient.embed({
+      input: texts,
+      model: this.config.model,
+    });
+
+    return response.embeddings;
+  }
+
+  async embedChunks(chunks: Chunk[]): Promise<number[][]> {
+    const response = await this.ollamaClient.embed({
+      input: chunks.map((c) => c.content),
+      model: this.config.model,
+    });
+
+    return response.embeddings;
+  }
+}
+
+export default OllamaEmbedding;

src/lib/models/providers/ollama/ollamaLLM.ts

@@ -0,0 +1,149 @@
+import z from 'zod';
+import BaseLLM from '../../base/llm';
+import {
+  GenerateObjectInput,
+  GenerateOptions,
+  GenerateTextInput,
+  GenerateTextOutput,
+  StreamTextOutput,
+} from '../../types';
+import { Ollama } from 'ollama';
+import { parse } from 'partial-json';
+
+type OllamaConfig = {
+  baseURL: string;
+  model: string;
+  options?: GenerateOptions;
+};
+
+class OllamaLLM extends BaseLLM<OllamaConfig> {
+  ollamaClient: Ollama;
+
+  constructor(protected config: OllamaConfig) {
+    super(config);
+
+    this.ollamaClient = new Ollama({
+      host: this.config.baseURL || 'http://localhost:11434',
+    });
+  }
+
+  withOptions(options: GenerateOptions) {
+    this.config.options = {
+      ...this.config.options,
+      ...options,
+    };
+    return this;
+  }
+
+  async generateText(input: GenerateTextInput): Promise<GenerateTextOutput> {
+    this.withOptions(input.options || {});
+
+    const res = await this.ollamaClient.chat({
+      model: this.config.model,
+      messages: input.messages,
+      options: {
+        top_p: this.config.options?.topP,
+        temperature: this.config.options?.temperature,
+        num_predict: this.config.options?.maxTokens,
+        frequency_penalty: this.config.options?.frequencyPenalty,
+        presence_penalty: this.config.options?.presencePenalty,
+        stop: this.config.options?.stopSequences,
+      },
+    });
+
+    return {
+      content: res.message.content,
+      additionalInfo: {
+        reasoning: res.message.thinking,
+      },
+    };
+  }
+
+  async *streamText(
+    input: GenerateTextInput,
+  ): AsyncGenerator<StreamTextOutput> {
+    this.withOptions(input.options || {});
+
+    const stream = await this.ollamaClient.chat({
+      model: this.config.model,
+      messages: input.messages,
+      stream: true,
+      options: {
+        top_p: this.config.options?.topP,
+        temperature: this.config.options?.temperature,
+        num_predict: this.config.options?.maxTokens,
+        frequency_penalty: this.config.options?.frequencyPenalty,
+        presence_penalty: this.config.options?.presencePenalty,
+        stop: this.config.options?.stopSequences,
+      },
+    });
+
+    for await (const chunk of stream) {
+      yield {
+        contentChunk: chunk.message.content,
+        done: chunk.done,
+        additionalInfo: {
+          reasoning: chunk.message.thinking,
+        },
+      };
+    }
+  }
+
+  async generateObject<T>(input: GenerateObjectInput): Promise<T> {
+    this.withOptions(input.options || {});
+
+    const response = await this.ollamaClient.chat({
+      model: this.config.model,
+      messages: input.messages,
+      format: z.toJSONSchema(input.schema),
+      options: {
+        top_p: this.config.options?.topP,
+        temperature: this.config.options?.temperature,
+        num_predict: this.config.options?.maxTokens,
+        frequency_penalty: this.config.options?.frequencyPenalty,
+        presence_penalty: this.config.options?.presencePenalty,
+        stop: this.config.options?.stopSequences,
+      },
+    });
+
+    try {
+      return input.schema.parse(JSON.parse(response.message.content)) as T;
+    } catch (err) {
+      throw new Error(`Error parsing response from Ollama: ${err}`);
+    }
+  }
+
+  async *streamObject<T>(input: GenerateObjectInput): AsyncGenerator<T> {
+    let recievedObj: string = '';
+
+    this.withOptions(input.options || {});
+
+    const stream = await this.ollamaClient.chat({
+      model: this.config.model,
+      messages: input.messages,
+      format: z.toJSONSchema(input.schema),
+      stream: true,
+      options: {
+        top_p: this.config.options?.topP,
+        temperature: this.config.options?.temperature,
+        num_predict: this.config.options?.maxTokens,
+        frequency_penalty: this.config.options?.frequencyPenalty,
+        presence_penalty: this.config.options?.presencePenalty,
+        stop: this.config.options?.stopSequences,
+      },
+    });
+
+    for await (const chunk of stream) {
+      recievedObj += chunk.message.content;
+
+      try {
+        yield parse(recievedObj) as T;
+      } catch (err) {
+        console.log('Error parsing partial object from Ollama:', err);
+        yield {} as T;
+      }
+    }
+  }
+}
+
+export default OllamaLLM;

src/lib/models/providers/openai/index.ts

@@ -1,10 +1,13 @@
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
-import { Model, ModelList, ProviderMetadata } from '../types';
-import BaseModelProvider from './baseProvider';
-import { ChatOpenAI, OpenAIEmbeddings } from '@langchain/openai';
 import { Embeddings } from '@langchain/core/embeddings';
 import { UIConfigField } from '@/lib/config/types';
 import { getConfiguredModelProviderById } from '@/lib/config/serverRegistry';
+import { Model, ModelList, ProviderMetadata } from '../../types';
+import OpenAIEmbedding from './openaiEmbedding';
+import BaseEmbedding from '../../base/embedding';
+import BaseModelProvider from '../../base/provider';
+import BaseLLM from '../../base/llm';
+import OpenAILLM from './openaiLLM';
 
 interface OpenAIConfig {
   apiKey: string;
@@ -145,7 +148,7 @@ class OpenAIProvider extends BaseModelProvider<OpenAIConfig> {
     };
   }
 
-  async loadChatModel(key: string): Promise<BaseChatModel> {
+  async loadChatModel(key: string): Promise<BaseLLM<any>> {
     const modelList = await this.getModelList();
 
     const exists = modelList.chat.find((m) => m.key === key);
@@ -156,17 +159,14 @@ class OpenAIProvider extends BaseModelProvider<OpenAIConfig> {
       );
     }
 
-    return new ChatOpenAI({
+    return new OpenAILLM({
       apiKey: this.config.apiKey,
-      temperature: 0.7,
       model: key,
-      configuration: {
-        baseURL: this.config.baseURL,
-      },
+      baseURL: this.config.baseURL,
     });
   }
 
-  async loadEmbeddingModel(key: string): Promise<Embeddings> {
+  async loadEmbeddingModel(key: string): Promise<BaseEmbedding<any>> {
     const modelList = await this.getModelList();
     const exists = modelList.embedding.find((m) => m.key === key);
 
@@ -176,12 +176,10 @@ class OpenAIProvider extends BaseModelProvider<OpenAIConfig> {
       );
     }
 
-    return new OpenAIEmbeddings({
+    return new OpenAIEmbedding({
       apiKey: this.config.apiKey,
       model: key,
-      configuration: {
-        baseURL: this.config.baseURL,
-      },
+      baseURL: this.config.baseURL,
     });
   }
 

src/lib/models/providers/openai/openaiEmbedding.ts

@@ -0,0 +1,41 @@
+import OpenAI from 'openai';
+import BaseEmbedding from '../../base/embedding';
+
+type OpenAIConfig = {
+  apiKey: string;
+  model: string;
+  baseURL?: string;
+};
+
+class OpenAIEmbedding extends BaseEmbedding<OpenAIConfig> {
+  openAIClient: OpenAI;
+
+  constructor(protected config: OpenAIConfig) {
+    super(config);
+
+    this.openAIClient = new OpenAI({
+      apiKey: config.apiKey,
+      baseURL: config.baseURL,
+    });
+  }
+
+  async embedText(texts: string[]): Promise<number[][]> {
+    const response = await this.openAIClient.embeddings.create({
+      model: this.config.model,
+      input: texts,
+    });
+
+    return response.data.map((embedding) => embedding.embedding);
+  }
+
+  async embedChunks(chunks: Chunk[]): Promise<number[][]> {
+    const response = await this.openAIClient.embeddings.create({
+      model: this.config.model,
+      input: chunks.map((c) => c.content),
+    });
+
+    return response.data.map((embedding) => embedding.embedding);
+  }
+}
+
+export default OpenAIEmbedding;

src/lib/models/providers/openai/openaiLLM.ts

@@ -0,0 +1,163 @@
+import OpenAI from 'openai';
+import BaseLLM from '../../base/llm';
+import { zodTextFormat, zodResponseFormat } from 'openai/helpers/zod';
+import {
+  GenerateObjectInput,
+  GenerateOptions,
+  GenerateTextInput,
+  GenerateTextOutput,
+  StreamTextOutput,
+} from '../../types';
+import { parse } from 'partial-json';
+
+type OpenAIConfig = {
+  apiKey: string;
+  model: string;
+  baseURL?: string;
+  options?: GenerateOptions;
+};
+
+class OpenAILLM extends BaseLLM<OpenAIConfig> {
+  openAIClient: OpenAI;
+
+  constructor(protected config: OpenAIConfig) {
+    super(config);
+
+    this.openAIClient = new OpenAI({
+      apiKey: this.config.apiKey,
+      baseURL: this.config.baseURL || 'https://api.openai.com/v1',
+    });
+  }
+
+  withOptions(options: GenerateOptions) {
+    this.config.options = {
+      ...this.config.options,
+      ...options,
+    };
+
+    return this;
+  }
+
+  async generateText(input: GenerateTextInput): Promise<GenerateTextOutput> {
+    this.withOptions(input.options || {});
+
+    const response = await this.openAIClient.chat.completions.create({
+      model: this.config.model,
+      messages: input.messages,
+      temperature: this.config.options?.temperature || 1.0,
+      top_p: this.config.options?.topP,
+      max_completion_tokens: this.config.options?.maxTokens,
+      stop: this.config.options?.stopSequences,
+      frequency_penalty: this.config.options?.frequencyPenalty,
+      presence_penalty: this.config.options?.presencePenalty,
+    });
+
+    if (response.choices && response.choices.length > 0) {
+      return {
+        content: response.choices[0].message.content!,
+        additionalInfo: {
+          finishReason: response.choices[0].finish_reason,
+        },
+      };
+    }
+
+    throw new Error('No response from OpenAI');
+  }
+
+  async *streamText(
+    input: GenerateTextInput,
+  ): AsyncGenerator<StreamTextOutput> {
+    this.withOptions(input.options || {});
+
+    const stream = await this.openAIClient.chat.completions.create({
+      model: this.config.model,
+      messages: input.messages,
+      temperature: this.config.options?.temperature || 1.0,
+      top_p: this.config.options?.topP,
+      max_completion_tokens: this.config.options?.maxTokens,
+      stop: this.config.options?.stopSequences,
+      frequency_penalty: this.config.options?.frequencyPenalty,
+      presence_penalty: this.config.options?.presencePenalty,
+      stream: true,
+    });
+
+    for await (const chunk of stream) {
+      if (chunk.choices && chunk.choices.length > 0) {
+        yield {
+          contentChunk: chunk.choices[0].delta.content || '',
+          done: chunk.choices[0].finish_reason !== null,
+          additionalInfo: {
+            finishReason: chunk.choices[0].finish_reason,
+          },
+        };
+      }
+    }
+  }
+
+  async generateObject<T>(input: GenerateObjectInput): Promise<T> {
+    this.withOptions(input.options || {});
+
+    const response = await this.openAIClient.chat.completions.parse({
+      messages: input.messages,
+      model: this.config.model,
+      temperature: this.config.options?.temperature || 1.0,
+      top_p: this.config.options?.topP,
+      max_completion_tokens: this.config.options?.maxTokens,
+      stop: this.config.options?.stopSequences,
+      frequency_penalty: this.config.options?.frequencyPenalty,
+      presence_penalty: this.config.options?.presencePenalty,
+      response_format: zodResponseFormat(input.schema, 'object'),
+    });
+
+    if (response.choices && response.choices.length > 0) {
+      try {
+        return input.schema.parse(response.choices[0].message.parsed) as T;
+      } catch (err) {
+        throw new Error(`Error parsing response from OpenAI: ${err}`);
+      }
+    }
+
+    throw new Error('No response from OpenAI');
+  }
+
+  async *streamObject<T>(input: GenerateObjectInput): AsyncGenerator<T> {
+    let recievedObj: string = '';
+
+    this.withOptions(input.options || {});
+
+    const stream = this.openAIClient.responses.stream({
+      model: this.config.model,
+      input: input.messages,
+      temperature: this.config.options?.temperature || 1.0,
+      top_p: this.config.options?.topP,
+      max_completion_tokens: this.config.options?.maxTokens,
+      stop: this.config.options?.stopSequences,
+      frequency_penalty: this.config.options?.frequencyPenalty,
+      presence_penalty: this.config.options?.presencePenalty,
+      text: {
+        format: zodTextFormat(input.schema, 'object'),
+      },
+    });
+
+    for await (const chunk of stream) {
+      if (chunk.type === 'response.output_text.delta' && chunk.delta) {
+        recievedObj += chunk.delta;
+
+        try {
+          yield parse(recievedObj) as T;
+        } catch (err) {
+          console.log('Error parsing partial object from OpenAI:', err);
+          yield {} as T;
+        }
+      } else if (chunk.type === 'response.output_text.done' && chunk.text) {
+        try {
+          yield parse(chunk.text) as T;
+        } catch (err) {
+          throw new Error(`Error parsing response from OpenAI: ${err}`);
+        }
+      }
+    }
+  }
+}
+
+export default OpenAILLM;

src/lib/models/types.ts

@@ -1,3 +1,5 @@
+import z from 'zod';
+
 type Model = {
   name: string;
   key: string;
@@ -25,10 +27,59 @@ type ModelWithProvider = {
   providerId: string;
 };
 
+type GenerateOptions = {
+  temperature?: number;
+  maxTokens?: number;
+  topP?: number;
+  stopSequences?: string[];
+  frequencyPenalty?: number;
+  presencePenalty?: number;
+};
+
+type GenerateTextInput = {
+  messages: Message[];
+  options?: GenerateOptions;
+};
+
+type GenerateTextOutput = {
+  content: string;
+  additionalInfo?: Record<string, any>;
+};
+
+type StreamTextOutput = {
+  contentChunk: string;
+  additionalInfo?: Record<string, any>;
+  done?: boolean;
+};
+
+type GenerateObjectInput = {
+  schema: z.ZodTypeAny;
+  messages: Message[];
+  options?: GenerateOptions;
+};
+
+type GenerateObjectOutput<T> = {
+  object: T;
+  additionalInfo?: Record<string, any>;
+};
+
+type StreamObjectOutput<T> = {
+  objectChunk: Partial<T>;
+  additionalInfo?: Record<string, any>;
+  done?: boolean;
+};
+
 export type {
   Model,
   ModelList,
   ProviderMetadata,
   MinimalProvider,
   ModelWithProvider,
+  GenerateOptions,
+  GenerateTextInput,
+  GenerateTextOutput,
+  StreamTextOutput,
+  GenerateObjectInput,
+  GenerateObjectOutput,
+  StreamObjectOutput,
 };

src/lib/prompts/media/image.ts

@@ -0,0 +1,26 @@
+import { BaseMessageLike } from "@langchain/core/messages";
+
+export const imageSearchPrompt = `
+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.
+Output only the rephrased query wrapped in an XML <query> element. Do not include any explanation or additional text.
+`;
+
+export const imageSearchFewShots: BaseMessageLike[] = [
+    [
+        'user',
+        '<conversation>\n</conversation>\n<follow_up>\nWhat is a cat?\n</follow_up>',
+    ],
+    ['assistant', '<query>A cat</query>'],
+
+    [
+        'user',
+        '<conversation>\n</conversation>\n<follow_up>\nWhat is a car? How does it work?\n</follow_up>',
+    ],
+    ['assistant', '<query>Car working</query>'],
+    [
+        'user',
+        '<conversation>\n</conversation>\n<follow_up>\nHow does an AC work?\n</follow_up>',
+    ],
+    ['assistant', '<query>AC working</query>']
+]

src/lib/prompts/media/videos.ts

@@ -0,0 +1,25 @@
+import { BaseMessageLike } from "@langchain/core/messages";
+
+export const videoSearchPrompt = `
+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.
+Output only the rephrased query wrapped in an XML <query> element. Do not include any explanation or additional text.
+`;
+
+export const videoSearchFewShots: BaseMessageLike[] = [
+    [
+        'user',
+        '<conversation>\n</conversation>\n<follow_up>\nHow does a car work?\n</follow_up>',
+    ],
+    ['assistant', '<query>How does a car work?</query>'],
+    [
+        'user',
+        '<conversation>\n</conversation>\n<follow_up>\nWhat is the theory of relativity?\n</follow_up>',
+    ],
+    ['assistant', '<query>Theory of relativity</query>'],
+    [
+        'user',
+        '<conversation>\n</conversation>\n<follow_up>\nHow does an AC work?\n</follow_up>',
+    ],
+    ['assistant', '<query>AC working</query>'],
+]

src/lib/prompts/suggestions/index.ts

@@ -0,0 +1,15 @@
+export const suggestionGeneratorPrompt = `
+You are an AI suggestion generator for an AI powered search engine. You will be given a conversation below. You need to generate 4-5 suggestions based on the conversation. The suggestion should be relevant to the conversation that can be used by the user to ask the chat model for more information.
+You need to make sure the suggestions are relevant to the conversation and are helpful to the user. Keep a note that the user might use these suggestions to ask a chat model for more information. 
+Make sure the suggestions are medium in length and are informative and relevant to the conversation.
+
+Provide these suggestions separated by newlines between the XML tags <suggestions> and </suggestions>. For example:
+
+<suggestions>
+Tell me more about SpaceX and their recent projects
+What is the latest news on SpaceX?
+Who is the CEO of SpaceX?
+</suggestions>
+
+Today's date is ${new Date().toISOString()}
+`;

src/lib/types.ts

@@ -0,0 +1,9 @@
+type Message = {
+  role: 'user' | 'assistant' | 'system';
+  content: string;
+};
+
+type Chunk = {
+  content: string;
+  metadata: Record<string, any>;
+};

yarn.lock

@@ -746,19 +746,19 @@
     "@jridgewell/resolve-uri" "^3.1.0"
     "@jridgewell/sourcemap-codec" "^1.4.14"
 
-"@langchain/anthropic@^1.0.0":
-  version "1.0.0"
-  resolved "https://registry.yarnpkg.com/@langchain/anthropic/-/anthropic-1.0.0.tgz#48535c5682851bf8fddcf37aa7ca78d4d93da932"
-  integrity sha512-Lud/FrkFmXMYW5R9y0FC+RGdgjBBVQ2JAnG3A8E1I4+sqv5JgJttw3HKRpFkyBUSyacs6LMfSn5dbJ6TT9nMiQ==
+"@langchain/anthropic@^1.0.1":
+  version "1.0.1"
+  resolved "https://registry.yarnpkg.com/@langchain/anthropic/-/anthropic-1.0.1.tgz#a9f836b11ecbce282fc2afb8d707c52fd37711c4"
+  integrity sha512-yVKePAT+nNHtybyyPlWqiq6lqcoDlIuMgL9B4WMEU5gbmzL170iodiqcgcZNFQLOC1V2wCOzywq6Zr0kB24AFg==
   dependencies:
     "@anthropic-ai/sdk" "^0.65.0"
 
-"@langchain/classic@1.0.0":
-  version "1.0.0"
-  resolved "https://registry.yarnpkg.com/@langchain/classic/-/classic-1.0.0.tgz#acbc15eebba03499cf93e73d2c93703a3da0a46e"
-  integrity sha512-darZFvO5g5e3TqZ4rvZ938F94D4a34v2ZdWfyipmyu7WB4RXMshmYtWCp98o4ec3bfRD9S4+oHMmaPcnk5cs5A==
+"@langchain/classic@1.0.3":
+  version "1.0.3"
+  resolved "https://registry.yarnpkg.com/@langchain/classic/-/classic-1.0.3.tgz#92482cb4cb8692407b4ecde0df312f035934472f"
+  integrity sha512-XyoaiJSi4y7SzrZMCb3DdDfC+M3gqIQpVH2cOCh9xQf4244jNrncpLXF/MwOJYWxzTsjfcCAHIbFJ0kSH5nqmg==
   dependencies:
-    "@langchain/openai" "1.0.0-alpha.3"
+    "@langchain/openai" "1.1.1"
     "@langchain/textsplitters" "1.0.0"
     handlebars "^4.7.8"
     js-yaml "^4.1.0"
@@ -771,24 +771,24 @@
   optionalDependencies:
     langsmith "^0.3.64"
 
-"@langchain/community@^1.0.0":
-  version "1.0.0"
-  resolved "https://registry.yarnpkg.com/@langchain/community/-/community-1.0.0.tgz#8e587605b7c981882e20281aa9e644a166620145"
-  integrity sha512-CM4vUZHaFHq8HpWBMIWPO5bo/rmRPJ1/iaJk7s8CghkkQ0WLaZzDtoG/wJKJZMDJOUVCtZKTw+TytlGu00/9dg==
+"@langchain/community@^1.0.3":
+  version "1.0.3"
+  resolved "https://registry.yarnpkg.com/@langchain/community/-/community-1.0.3.tgz#278c82eee22ff37b120e182b07b7c23ffc6786ab"
+  integrity sha512-86L7qooSY8Fh5Sf2Tu/X8PvDJqvEXohyZUGusuv0XtnWGivwtecBm0vEbVPkLF07I2ZMtyAGzHJOblbveq6Nmg==
   dependencies:
-    "@langchain/classic" "1.0.0"
-    "@langchain/openai" "1.0.0"
+    "@langchain/classic" "1.0.3"
+    "@langchain/openai" "1.1.1"
     binary-extensions "^2.2.0"
-    expr-eval "^2.0.2"
     flat "^5.0.2"
     js-yaml "^4.1.0"
+    math-expression-evaluator "^2.0.0"
     uuid "^10.0.0"
     zod "^3.25.76 || ^4"
 
-"@langchain/core@^1.0.1":
-  version "1.0.1"
-  resolved "https://registry.yarnpkg.com/@langchain/core/-/core-1.0.1.tgz#c2bdbdff87649fe17b2c86bf535d749ac73a586c"
-  integrity sha512-hVM3EkojYOk4ISJQKjLuWYSH6kyyOFlZIrLFETDA1L0Z2/Iu0q32aJawZ0FDn6rlXE8QZjBt/9OaOL36rXc05w==
+"@langchain/core@^1.0.5":
+  version "1.0.5"
+  resolved "https://registry.yarnpkg.com/@langchain/core/-/core-1.0.5.tgz#1e20ecce80fa4d0b979ea05b24b879b8357d8092"
+  integrity sha512-9Hy/b9+j+mm0Bhnm8xD9B0KpBYTidroLrDHdbrHoMC2DqXoY2umvi1M3M/9D744qsMSaIMP0ZwFcy5YbqI/dGw==
   dependencies:
     "@cfworker/json-schema" "^4.0.2"
     ansi-styles "^5.0.0"
@@ -802,18 +802,18 @@
     uuid "^10.0.0"
     zod "^3.25.76 || ^4"
 
-"@langchain/google-genai@^1.0.0":
-  version "1.0.0"
-  resolved "https://registry.yarnpkg.com/@langchain/google-genai/-/google-genai-1.0.0.tgz#2785fa163788cb6214dffc1dc29fcd5bbb751493"
-  integrity sha512-ICUBZl/46nG6+Yhe5v7kp/2TQBGOzqEkpfKPLDeNyJ4x9OOL46xsW3ZZrHJjhGMQuR6/JMmQMTU9kLoYgsd1Tg==
+"@langchain/google-genai@^1.0.1":
+  version "1.0.1"
+  resolved "https://registry.yarnpkg.com/@langchain/google-genai/-/google-genai-1.0.1.tgz#3601645f652f24e3beb55acc61878070b49c24ed"
+  integrity sha512-a9Bzaswp1P+eA2V8hAWSBypqjxmH+/zhOY1TBdalQuPQBTRH35jBMVgX3CTTAheAzBUGQtlDD4/dR9tyemDbhw==
   dependencies:
     "@google/generative-ai" "^0.24.0"
     uuid "^11.1.0"
 
-"@langchain/groq@^1.0.0":
-  version "1.0.0"
-  resolved "https://registry.yarnpkg.com/@langchain/groq/-/groq-1.0.0.tgz#413b02158761ff406238467325cd4f9fe0990f3a"
-  integrity sha512-6fG9MEQHNXnxgObFHSPh+BPYyTGcoDnKd+GhI9l96cpHh+QNI+IvypicRCZVSsLdqzRCFHISvBQaH+SP5vgjIw==
+"@langchain/groq@^1.0.1":
+  version "1.0.1"
+  resolved "https://registry.yarnpkg.com/@langchain/groq/-/groq-1.0.1.tgz#7ec8822cd2f29eef4ae0f9c20f67268d1924ab96"
+  integrity sha512-vDQzv6A3mjG0/W/7vL4Iq+dnmhSbMHln+b7Rna810trjZzfNPZhAP6omqZyzCKIqjsQYUH4ODLnSUCNiarfYsQ==
   dependencies:
     groq-sdk "^0.19.0"
 
@@ -842,30 +842,30 @@
     "@langchain/langgraph-sdk" "~1.0.0"
     uuid "^10.0.0"
 
-"@langchain/ollama@^1.0.0":
-  version "1.0.0"
-  resolved "https://registry.yarnpkg.com/@langchain/ollama/-/ollama-1.0.0.tgz#803c353e9dfb1a9e7b20f1460a6a201fec29bb77"
-  integrity sha512-zqn6i7haMjvZW4FQWo0GrF4wYL5mLurdL0qoe+moYWYSCGaay4K7e/4dqM5C/MR16/HPFDzFbBRMkni2PDRBgA==
+"@langchain/langgraph@^1.0.1":
+  version "1.0.1"
+  resolved "https://registry.yarnpkg.com/@langchain/langgraph/-/langgraph-1.0.1.tgz#d0be714653e8a27665f86ea795c5c34189455406"
+  integrity sha512-7y8OTDLrHrpJ55Y5x7c7zU2BbqNllXwxM106Xrd+NaQB5CpEb4hbUfIwe4XmhhscKPwvhXAq3tjeUxw9MCiurQ==
+  dependencies:
+    "@langchain/langgraph-checkpoint" "^1.0.0"
+    "@langchain/langgraph-sdk" "~1.0.0"
+    uuid "^10.0.0"
+
+"@langchain/ollama@^1.0.1":
+  version "1.0.1"
+  resolved "https://registry.yarnpkg.com/@langchain/ollama/-/ollama-1.0.1.tgz#c63ac6db65110beef4020a5e2b167ad0bc678d33"
+  integrity sha512-Pe32hhTpMvnRlNFJxkdu6r1QzsONGz5uvoLiMU1TpgAUu7EyKr2osymlgjBLqDe2vMKUmqHb+yWRH0IppDBUOg==
   dependencies:
     ollama "^0.5.12"
     uuid "^10.0.0"
 
-"@langchain/openai@1.0.0", "@langchain/openai@^1.0.0":
-  version "1.0.0"
-  resolved "https://registry.yarnpkg.com/@langchain/openai/-/openai-1.0.0.tgz#03b15312286b30ce0149f6052620c6c95b4387bc"
-  integrity sha512-olKEUIjb3HBOiD/NR056iGJz4wiN6HhQ/u65YmGWYadWWoKOcGwheBw/FE0x6SH4zDlI3QmP+vMhuQoaww19BQ==
+"@langchain/openai@1.1.1", "@langchain/openai@^1.1.1":
+  version "1.1.1"
+  resolved "https://registry.yarnpkg.com/@langchain/openai/-/openai-1.1.1.tgz#67ddcf54ee7ac402f6b75b4b9e25447e78c56a93"
+  integrity sha512-0kUaXejo/sn6QAohWHDaAUapC4CJRkJIajGaWfJC+llSqpDBnmBE1oHg1M2fi1OCeP+ns9SxB6BTsq4Qbiqmig==
   dependencies:
     js-tiktoken "^1.0.12"
-    openai "^6.3.0"
-    zod "^3.25.76 || ^4"
-
-"@langchain/openai@1.0.0-alpha.3":
-  version "1.0.0-alpha.3"
-  resolved "https://registry.yarnpkg.com/@langchain/openai/-/openai-1.0.0-alpha.3.tgz#35c4e770e3421b75a226087af54fbeff147e201a"
-  integrity sha512-re2NXLYeLatPzoB6YRoFgB1fW6i5ygcLGa7PlNOhi3f93uU1vSlWMgjkO9dcN9ALmr/bhoruqJEn7U0Eva+6/w==
-  dependencies:
-    js-tiktoken "^1.0.12"
-    openai "^6.3.0"
+    openai "^6.9.0"
     zod "^3.25.76 || ^4"
 
 "@langchain/textsplitters@1.0.0", "@langchain/textsplitters@^1.0.0":
@@ -2607,11 +2607,6 @@ expand-template@^2.0.3:
   resolved "https://registry.yarnpkg.com/expand-template/-/expand-template-2.0.3.tgz#6e14b3fcee0f3a6340ecb57d2e8918692052a47c"
   integrity sha512-XYfuKMvj4O35f/pOXLObndIRvyQ+/+6AhODh+OKWj9S9498pHHn/IMszH+gt0fBCRWMNfk1ZSp5x3AifmnI2vg==
 
-expr-eval@^2.0.2:
-  version "2.0.2"
-  resolved "https://registry.yarnpkg.com/expr-eval/-/expr-eval-2.0.2.tgz#fa6f044a7b0c93fde830954eb9c5b0f7fbc7e201"
-  integrity sha512-4EMSHGOPSwAfBiibw3ndnP0AvjDWLsMvGOvWEZ2F96IGk0bIVdjQisOHxReSkE13mHcfbuCiXw+G4y0zv6N8Eg==
-
 fast-deep-equal@^3.1.1, fast-deep-equal@^3.1.3:
   version "3.1.3"
   resolved "https://registry.yarnpkg.com/fast-deep-equal/-/fast-deep-equal-3.1.3.tgz#3a7d56b559d6cbc3eb512325244e619a65c6c525"
@@ -3514,17 +3509,16 @@ kuler@^2.0.0:
   resolved "https://registry.yarnpkg.com/kuler/-/kuler-2.0.0.tgz#e2c570a3800388fb44407e851531c1d670b061b3"
   integrity sha512-Xq9nH7KlWZmXAtodXDDRE7vs6DU1gTU8zYDHDiWLSip45Egwq3plLHzPn27NgvzL2r1LMPC1vdqh98sQxtqj4A==
 
-langchain@^1.0.1:
-  version "1.0.1"
-  resolved "https://registry.yarnpkg.com/langchain/-/langchain-1.0.1.tgz#fb181176f4aa443ef02e9e5b563bcb4e170dfeb6"
-  integrity sha512-IT4JBVbKBh2AjaUFT9OsmOfeK3UbKy3SgdzZOuvet25sAaMpAR8IaM9XVddRs+OXQqVg6sOS01KUUVCJksVhHg==
+langchain@^1.0.4:
+  version "1.0.4"
+  resolved "https://registry.yarnpkg.com/langchain/-/langchain-1.0.4.tgz#c4fa22d927f41d56c356ecfccea5c08ae7b682ef"
+  integrity sha512-g7z2kKvnXOecybbVGHfI2ZmdmP309mxC1FYlq6WC/7RsKgX5MwY9gBjwK16mpKOaozOD9QCo1Ia7o2UcUBRb9Q==
   dependencies:
     "@langchain/langgraph" "^1.0.0"
     "@langchain/langgraph-checkpoint" "^1.0.0"
+    langsmith "~0.3.74"
     uuid "^10.0.0"
     zod "^3.25.76 || ^4"
-  optionalDependencies:
-    langsmith "^0.3.64"
 
 langsmith@^0.3.64:
   version "0.3.74"
@@ -3539,6 +3533,19 @@ langsmith@^0.3.64:
     semver "^7.6.3"
     uuid "^10.0.0"
 
+langsmith@~0.3.74:
+  version "0.3.79"
+  resolved "https://registry.yarnpkg.com/langsmith/-/langsmith-0.3.79.tgz#6c845644da26e7fdd8e9b80706091669fc43bda4"
+  integrity sha512-j5uiAsyy90zxlxaMuGjb7EdcL51Yx61SpKfDOI1nMPBbemGju+lf47he4e59Hp5K63CY8XWgFP42WeZ+zuIU4Q==
+  dependencies:
+    "@types/uuid" "^10.0.0"
+    chalk "^4.1.2"
+    console-table-printer "^2.12.1"
+    p-queue "^6.6.2"
+    p-retry "4"
+    semver "^7.6.3"
+    uuid "^10.0.0"
+
 language-subtag-registry@^0.3.20:
   version "0.3.22"
   resolved "https://registry.yarnpkg.com/language-subtag-registry/-/language-subtag-registry-0.3.22.tgz#2e1500861b2e457eba7e7ae86877cbd08fa1fd1d"
@@ -3686,6 +3693,11 @@ matcher@^3.0.0:
   dependencies:
     escape-string-regexp "^4.0.0"
 
+math-expression-evaluator@^2.0.0:
+  version "2.0.7"
+  resolved "https://registry.yarnpkg.com/math-expression-evaluator/-/math-expression-evaluator-2.0.7.tgz#dc99a80ce2bf7f9b7df878126feb5c506c1fdf5f"
+  integrity sha512-uwliJZ6BPHRq4eiqNWxZBDzKUiS5RIynFFcgchqhBOloVLVBpZpNG8jRYkedLcBvhph8TnRyWEuxPqiQcwIdog==
+
 math-intrinsics@^1.1.0:
   version "1.1.0"
   resolved "https://registry.yarnpkg.com/math-intrinsics/-/math-intrinsics-1.1.0.tgz#a0dd74be81e2aa5c2f27e65ce283605ee4e2b7f9"
@@ -4025,10 +4037,10 @@ onnxruntime-web@1.22.0-dev.20250409-89f8206ba4:
     platform "^1.3.6"
     protobufjs "^7.2.4"
 
-openai@^6.3.0:
-  version "6.5.0"
-  resolved "https://registry.yarnpkg.com/openai/-/openai-6.5.0.tgz#7dd9c4c0ca6e394c1d1e738b2000e084024685b2"
-  integrity sha512-bNqJ15Ijbs41KuJ2iYz/mGAruFHzQQt7zXo4EvjNLoB64aJdgn1jlMeDTsXjEg+idVYafg57QB/5Rd16oqvZ6A==
+openai@^6.9.0:
+  version "6.9.0"
+  resolved "https://registry.yarnpkg.com/openai/-/openai-6.9.0.tgz#acd15b2233c42b165981f3de8f4cfce27f844fce"
+  integrity sha512-n2sJRYmM+xfJ0l3OfH8eNnIyv3nQY7L08gZQu3dw6wSdfPtKAk92L83M2NIP5SS8Cl/bsBBG3yKzEOjkx0O+7A==
 
 openapi-types@^12.1.3:
   version "12.1.3"
@@ -5491,12 +5503,7 @@ yocto-queue@^0.1.0:
   resolved "https://registry.yarnpkg.com/yocto-queue/-/yocto-queue-0.1.0.tgz#0294eb3dee05028d31ee1a5fa2c556a6aaf10a1b"
   integrity sha512-rVksvsnNCdJ/ohGc6xgPwyN8eheCxsiLM8mxuE/t/mOVqJewPuO1miLpTHQiRgTKCLexL4MeAFVagts7HmNZ2Q==
 
-zod@^3.22.4:
-  version "3.22.4"
-  resolved "https://registry.yarnpkg.com/zod/-/zod-3.22.4.tgz#f31c3a9386f61b1f228af56faa9255e845cf3fff"
-  integrity sha512-iC+8Io04lddc+mVqQ9AZ7OQ2MrUKGN+oIQyq1vemgt46jwCwLfhq7/pwnBnNXXXZb8VTVLKwp9EDkx+ryxIWmg==
-
-"zod@^3.25.76 || ^4":
+"zod@^3.25.76 || ^4", zod@^4.1.12:
   version "4.1.12"
   resolved "https://registry.yarnpkg.com/zod/-/zod-4.1.12.tgz#64f1ea53d00eab91853195653b5af9eee68970f0"
   integrity sha512-JInaHOamG8pt5+Ey8kGmdcAcg3OL9reK8ltczgHTAwNhMys/6ThXHityHxVV2p3fkw/c+MAvBHFVYHFZDmjMCQ==