merge with master

README.md

@@ -67,9 +67,9 @@ Some cool features of AnythingLLM
 - Extremely efficient cost-saving measures for managing very large documents. You'll never pay to embed a massive document or transcript more than once. 90% more cost effective than other document chatbot solutions.
 - Full Developer API for custom integrations!
 
-### Supported LLMs, Embedders, Transcriptions models, and Vector Databases
+### Supported LLMs, Embedder Models, Speech models, and Vector Databases
 
-**Supported LLMs:**
+**Language Learning Models:**
 
 - [Any open-source llama.cpp compatible model](/server/storage/models/README.md#text-generation-llm-selection)
 - [OpenAI](https://openai.com)
@@ -88,9 +88,10 @@ Some cool features of AnythingLLM
 - [Groq](https://groq.com/)
 - [Cohere](https://cohere.com/)
 - [KoboldCPP](https://github.com/LostRuins/koboldcpp)
+- [LiteLLM](https://github.com/BerriAI/litellm)
 - [Text Generation Web UI](https://github.com/oobabooga/text-generation-webui)
 
-**Supported Embedding models:**
+**Embedder models:**
 
 - [AnythingLLM Native Embedder](/server/storage/models/README.md) (default)
 - [OpenAI](https://openai.com)
@@ -100,12 +101,22 @@ Some cool features of AnythingLLM
 - [LM Studio (all)](https://lmstudio.ai)
 - [Cohere](https://cohere.com/)
 
-**Supported Transcription models:**
+**Audio Transcription models:**
 
 - [AnythingLLM Built-in](https://github.com/Mintplex-Labs/anything-llm/tree/master/server/storage/models#audiovideo-transcription) (default)
 - [OpenAI](https://openai.com/)
 
-**Supported Vector Databases:**
+**TTS (text-to-speech) support:**
+
+- Native Browser Built-in (default)
+- [OpenAI TTS](https://platform.openai.com/docs/guides/text-to-speech/voice-options)
+- [ElevenLabs](https://elevenlabs.io/)
+
+**STT (speech-to-text) support:**
+
+- Native Browser Built-in (default)
+
+**Vector Databases:**
 
 - [LanceDB](https://github.com/lancedb/lancedb) (default)
 - [Astra DB](https://www.datastax.com/products/datastax-astra)
@@ -122,8 +133,9 @@ This monorepo consists of three main sections:
 
 - `frontend`: A viteJS + React frontend that you can run to easily create and manage all your content the LLM can use.
 - `server`: A NodeJS express server to handle all the interactions and do all the vectorDB management and LLM interactions.
-- `docker`: Docker instructions and build process + information for building from source.
 - `collector`: NodeJS express server that process and parses documents from the UI.
+- `docker`: Docker instructions and build process + information for building from source.
+- `embed`: Code specifically for generation of the [embed widget](./embed/README.md).
 
 ## 🛳 Self Hosting
 
@@ -132,9 +144,9 @@ Mintplex Labs & the community maintain a number of deployment methods, scripts,
 |----------------------------------------|----:|-----|---------------|------------|
 | [![Deploy on Docker][docker-btn]][docker-deploy] | [![Deploy on AWS][aws-btn]][aws-deploy] | [![Deploy on GCP][gcp-btn]][gcp-deploy] | [![Deploy on DigitalOcean][do-btn]][do-deploy] | [![Deploy on Render.com][render-btn]][render-deploy] |
 
-| Railway                                             |
+| Railway  |  RepoCloud |
-| --------------------------------------------------- |
+| --- | --- |
-| [![Deploy on Railway][railway-btn]][railway-deploy] |
+| [![Deploy on Railway][railway-btn]][railway-deploy] | [![Deploy on RepoCloud][repocloud-btn]][repocloud-deploy] |
 
 [or set up a production AnythingLLM instance without Docker →](./BARE_METAL.md)
 
@@ -223,3 +235,5 @@ This project is [MIT](./LICENSE) licensed.
 [render-deploy]: https://render.com/deploy?repo=https://github.com/Mintplex-Labs/anything-llm&branch=render
 [railway-btn]: https://railway.app/button.svg
 [railway-deploy]: https://railway.app/template/HNSCS1?referralCode=WFgJkn
+[repocloud-btn]: https://d16t0pc4846x52.cloudfront.net/deploylobe.svg
+[repocloud-deploy]: https://repocloud.io/details/?app_id=276

collector/utils/WhisperProviders/localWhisper.js

@@ -1,19 +1,23 @@
 const fs = require("fs");
 const path = require("path");
 const { v4 } = require("uuid");
+const defaultWhisper = "Xenova/whisper-small"; // Model Card: https://huggingface.co/Xenova/whisper-small
+const fileSize = {
+  "Xenova/whisper-small": "250mb",
+  "Xenova/whisper-large": "1.56GB",
+};
 
 class LocalWhisper {
-  constructor() {
+  constructor({ options }) {
-    // Model Card: https://huggingface.co/Xenova/whisper-small
+    this.model = options?.WhisperModelPref ?? defaultWhisper;
-    this.model = "Xenova/whisper-small";
+    this.fileSize = fileSize[this.model];
     this.cacheDir = path.resolve(
       process.env.STORAGE_DIR
         ? path.resolve(process.env.STORAGE_DIR, `models`)
         : path.resolve(__dirname, `../../../server/storage/models`)
     );
 
-    this.modelPath = path.resolve(this.cacheDir, "Xenova", "whisper-small");
+    this.modelPath = path.resolve(this.cacheDir, ...this.model.split("/"));
-
     // Make directory when it does not exist in existing installations
     if (!fs.existsSync(this.cacheDir))
       fs.mkdirSync(this.cacheDir, { recursive: true });
@@ -104,7 +108,7 @@ class LocalWhisper {
   async client() {
     if (!fs.existsSync(this.modelPath)) {
       this.#log(
-        `The native whisper model has never been run and will be downloaded right now. Subsequent runs will be faster. (~250MB)`
+        `The native whisper model has never been run and will be downloaded right now. Subsequent runs will be faster. (~${this.fileSize})`
       );
     }
 

collector/utils/extensions/Confluence/index.js

@@ -12,18 +12,23 @@ const {
 function validSpaceUrl(spaceUrl = "") {
   // Atlassian default URL match
   const atlassianPattern = new UrlPattern(
-    "https\\://(:subdomain).atlassian.net/wiki/spaces/(:spaceKey)/*"
+    "https\\://(:subdomain).atlassian.net/wiki/spaces/(:spaceKey)*"
   );
   const atlassianMatch = atlassianPattern.match(spaceUrl);
   if (atlassianMatch) {
     return { valid: true, result: atlassianMatch };
   }
 
-  // Custom Confluence URL match
+  let customMatch = null;
-  const customPattern = new UrlPattern(
+  [
-    "https\\://(:subdomain.):domain.:tld/wiki/spaces/(:spaceKey)/*"
+    "https\\://(:subdomain.):domain.:tld/wiki/spaces/(:spaceKey)*", // Custom Confluence space
-  );
+    "https\\://(:subdomain.):domain.:tld/display/(:spaceKey)*", // Custom Confluence space + Human-readable space tag.
-  const customMatch = customPattern.match(spaceUrl);
+  ].forEach((matchPattern) => {
+    if (!!customMatch) return;
+    const pattern = new UrlPattern(matchPattern);
+    customMatch = pattern.match(spaceUrl);
+  });
+
   if (customMatch) {
     customMatch.customDomain =
       (customMatch.subdomain ? `${customMatch.subdomain}.` : "") + //

docker/.env.example

@@ -82,6 +82,12 @@ GID='1000'
 # GENERIC_OPEN_AI_MODEL_TOKEN_LIMIT=4096
 # GENERIC_OPEN_AI_API_KEY=sk-123abc
 
+# LLM_PROVIDER='litellm'
+# LITE_LLM_MODEL_PREF='gpt-3.5-turbo'
+# LITE_LLM_MODEL_TOKEN_LIMIT=4096
+# LITE_LLM_BASE_PATH='http://127.0.0.1:4000'
+# LITE_LLM_API_KEY='sk-123abc'
+
 # LLM_PROVIDER='cohere'
 # COHERE_API_KEY=
 # COHERE_MODEL_PREF='command-r'
@@ -118,6 +124,10 @@ GID='1000'
 # COHERE_API_KEY=
 # EMBEDDING_MODEL_PREF='embed-english-v3.0'
 
+# EMBEDDING_ENGINE='voyageai'
+# VOYAGEAI_API_KEY=
+# EMBEDDING_MODEL_PREF='voyage-large-2-instruct'
+
 ###########################################
 ######## Vector Database Selection ########
 ###########################################

embed/index.html

@@ -1,4 +1,8 @@
 <!doctype html>
+
+<head>
+  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+</head>
 <html lang="en">
 
   <body>

embed/src/App.jsx

@@ -28,18 +28,18 @@ export default function App() {
 
   const position = embedSettings.position || "bottom-right";
   const windowWidth = embedSettings.windowWidth
-    ? `md:max-w-[${embedSettings.windowWidth}]`
+    ? `max-w-[${embedSettings.windowWidth}]`
-    : "md:max-w-[400px]";
+    : "max-w-[400px]";
   const windowHeight = embedSettings.windowHeight
-    ? `md:max-h-[${embedSettings.windowHeight}]`
+    ? `max-h-[${embedSettings.windowHeight}]`
-    : "md:max-h-[700px]";
+    : "max-h-[700px]";
 
   return (
     <>
       <Head />
       <div className={`fixed inset-0 z-50 ${isChatOpen ? "block" : "hidden"}`}>
         <div
-          className={`${windowHeight} ${windowWidth} h-full w-full bg-white md:fixed md:bottom-0 md:right-0 md:mb-4 md:mr-4 md:rounded-2xl md:border md:border-gray-300 md:shadow-[0_4px_14px_rgba(0,0,0,0.25)] ${positionClasses[position]}`}
+          className={`${windowHeight} ${windowWidth} h-full w-full bg-white fixed bottom-0 right-0 mb-4 md:mr-4 rounded-2xl border border-gray-300 shadow-[0_4px_14px_rgba(0,0,0,0.25)] ${positionClasses[position]}`}
           id="anything-llm-chat"
         >
           {isChatOpen && (

frontend/.gitignore

@@ -9,10 +9,8 @@ lerna-debug.log*
 
 node_modules
 dist
-lib
 dist-ssr
 *.local
-!frontend/components/lib
 
 # Editor directories and files
 .vscode/*

frontend/src/components/EmbeddingSelection/VoyageAiOptions/index.jsx

@@ -0,0 +1,50 @@
+export default function VoyageAiOptions({ settings }) {
+  return (
+    <div className="w-full flex flex-col gap-y-4">
+      <div className="w-full flex items-center gap-4">
+        <div className="flex flex-col w-60">
+          <label className="text-white text-sm font-semibold block mb-4">
+            API Key
+          </label>
+          <input
+            type="password"
+            name="VoyageAiApiKey"
+            className="bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+            placeholder="Voyage AI API Key"
+            defaultValue={settings?.VoyageAiApiKey ? "*".repeat(20) : ""}
+            required={true}
+            autoComplete="off"
+            spellCheck={false}
+          />
+        </div>
+        <div className="flex flex-col w-60">
+          <label className="text-white text-sm font-semibold block mb-4">
+            Model Preference
+          </label>
+          <select
+            name="EmbeddingModelPref"
+            required={true}
+            defaultValue={settings?.EmbeddingModelPref}
+            className="bg-zinc-900 border-gray-500 text-white text-sm rounded-lg block w-full p-2.5"
+          >
+            <optgroup label="Available embedding models">
+              {[
+                "voyage-large-2-instruct",
+                "voyage-law-2",
+                "voyage-code-2",
+                "voyage-large-2",
+                "voyage-2",
+              ].map((model) => {
+                return (
+                  <option key={model} value={model}>
+                    {model}
+                  </option>
+                );
+              })}
+            </optgroup>
+          </select>
+        </div>
+      </div>
+    </div>
+  );
+}

frontend/src/components/Footer/index.jsx

@@ -14,6 +14,8 @@ import {
 import React, { useEffect, useState } from "react";
 import SettingsButton from "../SettingsButton";
 import { isMobile } from "react-device-detect";
+import { Tooltip } from "react-tooltip";
+import { v4 } from "uuid";
 
 export const MAX_ICONS = 3;
 export const ICON_COMPONENTS = {
@@ -47,36 +49,48 @@ export default function Footer() {
     return (
       <div className="flex justify-center mb-2">
         <div className="flex space-x-4">
+          <ToolTipWrapper id="open-github">
             <a
               href={paths.github()}
               target="_blank"
               rel="noreferrer"
               className="transition-all duration-300 p-2 rounded-full text-white bg-sidebar-button hover:bg-menu-item-selected-gradient hover:border-slate-100 hover:border-opacity-50 border-transparent border"
               aria-label="Find us on Github"
+              data-tooltip-id="open-github"
+              data-tooltip-content="View source code on Github"
             >
               <GithubLogo weight="fill" className="h-5 w-5 " />
             </a>
+          </ToolTipWrapper>
+          <ToolTipWrapper id="open-documentation">
             <a
               href={paths.docs()}
               target="_blank"
               rel="noreferrer"
-            className="transition-all duration-300 p-2 rounded-full text-white bg-sidebar-button hover:bg-menu-item-selected-gradient hover:border-slate-100 hover:border-opacity-50 border-transparent border"
+              className="w-fit transition-all duration-300 p-2 rounded-full text-white bg-sidebar-button hover:bg-menu-item-selected-gradient hover:border-slate-100 hover:border-opacity-50 border-transparent border"
               aria-label="Docs"
+              data-tooltip-id="open-documentation"
+              data-tooltip-content="Open AnythingLLM help docs"
             >
               <BookOpen weight="fill" className="h-5 w-5 " />
             </a>
+          </ToolTipWrapper>
+          <ToolTipWrapper id="open-discord">
             <a
               href={paths.discord()}
               target="_blank"
               rel="noreferrer"
               className="transition-all duration-300 p-2 rounded-full text-white bg-sidebar-button hover:bg-menu-item-selected-gradient hover:border-slate-100 hover:border-opacity-50 border-transparent border"
               aria-label="Join our Discord server"
+              data-tooltip-id="open-discord"
+              data-tooltip-content="Join the AnythingLLM Discord"
             >
               <DiscordLogo
                 weight="fill"
                 className="h-5 w-5 stroke-slate-200 group-hover:stroke-slate-200"
               />
             </a>
+          </ToolTipWrapper>
           {!isMobile && <SettingsButton />}
         </div>
       </div>
@@ -105,3 +119,17 @@ export default function Footer() {
     </div>
   );
 }
+
+export function ToolTipWrapper({ id = v4(), children }) {
+  return (
+    <div className="flex w-fit">
+      {children}
+      <Tooltip
+        id={id}
+        place="top"
+        delayShow={300}
+        className="tooltip !text-xs z-99"
+      />
+    </div>
+  );
+}

frontend/src/components/LLMSelection/GeminiLLMOptions/index.jsx

@@ -19,6 +19,7 @@ export default function GeminiLLMOptions({ settings }) {
         </div>
 
         {!settings?.credentialsOnly && (
+          <>
             <div className="flex flex-col w-60">
               <label className="text-white text-sm font-semibold block mb-4">
                 Chat Model Selection
@@ -38,6 +39,27 @@ export default function GeminiLLMOptions({ settings }) {
                 })}
               </select>
             </div>
+            <div className="flex flex-col w-60">
+              <label className="text-white text-sm font-semibold block mb-4">
+                Safety Setting
+              </label>
+              <select
+                name="GeminiSafetySetting"
+                defaultValue={
+                  settings?.GeminiSafetySetting || "BLOCK_MEDIUM_AND_ABOVE"
+                }
+                required={true}
+                className="bg-zinc-900 border-gray-500 text-white text-sm rounded-lg block w-full p-2.5"
+              >
+                <option value="BLOCK_NONE">None</option>
+                <option value="BLOCK_ONLY_HIGH">Block few</option>
+                <option value="BLOCK_MEDIUM_AND_ABOVE">
+                  Block some (default)
+                </option>
+                <option value="BLOCK_LOW_AND_ABOVE">Block most</option>
+              </select>
+            </div>
+          </>
         )}
       </div>
     </div>

frontend/src/components/LLMSelection/GenericOpenAiOptions/index.jsx

@@ -1,5 +1,6 @@
 export default function GenericOpenAiOptions({ settings }) {
   return (
+    <div className="flex flex-col gap-y-4">
       <div className="flex gap-4 flex-wrap">
         <div className="flex flex-col w-60">
           <label className="text-white text-sm font-semibold block mb-4">
@@ -45,6 +46,8 @@ export default function GenericOpenAiOptions({ settings }) {
             autoComplete="off"
           />
         </div>
+      </div>
+      <div className="flex gap-x-4 flex-wrap">
         <div className="flex flex-col w-60">
           <label className="text-white text-sm font-semibold block mb-4">
             Token context window
@@ -77,5 +80,6 @@ export default function GenericOpenAiOptions({ settings }) {
           />
         </div>
       </div>
+    </div>
   );
 }

frontend/src/components/LLMSelection/LiteLLMOptions/index.jsx

@@ -0,0 +1,148 @@
+import { useEffect, useState } from "react";
+import System from "@/models/system";
+
+export default function LiteLLMOptions({ settings }) {
+  const [basePathValue, setBasePathValue] = useState(settings?.LiteLLMBasePath);
+  const [basePath, setBasePath] = useState(settings?.LiteLLMBasePath);
+  const [apiKeyValue, setApiKeyValue] = useState(settings?.LiteLLMAPIKey);
+  const [apiKey, setApiKey] = useState(settings?.LiteLLMAPIKey);
+
+  return (
+    <div className="w-full flex flex-col gap-y-4">
+      <div className="w-full flex items-center gap-4">
+        <div className="flex flex-col w-60">
+          <label className="text-white text-sm font-semibold block mb-4">
+            Base URL
+          </label>
+          <input
+            type="url"
+            name="LiteLLMBasePath"
+            className="bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+            placeholder="http://127.0.0.1:4000"
+            defaultValue={settings?.LiteLLMBasePath}
+            required={true}
+            autoComplete="off"
+            spellCheck={false}
+            onChange={(e) => setBasePathValue(e.target.value)}
+            onBlur={() => setBasePath(basePathValue)}
+          />
+        </div>
+        <LiteLLMModelSelection
+          settings={settings}
+          basePath={basePath}
+          apiKey={apiKey}
+        />
+        <div className="flex flex-col w-60">
+          <label className="text-white text-sm font-semibold block mb-4">
+            Token context window
+          </label>
+          <input
+            type="number"
+            name="LiteLLMTokenLimit"
+            className="bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+            placeholder="4096"
+            min={1}
+            onScroll={(e) => e.target.blur()}
+            defaultValue={settings?.LiteLLMTokenLimit}
+            required={true}
+            autoComplete="off"
+          />
+        </div>
+      </div>
+      <div className="w-full flex items-center gap-4">
+        <div className="flex flex-col w-60">
+          <div className="flex flex-col gap-y-1 mb-4">
+            <label className="text-white text-sm font-semibold flex items-center gap-x-2">
+              API Key <p className="!text-xs !italic !font-thin">optional</p>
+            </label>
+          </div>
+          <input
+            type="password"
+            name="LiteLLMAPIKey"
+            className="bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+            placeholder="sk-mysecretkey"
+            defaultValue={settings?.LiteLLMAPIKey ? "*".repeat(20) : ""}
+            autoComplete="off"
+            spellCheck={false}
+            onChange={(e) => setApiKeyValue(e.target.value)}
+            onBlur={() => setApiKey(apiKeyValue)}
+          />
+        </div>
+      </div>
+    </div>
+  );
+}
+
+function LiteLLMModelSelection({ settings, basePath = null, apiKey = null }) {
+  const [customModels, setCustomModels] = useState([]);
+  const [loading, setLoading] = useState(true);
+
+  useEffect(() => {
+    async function findCustomModels() {
+      if (!basePath) {
+        setCustomModels([]);
+        setLoading(false);
+        return;
+      }
+      setLoading(true);
+      const { models } = await System.customModels(
+        "litellm",
+        typeof apiKey === "boolean" ? null : apiKey,
+        basePath
+      );
+      setCustomModels(models || []);
+      setLoading(false);
+    }
+    findCustomModels();
+  }, [basePath, apiKey]);
+
+  if (loading || customModels.length == 0) {
+    return (
+      <div className="flex flex-col w-60">
+        <label className="text-white text-sm font-semibold block mb-4">
+          Chat Model Selection
+        </label>
+        <select
+          name="LiteLLMModelPref"
+          disabled={true}
+          className="bg-zinc-900 border-gray-500 text-white text-sm rounded-lg block w-full p-2.5"
+        >
+          <option disabled={true} selected={true}>
+            {basePath?.includes("/v1")
+              ? "-- loading available models --"
+              : "-- waiting for URL --"}
+          </option>
+        </select>
+      </div>
+    );
+  }
+
+  return (
+    <div className="flex flex-col w-60">
+      <label className="text-white text-sm font-semibold block mb-4">
+        Chat Model Selection
+      </label>
+      <select
+        name="LiteLLMModelPref"
+        required={true}
+        className="bg-zinc-900 border-gray-500 text-white text-sm rounded-lg block w-full p-2.5"
+      >
+        {customModels.length > 0 && (
+          <optgroup label="Your loaded models">
+            {customModels.map((model) => {
+              return (
+                <option
+                  key={model.id}
+                  value={model.id}
+                  selected={settings.LiteLLMModelPref === model.id}
+                >
+                  {model.id}
+                </option>
+              );
+            })}
+          </optgroup>
+        )}
+      </select>
+    </div>
+  );
+}

frontend/src/components/Modals/MangeWorkspace/DataConnectors/Connectors/Github/index.jsx

@@ -3,7 +3,7 @@ import System from "@/models/system";
 import showToast from "@/utils/toast";
 import pluralize from "pluralize";
 import { TagsInput } from "react-tag-input-component";
-import { Warning } from "@phosphor-icons/react";
+import { Info, Warning } from "@phosphor-icons/react";
 import { Tooltip } from "react-tooltip";
 
 const DEFAULT_BRANCHES = ["main", "master"];
@@ -92,45 +92,7 @@ export default function GithubOptions() {
                     <p className="font-bold text-white">Github Access Token</p>{" "}
                     <p className="text-xs text-white/50 font-light flex items-center">
                       optional
-                      {!accessToken && (
+                      <PATTooltip accessToken={accessToken} />
-                        <Warning
-                          size={14}
-                          className="ml-1 text-orange-500 cursor-pointer"
-                          data-tooltip-id="access-token-tooltip"
-                          data-tooltip-place="right"
-                        />
-                      )}
-                      <Tooltip
-                        delayHide={300}
-                        id="access-token-tooltip"
-                        className="max-w-xs"
-                        clickable={true}
-                      >
-                        <p className="text-sm">
-                          Without a{" "}
-                          <a
-                            href="https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens"
-                            rel="noreferrer"
-                            target="_blank"
-                            className="underline"
-                            onClick={(e) => e.stopPropagation()}
-                          >
-                            Personal Access Token
-                          </a>
-                          , the GitHub API may limit the number of files that
-                          can be collected due to rate limits. You can{" "}
-                          <a
-                            href="https://github.com/settings/personal-access-tokens/new"
-                            rel="noreferrer"
-                            target="_blank"
-                            className="underline"
-                            onClick={(e) => e.stopPropagation()}
-                          >
-                            create a temporary Access Token
-                          </a>{" "}
-                          to avoid this issue.
-                        </p>
-                      </Tooltip>
                     </p>
                   </label>
                   <p className="text-xs font-normal text-white/50">
@@ -180,6 +142,7 @@ export default function GithubOptions() {
           </div>
 
           <div className="flex flex-col gap-y-2 w-full pr-10">
+            <PATAlert accessToken={accessToken} />
             <button
               type="submit"
               disabled={loading}
@@ -269,3 +232,78 @@ function GitHubBranchSelection({ repo, accessToken }) {
     </div>
   );
 }
+
+function PATAlert({ accessToken }) {
+  if (!!accessToken) return null;
+  return (
+    <div className="flex flex-col md:flex-row md:items-center gap-x-2 text-white mb-4 bg-blue-800/30 w-fit rounded-lg px-4 py-2">
+      <div className="gap-x-2 flex items-center">
+        <Info className="shrink-0" size={25} />
+        <p className="text-sm">
+          Without filling out the <b>Github Access Token</b> this data connector
+          will only be able to collect the <b>top-level</b> files of the repo
+          due to GitHub's public API rate-limits.
+          <br />
+          <br />
+          <a
+            href="https://github.com/settings/personal-access-tokens/new"
+            rel="noreferrer"
+            target="_blank"
+            className="underline"
+            onClick={(e) => e.stopPropagation()}
+          >
+            {" "}
+            Get a free Personal Access Token with a GitHub account here.
+          </a>
+        </p>
+      </div>
+    </div>
+  );
+}
+
+function PATTooltip({ accessToken }) {
+  if (!!accessToken) return null;
+  return (
+    <>
+      {!accessToken && (
+        <Warning
+          size={14}
+          className="ml-1 text-orange-500 cursor-pointer"
+          data-tooltip-id="access-token-tooltip"
+          data-tooltip-place="right"
+        />
+      )}
+      <Tooltip
+        delayHide={300}
+        id="access-token-tooltip"
+        className="max-w-xs"
+        clickable={true}
+      >
+        <p className="text-sm">
+          Without a{" "}
+          <a
+            href="https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens"
+            rel="noreferrer"
+            target="_blank"
+            className="underline"
+            onClick={(e) => e.stopPropagation()}
+          >
+            Personal Access Token
+          </a>
+          , the GitHub API may limit the number of files that can be collected
+          due to rate limits. You can{" "}
+          <a
+            href="https://github.com/settings/personal-access-tokens/new"
+            rel="noreferrer"
+            target="_blank"
+            className="underline"
+            onClick={(e) => e.stopPropagation()}
+          >
+            create a temporary Access Token
+          </a>{" "}
+          to avoid this issue.
+        </p>
+      </Tooltip>
+    </>
+  );
+}

frontend/src/components/SettingsButton/index.jsx

@@ -3,6 +3,7 @@ import paths from "@/utils/paths";
 import { ArrowUUpLeft, Wrench } from "@phosphor-icons/react";
 import { Link } from "react-router-dom";
 import { useMatch } from "react-router-dom";
+import { ToolTipWrapper } from "../Footer";
 
 export default function SettingsButton() {
   const isInSettings = !!useMatch("/settings/*");
@@ -12,22 +13,32 @@ export default function SettingsButton() {
 
   if (isInSettings)
     return (
+      <ToolTipWrapper id="go-home">
         <Link
           to={paths.home()}
           className="transition-all duration-300 p-2 rounded-full text-white bg-sidebar-button hover:bg-menu-item-selected-gradient hover:border-slate-100 hover:border-opacity-50 border-transparent border"
           aria-label="Home"
+          data-tooltip-id="go-home"
+          data-tooltip-content="Back to workspaces"
         >
           <ArrowUUpLeft className="h-5 w-5" weight="fill" />
         </Link>
+      </ToolTipWrapper>
     );
 
   return (
+    <ToolTipWrapper id="open-settings">
       <Link
-      to={!!user?.role ? paths.settings.system() : paths.settings.appearance()}
+        to={
+          !!user?.role ? paths.settings.system() : paths.settings.appearance()
+        }
         className="transition-all duration-300 p-2 rounded-full text-white bg-sidebar-button hover:bg-menu-item-selected-gradient hover:border-slate-100 hover:border-opacity-50 border-transparent border"
         aria-label="Settings"
+        data-tooltip-id="open-settings"
+        data-tooltip-content="Open settings"
       >
         <Wrench className="h-5 w-5" weight="fill" />
       </Link>
+    </ToolTipWrapper>
   );
 }

frontend/src/components/SettingsSidebar/index.jsx

@@ -329,7 +329,7 @@ const SidebarOptions = ({ user = null }) => (
     <Option
       href={paths.settings.embedSetup()}
       childLinks={[paths.settings.embedChats()]}
-      btnText="Embedded Chat"
+      btnText="Chat Embed Widgets"
       icon={<CodeBlock className="h-5 w-5 flex-shrink-0" />}
       user={user}
       flex={true}
@@ -338,7 +338,7 @@ const SidebarOptions = ({ user = null }) => (
         <>
           <Option
             href={paths.settings.embedChats()}
-            btnText="Embedded Chat History"
+            btnText="Chat Embed History"
             icon={<Barcode className="h-5 w-5 flex-shrink-0" />}
             user={user}
             flex={true}

frontend/src/components/TextToSpeech/ElevenLabsOptions/index.jsx

@@ -84,6 +84,7 @@ function ElevenLabsModelSelection({ apiKey, settings }) {
       <select
         name="TTSElevenLabsVoiceModel"
         required={true}
+        defaultValue={settings?.TTSElevenLabsVoiceModel}
         className="bg-zinc-900 border-gray-500 text-white text-sm rounded-lg block w-full p-2.5"
       >
         {Object.keys(groupedModels)
@@ -91,11 +92,7 @@ function ElevenLabsModelSelection({ apiKey, settings }) {
           .map((organization) => (
             <optgroup key={organization} label={organization}>
               {groupedModels[organization].map((model) => (
-                <option
+                <option key={model.id} value={model.id}>
-                  key={model.id}
-                  value={model.id}
-                  selected={settings?.OpenAiModelPref === model.id}
-                >
                   {model.name}
                 </option>
               ))}

frontend/src/components/TextToSpeech/OpenAiOptions/index.jsx

@@ -35,7 +35,11 @@ export default function OpenAiTextToSpeechOptions({ settings }) {
         >
           {["alloy", "echo", "fable", "onyx", "nova", "shimmer"].map(
             (voice) => {
-              return <option value={voice}>{toProperCase(voice)}</option>;
+              return (
+                <option key={voice} value={voice}>
+                  {toProperCase(voice)}
+                </option>
+              );
             }
           )}
         </select>

frontend/src/components/TranscriptionSelection/NativeTranscriptionOptions/index.jsx

@@ -1,38 +1,89 @@
 import { Gauge } from "@phosphor-icons/react";
-export default function NativeTranscriptionOptions() {
+import { useState } from "react";
+
+export default function NativeTranscriptionOptions({ settings }) {
+  const [model, setModel] = useState(settings?.WhisperModelPref);
+
   return (
     <div className="w-full flex flex-col gap-y-4">
-      <div className="flex flex-col md:flex-row md:items-center gap-x-2 text-white mb-4 bg-blue-800/30 w-fit rounded-lg px-4 py-2">
+      <LocalWarning model={model} />
-        <div className="gap-x-2 flex items-center">
-          <Gauge size={25} />
-          <p className="text-sm">
-            Using the local whisper model on machines with limited RAM or CPU
-            can stall AnythingLLM when processing media files.
-            <br />
-            We recommend at least 2GB of RAM and upload files &lt;10Mb.
-            <br />
-            <br />
-            <i>
-              The built-in model will automatically download on the first use.
-            </i>
-          </p>
-        </div>
-      </div>
       <div className="w-full flex items-center gap-4">
         <div className="flex flex-col w-60">
           <label className="text-white text-sm font-semibold block mb-4">
             Model Selection
           </label>
           <select
-            disabled={true}
+            name="WhisperModelPref"
+            defaultValue={model}
+            onChange={(e) => setModel(e.target.value)}
             className="bg-zinc-900 border-gray-500 text-white text-sm rounded-lg block w-full p-2.5"
           >
-            <option disabled={true} selected={true}>
+            {["Xenova/whisper-small", "Xenova/whisper-large"].map(
-              Xenova/whisper-small
+              (value, i) => {
+                return (
+                  <option key={i} value={value}>
+                    {value}
                   </option>
+                );
+              }
+            )}
           </select>
         </div>
       </div>
     </div>
   );
 }
+
+function LocalWarning({ model }) {
+  switch (model) {
+    case "Xenova/whisper-small":
+      return <WhisperSmall />;
+    case "Xenova/whisper-large":
+      return <WhisperLarge />;
+    default:
+      return <WhisperSmall />;
+  }
+}
+
+function WhisperSmall() {
+  return (
+    <div className="flex flex-col md:flex-row md:items-center gap-x-2 text-white mb-4 bg-blue-800/30 w-fit rounded-lg px-4 py-2">
+      <div className="gap-x-2 flex items-center">
+        <Gauge size={25} />
+        <p className="text-sm">
+          Running the <b>whisper-small</b> model on a machine with limited RAM
+          or CPU can stall AnythingLLM when processing media files.
+          <br />
+          We recommend at least 2GB of RAM and upload files &lt;10Mb.
+          <br />
+          <br />
+          <i>
+            This model will automatically download on the first use. (250mb)
+          </i>
+        </p>
+      </div>
+    </div>
+  );
+}
+
+function WhisperLarge() {
+  return (
+    <div className="flex flex-col md:flex-row md:items-center gap-x-2 text-white mb-4 bg-blue-800/30 w-fit rounded-lg px-4 py-2">
+      <div className="gap-x-2 flex items-center">
+        <Gauge size={25} />
+        <p className="text-sm">
+          Using the <b>whisper-large</b> model on machines with limited RAM or
+          CPU can stall AnythingLLM when processing media files. This model is
+          substantially larger than the whisper-small.
+          <br />
+          We recommend at least 8GB of RAM and upload files &lt;10Mb.
+          <br />
+          <br />
+          <i>
+            This model will automatically download on the first use. (1.56GB)
+          </i>
+        </p>
+      </div>
+    </div>
+  );
+}

frontend/src/components/WorkspaceChat/ChatContainer/ChatHistory/Citation/index.jsx

@@ -115,6 +115,11 @@ function SkeletonLine() {
   );
 }
 
+function omitChunkHeader(text) {
+  if (!text.startsWith("<document_metadata>")) return text;
+  return text.split("</document_metadata>")[1].trim();
+}
+
 function CitationDetailModal({ source, onClose }) {
   const { references, title, chunks } = source;
   const { isUrl, text: webpageUrl, href: linkTo } = parseChunkSource(source);
@@ -167,7 +172,7 @@ function CitationDetailModal({ source, onClose }) {
               <div key={idx} className="pt-6 text-white">
                 <div className="flex flex-col w-full justify-start pb-6 gap-y-1">
                   <p className="text-white whitespace-pre-line">
-                    {HTMLDecode(text)}
+                    {HTMLDecode(omitChunkHeader(text))}
                   </p>
 
                   {!!score && (

frontend/src/components/WorkspaceChat/ChatContainer/ChatHistory/HistoricalMessage/Actions/index.jsx

@@ -32,8 +32,7 @@ const Actions = ({
     <div className="flex w-full justify-between items-center">
       <div className="flex justify-start items-center gap-x-4">
         <CopyMessage message={message} />
-        {isLastMessage &&
+        {isLastMessage && (
-          !message?.includes("Workspace chat memory was reset!") && (
           <RegenerateMessage
             regenerateMessage={regenerateMessage}
             slug={slug}
@@ -127,6 +126,7 @@ function CopyMessage({ message }) {
 }
 
 function RegenerateMessage({ regenerateMessage, chatId }) {
+  if (!chatId) return null;
   return (
     <div className="mt-3 relative">
       <button

frontend/src/components/WorkspaceChat/ChatContainer/ChatHistory/HistoricalMessage/index.jsx

@@ -57,7 +57,7 @@ const HistoricalMessage = ({
           <div className="flex gap-x-5">
             <div className="relative w-[35px] h-[35px] rounded-full flex-shrink-0 overflow-hidden" />
             <Actions
-              message={DOMPurify.sanitize(message)}
+              message={message}
               feedbackScore={feedbackScore}
               chatId={chatId}
               slug={workspace?.slug}

frontend/src/components/WorkspaceChat/ChatContainer/PromptInput/SlashCommands/SlashPresets/EditPresetModal.jsx

@@ -31,10 +31,7 @@ export default function EditPresetModal({
   };
 
   const handleDelete = async () => {
-    const confirmDelete = window.confirm(
+    if (!window.confirm("Are you sure you want to delete this preset?")) return;
-      "Are you sure you want to delete this preset?"
-    );
-    if (!confirmDelete) return;
 
     setDeleting(true);
     await onDelete(preset.id);

frontend/src/components/WorkspaceChat/ChatContainer/PromptInput/index.jsx

@@ -3,7 +3,6 @@ import SlashCommandsButton, {
   SlashCommands,
   useSlashCommands,
 } from "./SlashCommands";
-import { isMobile } from "react-device-detect";
 import debounce from "lodash.debounce";
 import { PaperPlaneRight } from "@phosphor-icons/react";
 import StopGenerationButton from "./StopGenerationButton";
@@ -13,6 +12,7 @@ import AvailableAgentsButton, {
 } from "./AgentMenu";
 import TextSizeButton from "./TextSizeMenu";
 import SpeechToText from "./SpeechToText";
+import { Tooltip } from "react-tooltip";
 
 export const PROMPT_INPUT_EVENT = "set_prompt_input";
 export default function PromptInput({
@@ -83,7 +83,6 @@ export default function PromptInput({
   };
 
   const adjustTextArea = (event) => {
-    if (isMobile) return false;
     const element = event.target;
     element.style.height = "auto";
     element.style.height = `${element.scrollHeight}px`;
@@ -130,20 +129,31 @@ export default function PromptInput({
                   adjustTextArea(e);
                 }}
                 value={promptInput}
-                className="cursor-text max-h-[100px] md:min-h-[40px] mx-2 md:mx-0 py-2 w-full text-[16px] md:text-md text-white bg-transparent placeholder:text-white/60 resize-none active:outline-none focus:outline-none flex-grow"
+                className="cursor-text max-h-[50vh] md:max-h-[350px] md:min-h-[40px] mx-2 md:mx-0 py-2 w-full text-[16px] md:text-md text-white bg-transparent placeholder:text-white/60 resize-none active:outline-none focus:outline-none flex-grow"
                 placeholder={"Send a message"}
               />
               {buttonDisabled ? (
                 <StopGenerationButton />
               ) : (
+                <>
                   <button
                     ref={formRef}
                     type="submit"
                     className="inline-flex justify-center rounded-2xl cursor-pointer text-white/60 hover:text-white group ml-4"
+                    data-tooltip-id="send-prompt"
+                    data-tooltip-content="Send prompt message to workspace"
+                    aria-label="Send prompt message to workspace"
                   >
                     <PaperPlaneRight className="w-7 h-7 my-3" weight="fill" />
                     <span className="sr-only">Send message</span>
                   </button>
+                  <Tooltip
+                    id="send-prompt"
+                    place="bottom"
+                    delayShow={300}
+                    className="tooltip !text-xs z-99"
+                  />
+                </>
               )}
             </div>
             <div className="flex justify-between py-3.5">

frontend/src/pages/GeneralSettings/AudioPreference/tts.jsx

@@ -47,7 +47,7 @@ export default function TextToSpeechProvider({ settings }) {
   const searchInputRef = useRef(null);
 
   const handleSubmit = async (e) => {
-    e.preventDefault();
+    e?.preventDefault();
     const form = e.target;
     const data = { TextToSpeechProvider: selectedProvider };
     const formData = new FormData(form);
@@ -110,10 +110,7 @@ export default function TextToSpeechProvider({ settings }) {
         </div>
         <div className="w-full justify-end flex">
           {hasChanges && (
-            <CTAButton
+            <CTAButton className="mt-3 mr-0 -mb-14 z-10">
-              onClick={() => handleSubmit()}
-              className="mt-3 mr-0 -mb-14 z-10"
-            >
               {saving ? "Saving..." : "Save changes"}
             </CTAButton>
           )}

frontend/src/pages/GeneralSettings/EmbeddingPreference/index.jsx

@@ -10,6 +10,8 @@ import LocalAiLogo from "@/media/llmprovider/localai.png";
 import OllamaLogo from "@/media/llmprovider/ollama.png";
 import LMStudioLogo from "@/media/llmprovider/lmstudio.png";
 import CohereLogo from "@/media/llmprovider/cohere.png";
+import VoyageAiLogo from "@/media/embeddingprovider/voyageai.png";
+
 import PreLoader from "@/components/Preloader";
 import ChangeWarningModal from "@/components/ChangeWarning";
 import OpenAiOptions from "@/components/EmbeddingSelection/OpenAiOptions";
@@ -19,6 +21,7 @@ import NativeEmbeddingOptions from "@/components/EmbeddingSelection/NativeEmbedd
 import OllamaEmbeddingOptions from "@/components/EmbeddingSelection/OllamaOptions";
 import LMStudioEmbeddingOptions from "@/components/EmbeddingSelection/LMStudioOptions";
 import CohereEmbeddingOptions from "@/components/EmbeddingSelection/CohereOptions";
+import VoyageAiOptions from "@/components/EmbeddingSelection/VoyageAiOptions";
 
 import EmbedderItem from "@/components/EmbeddingSelection/EmbedderItem";
 import { CaretUpDown, MagnifyingGlass, X } from "@phosphor-icons/react";
@@ -78,6 +81,13 @@ const EMBEDDERS = [
     options: (settings) => <CohereEmbeddingOptions settings={settings} />,
     description: "Run powerful embedding models from Cohere.",
   },
+  {
+    name: "Voyage AI",
+    value: "voyageai",
+    logo: VoyageAiLogo,
+    options: (settings) => <VoyageAiOptions settings={settings} />,
+    description: "Run powerful embedding models from Voyage AI.",
+  },
 ];
 
 export default function GeneralEmbeddingPreference() {

frontend/src/pages/GeneralSettings/LLMPreference/index.jsx

@@ -20,6 +20,7 @@ import GroqLogo from "@/media/llmprovider/groq.png";
 import KoboldCPPLogo from "@/media/llmprovider/koboldcpp.png";
 import TextGenWebUILogo from "@/media/llmprovider/text-generation-webui.png";
 import CohereLogo from "@/media/llmprovider/cohere.png";
+import LiteLLMLogo from "@/media/llmprovider/litellm.png";
 import PreLoader from "@/components/Preloader";
 import OpenAiOptions from "@/components/LLMSelection/OpenAiOptions";
 import GenericOpenAiOptions from "@/components/LLMSelection/GenericOpenAiOptions";
@@ -36,12 +37,13 @@ import PerplexityOptions from "@/components/LLMSelection/PerplexityOptions";
 import OpenRouterOptions from "@/components/LLMSelection/OpenRouterOptions";
 import GroqAiOptions from "@/components/LLMSelection/GroqAiOptions";
 import CohereAiOptions from "@/components/LLMSelection/CohereAiOptions";
+import KoboldCPPOptions from "@/components/LLMSelection/KoboldCPPOptions";
+import TextGenWebUIOptions from "@/components/LLMSelection/TextGenWebUIOptions";
+import LiteLLMOptions from "@/components/LLMSelection/LiteLLMOptions";
 
 import LLMItem from "@/components/LLMSelection/LLMItem";
 import { CaretUpDown, MagnifyingGlass, X } from "@phosphor-icons/react";
 import CTAButton from "@/components/lib/CTAButton";
-import KoboldCPPOptions from "@/components/LLMSelection/KoboldCPPOptions";
-import TextGenWebUIOptions from "@/components/LLMSelection/TextGenWebUIOptions";
 
 export const AVAILABLE_LLM_PROVIDERS = [
   {
@@ -184,6 +186,14 @@ export const AVAILABLE_LLM_PROVIDERS = [
     description: "Run Cohere's powerful Command models.",
     requiredConfig: ["CohereApiKey"],
   },
+  {
+    name: "LiteLLM",
+    value: "litellm",
+    logo: LiteLLMLogo,
+    options: (settings) => <LiteLLMOptions settings={settings} />,
+    description: "Run LiteLLM's OpenAI compatible proxy for various LLMs.",
+    requiredConfig: ["LiteLLMBasePath"],
+  },
   {
     name: "Generic OpenAI",
     value: "generic-openai",

frontend/src/pages/GeneralSettings/TranscriptionPreference/index.jsx

@@ -12,6 +12,23 @@ import LLMItem from "@/components/LLMSelection/LLMItem";
 import { CaretUpDown, MagnifyingGlass, X } from "@phosphor-icons/react";
 import CTAButton from "@/components/lib/CTAButton";
 
+const PROVIDERS = [
+  {
+    name: "OpenAI",
+    value: "openai",
+    logo: OpenAiLogo,
+    options: (settings) => <OpenAiWhisperOptions settings={settings} />,
+    description: "Leverage the OpenAI Whisper-large model using your API key.",
+  },
+  {
+    name: "AnythingLLM Built-In",
+    value: "local",
+    logo: AnythingLLMIcon,
+    options: (settings) => <NativeTranscriptionOptions settings={settings} />,
+    description: "Run a built-in whisper model on this instance privately.",
+  },
+];
+
 export default function TranscriptionModelPreference() {
   const [saving, setSaving] = useState(false);
   const [hasChanges, setHasChanges] = useState(false);
@@ -68,24 +85,6 @@ export default function TranscriptionModelPreference() {
     fetchKeys();
   }, []);
 
-  const PROVIDERS = [
-    {
-      name: "OpenAI",
-      value: "openai",
-      logo: OpenAiLogo,
-      options: <OpenAiWhisperOptions settings={settings} />,
-      description:
-        "Leverage the OpenAI Whisper-large model using your API key.",
-    },
-    {
-      name: "AnythingLLM Built-In",
-      value: "local",
-      logo: AnythingLLMIcon,
-      options: <NativeTranscriptionOptions settings={settings} />,
-      description: "Run a built-in whisper model on this instance privately.",
-    },
-  ];
-
   useEffect(() => {
     const filtered = PROVIDERS.filter((provider) =>
       provider.name.toLowerCase().includes(searchQuery.toLowerCase())
@@ -228,7 +227,7 @@ export default function TranscriptionModelPreference() {
                 {selectedProvider &&
                   PROVIDERS.find(
                     (provider) => provider.value === selectedProvider
-                  )?.options}
+                  )?.options(settings)}
               </div>
             </div>
           </form>

frontend/src/pages/OnboardingFlow/Steps/DataHandling/index.jsx

@@ -17,6 +17,8 @@ import OpenRouterLogo from "@/media/llmprovider/openrouter.jpeg";
 import GroqLogo from "@/media/llmprovider/groq.png";
 import KoboldCPPLogo from "@/media/llmprovider/koboldcpp.png";
 import TextGenWebUILogo from "@/media/llmprovider/text-generation-webui.png";
+import LiteLLMLogo from "@/media/llmprovider/litellm.png";
+
 import CohereLogo from "@/media/llmprovider/cohere.png";
 import ZillizLogo from "@/media/vectordbs/zilliz.png";
 import AstraDBLogo from "@/media/vectordbs/astraDB.png";
@@ -26,6 +28,8 @@ import LanceDbLogo from "@/media/vectordbs/lancedb.png";
 import WeaviateLogo from "@/media/vectordbs/weaviate.png";
 import QDrantLogo from "@/media/vectordbs/qdrant.png";
 import MilvusLogo from "@/media/vectordbs/milvus.png";
+import VoyageAiLogo from "@/media/embeddingprovider/voyageai.png";
+
 import React, { useState, useEffect } from "react";
 import paths from "@/utils/paths";
 import { useNavigate } from "react-router-dom";
@@ -168,6 +172,13 @@ export const LLM_SELECTION_PRIVACY = {
     ],
     logo: CohereLogo,
   },
+  litellm: {
+    name: "LiteLLM",
+    description: [
+      "Your model and chats are only accessible on the server running LiteLLM",
+    ],
+    logo: LiteLLMLogo,
+  },
 };
 
 export const VECTOR_DB_PRIVACY = {
@@ -283,6 +294,13 @@ export const EMBEDDING_ENGINE_PRIVACY = {
     ],
     logo: CohereLogo,
   },
+  voyageai: {
+    name: "Voyage AI",
+    description: [
+      "Data sent to Voyage AI's servers is shared according to the terms of service of voyageai.com.",
+    ],
+    logo: VoyageAiLogo,
+  },
 };
 
 export default function DataHandling({ setHeader, setForwardBtn, setBackBtn }) {

frontend/src/pages/OnboardingFlow/Steps/LLMPreference/index.jsx

@@ -16,6 +16,8 @@ import OpenRouterLogo from "@/media/llmprovider/openrouter.jpeg";
 import GroqLogo from "@/media/llmprovider/groq.png";
 import KoboldCPPLogo from "@/media/llmprovider/koboldcpp.png";
 import TextGenWebUILogo from "@/media/llmprovider/text-generation-webui.png";
+import LiteLLMLogo from "@/media/llmprovider/litellm.png";
+
 import CohereLogo from "@/media/llmprovider/cohere.png";
 import OpenAiOptions from "@/components/LLMSelection/OpenAiOptions";
 import GenericOpenAiOptions from "@/components/LLMSelection/GenericOpenAiOptions";
@@ -32,14 +34,15 @@ import PerplexityOptions from "@/components/LLMSelection/PerplexityOptions";
 import OpenRouterOptions from "@/components/LLMSelection/OpenRouterOptions";
 import GroqAiOptions from "@/components/LLMSelection/GroqAiOptions";
 import CohereAiOptions from "@/components/LLMSelection/CohereAiOptions";
+import KoboldCPPOptions from "@/components/LLMSelection/KoboldCPPOptions";
+import TextGenWebUIOptions from "@/components/LLMSelection/TextGenWebUIOptions";
+import LiteLLMOptions from "@/components/LLMSelection/LiteLLMOptions";
 
 import LLMItem from "@/components/LLMSelection/LLMItem";
 import System from "@/models/system";
 import paths from "@/utils/paths";
 import showToast from "@/utils/toast";
 import { useNavigate } from "react-router-dom";
-import KoboldCPPOptions from "@/components/LLMSelection/KoboldCPPOptions";
-import TextGenWebUIOptions from "@/components/LLMSelection/TextGenWebUIOptions";
 
 const TITLE = "LLM Preference";
 const DESCRIPTION =
@@ -162,6 +165,13 @@ const LLMS = [
     options: (settings) => <CohereAiOptions settings={settings} />,
     description: "Run Cohere's powerful Command models.",
   },
+  {
+    name: "LiteLLM",
+    value: "litellm",
+    logo: LiteLLMLogo,
+    options: (settings) => <LiteLLMOptions settings={settings} />,
+    description: "Run LiteLLM's OpenAI compatible proxy for various LLMs.",
+  },
   {
     name: "Generic OpenAI",
     value: "generic-openai",

frontend/src/pages/WorkspaceSettings/AgentConfig/SQLConnectorSelection/DBConnection.jsx

@@ -0,0 +1,47 @@
+import PostgreSQLLogo from "./icons/postgresql.png";
+import MySQLLogo from "./icons/mysql.png";
+import MSSQLLogo from "./icons/mssql.png";
+import { X } from "@phosphor-icons/react";
+
+export const DB_LOGOS = {
+  postgresql: PostgreSQLLogo,
+  mysql: MySQLLogo,
+  "sql-server": MSSQLLogo,
+};
+
+export default function DBConnection({ connection, onRemove }) {
+  const { database_id, engine } = connection;
+  function removeConfirmation() {
+    if (
+      !window.confirm(
+        `Delete ${database_id} from the list of available SQL connections? This cannot be undone.`
+      )
+    ) {
+      return false;
+    }
+    onRemove(database_id);
+  }
+
+  return (
+    <div className="flex gap-x-4 items-center">
+      <img
+        src={DB_LOGOS?.[engine] ?? null}
+        alt={`${engine} logo`}
+        className="w-10 h-10 rounded-md"
+      />
+      <div className="flex w-full items-center justify-between">
+        <div className="flex flex-col">
+          <div className="text-sm font-semibold text-white">{database_id}</div>
+          <div className="mt-1 text-xs text-[#D2D5DB]">{engine}</div>
+        </div>
+        <button
+          type="button"
+          onClick={removeConfirmation}
+          className="border-none text-white/40 hover:text-red-500"
+        >
+          <X size={24} />
+        </button>
+      </div>
+    </div>
+  );
+}

frontend/src/pages/WorkspaceSettings/AgentConfig/SQLConnectorSelection/NewConnectionModal.jsx

@@ -0,0 +1,271 @@
+import { useState } from "react";
+import { createPortal } from "react-dom";
+import ModalWrapper from "@/components/ModalWrapper";
+import { WarningOctagon, X } from "@phosphor-icons/react";
+import { DB_LOGOS } from "./DBConnection";
+
+function assembleConnectionString({
+  engine,
+  username = "",
+  password = "",
+  host = "",
+  port = "",
+  database = "",
+}) {
+  if ([username, password, host, database].every((i) => !!i) === false)
+    return `Please fill out all the fields above.`;
+  switch (engine) {
+    case "postgresql":
+      return `postgres://${username}:${password}@${host}:${port}/${database}`;
+    case "mysql":
+      return `mysql://${username}:${password}@${host}:${port}/${database}`;
+    case "sql-server":
+      return `mssql://${username}:${password}@${host}:${port}/${database}`;
+    default:
+      return null;
+  }
+}
+
+const DEFAULT_ENGINE = "postgresql";
+const DEFAULT_CONFIG = {
+  username: null,
+  password: null,
+  host: null,
+  port: null,
+  database: null,
+};
+
+export default function NewSQLConnection({ isOpen, closeModal, onSubmit }) {
+  const [engine, setEngine] = useState(DEFAULT_ENGINE);
+  const [config, setConfig] = useState(DEFAULT_CONFIG);
+  if (!isOpen) return null;
+
+  function handleClose() {
+    setEngine(DEFAULT_ENGINE);
+    setConfig(DEFAULT_CONFIG);
+    closeModal();
+  }
+
+  function onFormChange() {
+    const form = new FormData(document.getElementById("sql-connection-form"));
+    setConfig({
+      username: form.get("username").trim(),
+      password: form.get("password"),
+      host: form.get("host").trim(),
+      port: form.get("port").trim(),
+      database: form.get("database").trim(),
+    });
+  }
+
+  async function handleUpdate(e) {
+    e.preventDefault();
+    e.stopPropagation();
+    const form = new FormData(e.target);
+    onSubmit({
+      engine,
+      database_id: form.get("name"),
+      connectionString: assembleConnectionString({ engine, ...config }),
+    });
+    handleClose();
+    return false;
+  }
+
+  // Cannot do nested forms, it will cause all sorts of issues, so we portal this out
+  // to the parent container form so we don't have nested forms.
+  return createPortal(
+    <ModalWrapper isOpen={isOpen}>
+      <div className="relative w-1/3 max-h-full ">
+        <div className="relative bg-main-gradient rounded-xl shadow-[0_4px_14px_rgba(0,0,0,0.25)] max-h-[90vh] overflow-y-scroll no-scroll">
+          <div className="flex items-start justify-between p-4 border-b rounded-t border-gray-500/50">
+            <h3 className="text-xl font-semibold text-white">
+              New SQL Connection
+            </h3>
+            <button
+              onClick={handleClose}
+              type="button"
+              className="border-none transition-all duration-300 text-gray-400 bg-transparent hover:border-white/60 rounded-lg text-sm p-1.5 ml-auto inline-flex items-center bg-sidebar-button hover:bg-menu-item-selected-gradient hover:border-slate-100 hover:border-opacity-50 border-transparent border"
+              data-modal-hide="staticModal"
+            >
+              <X className="text-gray-300 text-lg" />
+            </button>
+          </div>
+
+          <form
+            id="sql-connection-form"
+            onSubmit={handleUpdate}
+            onChange={onFormChange}
+          >
+            <div className="py-[17px] px-[20px] flex flex-col gap-y-6">
+              <p className="text-sm text-white">
+                Add the connection information for your database below and it
+                will be available for future SQL agent calls.
+              </p>
+              <div className="flex flex-col w-full">
+                <div className="border border-red-800 bg-zinc-800 p-4 rounded-lg flex items-center gap-x-2 text-sm text-red-400">
+                  <WarningOctagon size={28} className="shrink-0" />
+                  <p>
+                    <b>WARNING:</b> The SQL agent has been <i>instructed</i> to
+                    only perform non-modifying queries. This <b>does not</b>{" "}
+                    prevent a hallucination from still deleting data. Only
+                    connect with a user who has <b>READ_ONLY</b> permissions.
+                  </p>
+                </div>
+
+                <label className="text-white text-sm font-semibold block my-4">
+                  Select your SQL engine
+                </label>
+                <div className="flex w-full flex-wrap gap-x-4">
+                  <DBEngine
+                    provider="postgresql"
+                    active={engine === "postgresql"}
+                    onClick={() => setEngine("postgresql")}
+                  />
+                  <DBEngine
+                    provider="mysql"
+                    active={engine === "mysql"}
+                    onClick={() => setEngine("mysql")}
+                  />
+                  <DBEngine
+                    provider="sql-server"
+                    active={engine === "sql-server"}
+                    onClick={() => setEngine("sql-server")}
+                  />
+                </div>
+              </div>
+
+              <div className="flex flex-col w-full">
+                <label className="text-white text-sm font-semibold block mb-4">
+                  Connection name
+                </label>
+                <input
+                  type="text"
+                  name="name"
+                  className="border-none bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+                  placeholder="a unique name to identify this SQL connection"
+                  required={true}
+                  autoComplete="off"
+                  spellCheck={false}
+                />
+              </div>
+
+              <div className="flex gap-x-2">
+                <div className="flex flex-col w-60">
+                  <label className="text-white text-sm font-semibold block mb-4">
+                    Database user
+                  </label>
+                  <input
+                    type="text"
+                    name="username"
+                    className="border-none bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+                    placeholder="root"
+                    required={true}
+                    autoComplete="off"
+                    spellCheck={false}
+                  />
+                </div>
+                <div className="flex flex-col w-60">
+                  <label className="text-white text-sm font-semibold block mb-4">
+                    Database user password
+                  </label>
+                  <input
+                    type="text"
+                    name="password"
+                    className="border-none bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+                    placeholder="password123"
+                    required={true}
+                    autoComplete="off"
+                    spellCheck={false}
+                  />
+                </div>
+              </div>
+
+              <div className="flex gap-x-2">
+                <div className="flex flex-col w-full">
+                  <label className="text-white text-sm font-semibold block mb-4">
+                    Server endpoint
+                  </label>
+                  <input
+                    type="text"
+                    name="host"
+                    className="border-none bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+                    placeholder="the hostname or endpoint for your database"
+                    required={true}
+                    autoComplete="off"
+                    spellCheck={false}
+                  />
+                </div>
+                <div className="flex flex-col w-30">
+                  <label className="text-white text-sm font-semibold block mb-4">
+                    Port
+                  </label>
+                  <input
+                    type="text"
+                    name="port"
+                    className="border-none bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+                    placeholder="3306"
+                    required={false}
+                    autoComplete="off"
+                    spellCheck={false}
+                  />
+                </div>
+              </div>
+
+              <div className="flex flex-col w-60">
+                <label className="text-white text-sm font-semibold block mb-4">
+                  Database
+                </label>
+                <input
+                  type="text"
+                  name="database"
+                  className="border-none bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-full p-2.5"
+                  placeholder="the database the agent will interact with"
+                  required={true}
+                  autoComplete="off"
+                  spellCheck={false}
+                />
+              </div>
+              <p className="text-white/40 text-sm">
+                {assembleConnectionString({ engine, ...config })}
+              </p>
+            </div>
+            <div className="flex w-full justify-between items-center p-3 space-x-2 border-t rounded-b border-gray-500/50">
+              <button
+                type="button"
+                onClick={handleClose}
+                className="border-none text-xs px-2 py-1 font-semibold rounded-lg bg-white hover:bg-transparent border-2 border-transparent hover:border-white hover:text-white h-[32px] w-fit -mr-8 whitespace-nowrap shadow-[0_4px_14px_rgba(0,0,0,0.25)]"
+              >
+                Cancel
+              </button>
+              <button
+                type="submit"
+                form="sql-connection-form"
+                className="border-none text-xs px-2 py-1 font-semibold rounded-lg bg-[#46C8FF] hover:bg-[#2C2F36] border-2 border-transparent hover:border-[#46C8FF] hover:text-white h-[32px] w-fit -mr-8 whitespace-nowrap shadow-[0_4px_14px_rgba(0,0,0,0.25)]"
+              >
+                Save connection
+              </button>
+            </div>
+          </form>
+        </div>
+      </div>
+    </ModalWrapper>,
+    document.getElementById("workspace-agent-settings-container")
+  );
+}
+
+function DBEngine({ provider, active, onClick }) {
+  return (
+    <button
+      type="button"
+      onClick={onClick}
+      className={`flex flex-col p-4 border border-white/40 bg-zinc-800 rounded-lg w-fit hover:bg-zinc-700 ${
+        active ? "!bg-blue-500/50" : ""
+      }`}
+    >
+      <img
+        src={DB_LOGOS[provider]}
+        className="h-[100px] rounded-md"
+        alt="PostgreSQL"
+      />
+    </button>
+  );
+}

frontend/src/pages/WorkspaceSettings/AgentConfig/SQLConnectorSelection/index.jsx

@@ -0,0 +1,109 @@
+import React, { useState } from "react";
+import DBConnection from "./DBConnection";
+import { Plus } from "@phosphor-icons/react";
+import NewSQLConnection from "./NewConnectionModal";
+import { useModal } from "@/hooks/useModal";
+
+export default function AgentSQLConnectorSelection({
+  skill,
+  settings,
+  toggleSkill,
+  enabled = false,
+}) {
+  const { isOpen, openModal, closeModal } = useModal();
+  const [connections, setConnections] = useState(
+    settings?.preferences?.agent_sql_connections || []
+  );
+
+  return (
+    <>
+      <div className="border-b border-white/40 pb-4">
+        <div className="flex flex-col">
+          <div className="flex w-full justify-between items-center">
+            <label htmlFor="name" className="block input-label">
+              SQL Agent
+            </label>
+            <label className="border-none relative inline-flex cursor-pointer items-center mt-2">
+              <input
+                type="checkbox"
+                className="peer sr-only"
+                checked={enabled}
+                onClick={() => toggleSkill(skill)}
+              />
+              <div className="pointer-events-none peer h-6 w-11 rounded-full bg-stone-400 after:absolute after:left-[2px] after:top-[2px] after:h-5 after:w-5 after:rounded-full after:shadow-xl after:border after:border-gray-600 after:bg-white after:box-shadow-md after:transition-all after:content-[''] peer-checked:bg-lime-300 peer-checked:after:translate-x-full peer-checked:after:border-white peer-focus:outline-none peer-focus:ring-4 peer-focus:ring-blue-800"></div>
+              <span className="ml-3 text-sm font-medium text-gray-900 dark:text-gray-300"></span>
+            </label>
+          </div>
+          <p className="text-white text-opacity-60 text-xs font-medium py-1.5">
+            Enable your agent to be able to leverage SQL to answer you questions
+            by connecting to various SQL database providers.
+          </p>
+        </div>
+        {enabled && (
+          <>
+            <input
+              name="system::agent_sql_connections"
+              type="hidden"
+              value={JSON.stringify(connections)}
+            />
+            <input
+              type="hidden"
+              value={JSON.stringify(
+                connections.filter((conn) => conn.action !== "remove")
+              )}
+            />
+            <div className="flex flex-col mt-2 gap-y-2">
+              <p className="text-white font-semibold text-sm">
+                Your database connections
+              </p>
+              <div className="flex flex-col gap-y-3">
+                {connections
+                  .filter((connection) => connection.action !== "remove")
+                  .map((connection) => (
+                    <DBConnection
+                      key={connection.database_id}
+                      connection={connection}
+                      onRemove={(databaseId) => {
+                        setConnections((prev) =>
+                          prev.map((conn) => {
+                            if (conn.database_id === databaseId)
+                              return { ...conn, action: "remove" };
+                            return conn;
+                          })
+                        );
+                      }}
+                    />
+                  ))}
+                <button
+                  type="button"
+                  onClick={openModal}
+                  className="w-fit relative flex h-[40px] items-center border-none hover:bg-slate-600/20 rounded-lg"
+                >
+                  <div className="flex w-full gap-x-2 items-center p-4">
+                    <div className="bg-zinc-600 p-2 rounded-lg h-[24px] w-[24px] flex items-center justify-center">
+                      <Plus
+                        weight="bold"
+                        size={14}
+                        className="shrink-0 text-slate-100"
+                      />
+                    </div>
+                    <p className="text-left text-slate-100 text-sm">
+                      New SQL connection
+                    </p>
+                  </div>
+                </button>
+              </div>
+            </div>
+          </>
+        )}
+      </div>
+      <NewSQLConnection
+        isOpen={isOpen}
+        closeModal={closeModal}
+        onSubmit={(newDb) =>
+          setConnections((prev) => [...prev, { action: "add", ...newDb }])
+        }
+      />
+    </>
+  );
+}

frontend/src/pages/WorkspaceSettings/AgentConfig/index.jsx

@@ -5,6 +5,7 @@ import { castToType } from "@/utils/types";
 import { useEffect, useRef, useState } from "react";
 import AgentLLMSelection from "./AgentLLMSelection";
 import AgentWebSearchSelection from "./WebSearchSelection";
+import AgentSQLConnectorSelection from "./SQLConnectorSelection";
 import GenericSkill from "./GenericSkill";
 import Admin from "@/models/admin";
 import * as Skeleton from "react-loading-skeleton";
@@ -205,6 +206,12 @@ function AvailableAgentSkills({ skills, settings, toggleAgentSkill }) {
           toggleSkill={toggleAgentSkill}
           enabled={skills.includes("web-browsing")}
         />
+        <AgentSQLConnectorSelection
+          skill="sql-agent"
+          settings={settings}
+          toggleSkill={toggleAgentSkill}
+          enabled={skills.includes("sql-agent")}
+        />
       </div>
     </div>
   );

frontend/src/pages/WorkspaceSettings/ChatSettings/ChatTemperatureSettings/index.jsx

@@ -20,19 +20,23 @@ export default function ChatTemperatureSettings({
           LLM Temperature
         </label>
         <p className="text-white text-opacity-60 text-xs font-medium py-1.5">
-          This setting controls how &quot;random&quot; or dynamic your chat
+          This setting controls how &quot;creative&quot; your LLM responses will
-          responses will be.
+          be.
           <br />
-          The higher the number (1.0 maximum) the more random and incoherent.
+          The higher the number the more creative. For some models this can lead
+          to incoherent responses when set too high.
           <br />
-          <i>Recommended: {defaults.temp}</i>
+          <br />
+          <i>
+            Most LLMs have various acceptable ranges of valid values. Consult
+            your LLM provider for that information.
+          </i>
         </p>
       </div>
       <input
         name="openAiTemp"
         type="number"
         min={0.0}
-        max={1.0}
         step={0.1}
         onWheel={(e) => e.target.blur()}
         defaultValue={workspace?.openAiTemp ?? defaults.temp}

frontend/src/pages/WorkspaceSettings/GeneralAppearance/index.jsx

@@ -2,7 +2,6 @@ import Workspace from "@/models/workspace";
 import { castToType } from "@/utils/types";
 import showToast from "@/utils/toast";
 import { useEffect, useRef, useState } from "react";
-import VectorCount from "./VectorCount";
 import WorkspaceName from "./WorkspaceName";
 import SuggestedChatMessages from "./SuggestedChatMessages";
 import DeleteWorkspace from "./DeleteWorkspace";
@@ -51,7 +50,6 @@ export default function GeneralInfo({ slug }) {
         onSubmit={handleUpdate}
         className="w-1/2 flex flex-col gap-y-6"
       >
-        <VectorCount reload={true} workspace={workspace} />
         <WorkspaceName
           key={workspace.slug}
           workspace={workspace}

frontend/src/pages/WorkspaceSettings/VectorDatabase/VectorCount/index.jsx

@@ -28,9 +28,6 @@ export default function VectorCount({ reload, workspace }) {
   return (
     <div>
       <h3 className="input-label">Number of vectors</h3>
-      <p className="text-white text-opacity-60 text-xs font-medium py-1">
-        Total number of vectors in your vector database.
-      </p>
       <p className="text-white text-opacity-60 text-sm font-medium">
         {totalVectors}
       </p>

frontend/src/pages/WorkspaceSettings/VectorDatabase/index.jsx

@@ -6,6 +6,7 @@ import VectorDBIdentifier from "./VectorDBIdentifier";
 import MaxContextSnippets from "./MaxContextSnippets";
 import DocumentSimilarityThreshold from "./DocumentSimilarityThreshold";
 import ResetDatabase from "./ResetDatabase";
+import VectorCount from "./VectorCount";
 
 export default function VectorDatabase({ workspace }) {
   const [hasChanges, setHasChanges] = useState(false);
@@ -38,7 +39,10 @@ export default function VectorDatabase({ workspace }) {
       onSubmit={handleUpdate}
       className="w-1/2 flex flex-col gap-y-6"
     >
+      <div className="flex items-start gap-x-5">
         <VectorDBIdentifier workspace={workspace} />
+        <VectorCount reload={true} workspace={workspace} />
+      </div>
       <MaxContextSnippets workspace={workspace} setHasChanges={setHasChanges} />
       <DocumentSimilarityThreshold
         workspace={workspace}

server/.env.example

@@ -79,6 +79,12 @@ JWT_SECRET="my-random-string-for-seeding" # Please generate random string at lea
 # GENERIC_OPEN_AI_MODEL_TOKEN_LIMIT=4096
 # GENERIC_OPEN_AI_API_KEY=sk-123abc
 
+# LLM_PROVIDER='litellm'
+# LITE_LLM_MODEL_PREF='gpt-3.5-turbo'
+# LITE_LLM_MODEL_TOKEN_LIMIT=4096
+# LITE_LLM_BASE_PATH='http://127.0.0.1:4000'
+# LITE_LLM_API_KEY='sk-123abc'
+
 # LLM_PROVIDER='cohere'
 # COHERE_API_KEY=
 # COHERE_MODEL_PREF='command-r'
@@ -115,6 +121,10 @@ JWT_SECRET="my-random-string-for-seeding" # Please generate random string at lea
 # COHERE_API_KEY=
 # EMBEDDING_MODEL_PREF='embed-english-v3.0'
 
+# EMBEDDING_ENGINE='voyageai'
+# VOYAGEAI_API_KEY=
+# EMBEDDING_MODEL_PREF='voyage-large-2-instruct'
+
 ###########################################
 ######## Vector Database Selection ########
 ###########################################

server/.gitignore

@@ -19,3 +19,6 @@ public/
 documents
 vector-cache
 yarn-error.log
+
+# Local SSL Certs for HTTPS
+sslcert

server/endpoints/admin.js

@@ -350,6 +350,8 @@ function adminEndpoints(app) {
           agent_search_provider:
             (await SystemSettings.get({ label: "agent_search_provider" }))
               ?.value || null,
+          agent_sql_connections:
+            await SystemSettings.brief.agent_sql_connections(),
           default_agent_skills:
             safeJsonParse(
               (await SystemSettings.get({ label: "default_agent_skills" }))

server/endpoints/api/workspace/index.js

@@ -447,6 +447,76 @@ function apiWorkspaceEndpoints(app) {
     }
   );
 
+  app.post(
+    "/v1/workspace/:slug/update-pin",
+    [validApiKey],
+    async (request, response) => {
+      /*
+      #swagger.tags = ['Workspaces']
+      #swagger.description = 'Add or remove pin from a document in a workspace by its unique slug.'
+      #swagger.path = '/workspace/{slug}/update-pin'
+      #swagger.parameters['slug'] = {
+          in: 'path',
+          description: 'Unique slug of workspace to find',
+          required: true,
+          type: 'string'
+      }
+      #swagger.requestBody = {
+        description: 'JSON object with the document path and pin status to update.',
+        required: true,
+        type: 'object',
+        content: {
+          "application/json": {
+            example: {
+              docPath: "custom-documents/my-pdf.pdf-hash.json",
+              pinStatus: true
+            }
+          }
+        }
+      }
+      #swagger.responses[200] = {
+        description: 'OK',
+        content: {
+          "application/json": {
+            schema: {
+              type: 'object',
+              example: {
+                message: 'Pin status updated successfully'
+              }
+            }
+          }
+        }
+      }
+      #swagger.responses[404] = {
+        description: 'Document not found'
+      }
+      #swagger.responses[500] = {
+        description: 'Internal Server Error'
+      }
+      */
+      try {
+        const { slug = null } = request.params;
+        const { docPath, pinStatus = false } = reqBody(request);
+        const workspace = await Workspace.get({ slug });
+
+        const document = await Document.get({
+          workspaceId: workspace.id,
+          docpath: docPath,
+        });
+        if (!document) return response.sendStatus(404).end();
+
+        await Document.update(document.id, { pinned: pinStatus });
+        return response
+          .status(200)
+          .json({ message: "Pin status updated successfully" })
+          .end();
+      } catch (error) {
+        console.error("Error processing the pin status update:", error);
+        return response.status(500).end();
+      }
+    }
+  );
+
   app.post(
     "/v1/workspace/:slug/chat",
     [validApiKey],
@@ -533,6 +603,7 @@ function apiWorkspaceEndpoints(app) {
         });
         response.status(200).json({ ...result });
       } catch (e) {
+        console.log(e.message, e);
         response.status(500).json({
           id: uuidv4(),
           type: "abort",
@@ -655,7 +726,7 @@ function apiWorkspaceEndpoints(app) {
         });
         response.end();
       } catch (e) {
-        console.error(e);
+        console.log(e.message, e);
         writeResponseChunk(response, {
           id: uuidv4(),
           type: "abort",

server/index.js

@@ -40,7 +40,12 @@ app.use(
   })
 );
 
-require("express-ws")(app);
+if (!!process.env.ENABLE_HTTPS) {
+  bootSSL(app, process.env.SERVER_PORT || 3001);
+} else {
+  require("express-ws")(app); // load WebSockets in non-SSL mode.
+}
+
 app.use("/api", apiRouter);
 systemEndpoints(apiRouter);
 extensionEndpoints(apiRouter);
@@ -113,8 +118,6 @@ app.all("*", function (_, response) {
   response.sendStatus(404);
 });
 
-if (!!process.env.ENABLE_HTTPS) {
+// In non-https mode we need to boot at the end since the server has not yet
-  bootSSL(app, process.env.SERVER_PORT || 3001);
+// started and is `.listen`ing.
-} else {
+if (!process.env.ENABLE_HTTPS) bootHTTP(app, process.env.SERVER_PORT || 3001);
-  bootHTTP(app, process.env.SERVER_PORT || 3001);
-}

server/models/systemSettings.js

@@ -1,6 +1,9 @@
 const path = require("path");
 const prisma = require("../utils/prisma");
-const { isValidUrl } = require("../utils/http");
+const { isValidUrl, safeJsonParse } = require("../utils/http");
+const { default: slugify } = require("slugify");
+const { v4 } = require("uuid");
+
 process.env.NODE_ENV === "development"
   ? require("dotenv").config({ path: `.env.${process.env.NODE_ENV}` })
   : require("dotenv").config({
@@ -28,6 +31,7 @@ const SystemSettings = {
     "text_splitter_chunk_overlap",
     "agent_search_provider",
     "default_agent_skills",
+    "agent_sql_connections",
   ],
   validations: {
     footer_data: (updates) => {
@@ -69,6 +73,7 @@ const SystemSettings = {
     },
     agent_search_provider: (update) => {
       try {
+        if (update === "none") return null;
         if (!["google-search-engine", "serper-dot-dev"].includes(update))
           throw new Error("Invalid SERP provider.");
         return String(update);
@@ -89,6 +94,22 @@ const SystemSettings = {
         return JSON.stringify([]);
       }
     },
+    agent_sql_connections: async (updates) => {
+      const existingConnections = safeJsonParse(
+        (await SystemSettings.get({ label: "agent_sql_connections" }))?.value,
+        []
+      );
+      try {
+        const updatedConnections = mergeConnections(
+          existingConnections,
+          safeJsonParse(updates, [])
+        );
+        return JSON.stringify(updatedConnections);
+      } catch (e) {
+        console.error(`Failed to merge connections`);
+        return JSON.stringify(existingConnections ?? []);
+      }
+    },
   },
   currentSettings: async function () {
     const { hasVectorCachedFiles } = require("../utils/files");
@@ -134,6 +155,8 @@ const SystemSettings = {
       // - then it can be shared.
       // --------------------------------------------------------
       WhisperProvider: process.env.WHISPER_PROVIDER || "local",
+      WhisperModelPref:
+        process.env.WHISPER_MODEL_PREF || "Xenova/whisper-small",
 
       // --------------------------------------------------------
       // TTS/STT  Selection Settings & Configs
@@ -208,12 +231,19 @@ const SystemSettings = {
   // that takes no user input for the keys being modified.
   _updateSettings: async function (updates = {}) {
     try {
-      const updatePromises = Object.keys(updates).map((key) => {
+      const updatePromises = [];
-        const validatedValue = this.validations.hasOwnProperty(key)
+      for (const key of Object.keys(updates)) {
-          ? this.validations[key](updates[key])
+        let validatedValue = updates[key];
-          : updates[key];
+        if (this.validations.hasOwnProperty(key)) {
+          if (this.validations[key].constructor.name === "AsyncFunction") {
+            validatedValue = await this.validations[key](updates[key]);
+          } else {
+            validatedValue = this.validations[key](updates[key]);
+          }
+        }
 
-        return prisma.system_settings.upsert({
+        updatePromises.push(
+          prisma.system_settings.upsert({
             where: { label: key },
             update: {
               value: validatedValue === null ? null : String(validatedValue),
@@ -222,8 +252,9 @@ const SystemSettings = {
               label: key,
               value: validatedValue === null ? null : String(validatedValue),
             },
-        });
+          })
-      });
+        );
+      }
 
       await Promise.all(updatePromises);
       return { success: true, error: null };
@@ -328,6 +359,8 @@ const SystemSettings = {
       // Gemini Keys
       GeminiLLMApiKey: !!process.env.GEMINI_API_KEY,
       GeminiLLMModelPref: process.env.GEMINI_LLM_MODEL_PREF || "gemini-pro",
+      GeminiSafetySetting:
+        process.env.GEMINI_SAFETY_SETTING || "BLOCK_MEDIUM_AND_ABOVE",
 
       // LMStudio Keys
       LMStudioBasePath: process.env.LMSTUDIO_BASE_PATH,
@@ -384,6 +417,12 @@ const SystemSettings = {
       TextGenWebUITokenLimit: process.env.TEXT_GEN_WEB_UI_MODEL_TOKEN_LIMIT,
       TextGenWebUIAPIKey: !!process.env.TEXT_GEN_WEB_UI_API_KEY,
 
+      // LiteLLM Keys
+      LiteLLMModelPref: process.env.LITE_LLM_MODEL_PREF,
+      LiteLLMTokenLimit: process.env.LITE_LLM_MODEL_TOKEN_LIMIT,
+      LiteLLMBasePath: process.env.LITE_LLM_BASE_PATH,
+      LiteLLMApiKey: !!process.env.LITE_LLM_API_KEY,
+
       // Generic OpenAI Keys
       GenericOpenAiBasePath: process.env.GENERIC_OPEN_AI_BASE_PATH,
       GenericOpenAiModelPref: process.env.GENERIC_OPEN_AI_MODEL_PREF,
@@ -394,8 +433,63 @@ const SystemSettings = {
       // Cohere API Keys
       CohereApiKey: !!process.env.COHERE_API_KEY,
       CohereModelPref: process.env.COHERE_MODEL_PREF,
+
+      // VoyageAi API Keys
+      VoyageAiApiKey: !!process.env.VOYAGEAI_API_KEY,
     };
   },
+
+  // For special retrieval of a key setting that does not expose any credential information
+  brief: {
+    agent_sql_connections: async function () {
+      const setting = await SystemSettings.get({
+        label: "agent_sql_connections",
+      });
+      if (!setting) return [];
+      return safeJsonParse(setting.value, []).map((dbConfig) => {
+        const { connectionString, ...rest } = dbConfig;
+        return rest;
+      });
+    },
+  },
 };
 
+function mergeConnections(existingConnections = [], updates = []) {
+  let updatedConnections = [...existingConnections];
+  const existingDbIds = existingConnections.map((conn) => conn.database_id);
+
+  // First remove all 'action:remove' candidates from existing connections.
+  const toRemove = updates
+    .filter((conn) => conn.action === "remove")
+    .map((conn) => conn.database_id);
+  updatedConnections = updatedConnections.filter(
+    (conn) => !toRemove.includes(conn.database_id)
+  );
+
+  // Next add all 'action:add' candidates into the updatedConnections; We DO NOT validate the connection strings.
+  // but we do validate their database_id is unique.
+  updates
+    .filter((conn) => conn.action === "add")
+    .forEach((update) => {
+      if (!update.connectionString) return; // invalid connection string
+
+      // Remap name to be unique to entire set.
+      if (existingDbIds.includes(update.database_id)) {
+        update.database_id = slugify(
+          `${update.database_id}-${v4().slice(0, 4)}`
+        );
+      } else {
+        update.database_id = slugify(update.database_id);
+      }
+
+      updatedConnections.push({
+        engine: update.engine,
+        database_id: update.database_id,
+        connectionString: update.connectionString,
+      });
+    });
+
+  return updatedConnections;
+}
+
 module.exports.SystemSettings = SystemSettings;

server/package.json

@@ -58,11 +58,14 @@
     "langchain": "0.1.36",
     "mime": "^3.0.0",
     "moment": "^2.29.4",
+    "mssql": "^10.0.2",
     "multer": "^1.4.5-lts.1",
+    "mysql2": "^3.9.7",
     "node-html-markdown": "^1.3.0",
     "node-llama-cpp": "^2.8.0",
     "ollama": "^0.5.0",
     "openai": "4.38.5",
+    "pg": "^8.11.5",
     "pinecone-client": "^1.1.0",
     "pluralize": "^8.0.0",
     "posthog-node": "^3.1.1",
@@ -72,6 +75,7 @@
     "sqlite3": "^5.1.6",
     "swagger-autogen": "^2.23.5",
     "swagger-ui-express": "^5.0.0",
+    "url-pattern": "^1.0.3",
     "uuid": "^9.0.0",
     "uuid-apikey": "^1.5.3",
     "vectordb": "0.4.11",

server/swagger/openapi.json

@@ -2000,6 +2000,69 @@
         }
       }
     },
+    "/workspace/{slug}/update-pin": {
+      "post": {
+        "tags": [
+          "Workspaces"
+        ],
+        "description": "Add or remove pin from a document in a workspace by its unique slug.",
+        "parameters": [
+          {
+            "name": "slug",
+            "in": "path",
+            "required": true,
+            "schema": {
+              "type": "string"
+            },
+            "description": "Unique slug of workspace to find"
+          },
+          {
+            "name": "Authorization",
+            "in": "header",
+            "schema": {
+              "type": "string"
+            }
+          }
+        ],
+        "responses": {
+          "200": {
+            "description": "OK",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "type": "object",
+                  "example": {
+                    "message": "Pin status updated successfully"
+                  }
+                }
+              }
+            }
+          },
+          "403": {
+            "description": "Forbidden"
+          },
+          "404": {
+            "description": "Document not found"
+          },
+          "500": {
+            "description": "Internal Server Error"
+          }
+        },
+        "requestBody": {
+          "description": "JSON object with the document path and pin status to update.",
+          "required": true,
+          "type": "object",
+          "content": {
+            "application/json": {
+              "example": {
+                "docPath": "custom-documents/my-pdf.pdf-hash.json",
+                "pinStatus": true
+              }
+            }
+          }
+        }
+      }
+    },
     "/v1/workspace/{slug}/chat": {
       "post": {
         "tags": [

server/utils/AiProviders/anthropic/index.js

@@ -3,6 +3,7 @@ const {
   writeResponseChunk,
   clientAbortedHandler,
 } = require("../../helpers/chat/responses");
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 
 class AnthropicLLM {
   constructor(embedder = null, modelPreference = null) {
@@ -23,11 +24,7 @@ class AnthropicLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      throw new Error(
-        "INVALID ANTHROPIC SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use Anthropic as your LLM."
-      );
-    this.embedder = embedder;
     this.defaultTemp = 0.7;
   }
 

server/utils/AiProviders/azureOpenAi/index.js

@@ -1,4 +1,4 @@
-const { AzureOpenAiEmbedder } = require("../../EmbeddingEngines/azureOpenAi");
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 const {
   writeResponseChunk,
   clientAbortedHandler,
@@ -23,11 +23,7 @@ class AzureOpenAiLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      console.warn(
-        "No embedding provider defined for AzureOpenAiLLM - falling back to AzureOpenAiEmbedder for embedding!"
-      );
-    this.embedder = !embedder ? new AzureOpenAiEmbedder() : embedder;
     this.defaultTemp = 0.7;
   }
 

server/utils/AiProviders/cohere/index.js

@@ -19,7 +19,8 @@ class CohereLLM {
       system: this.promptWindowLimit() * 0.15,
       user: this.promptWindowLimit() * 0.7,
     };
-    this.embedder = !!embedder ? embedder : new NativeEmbedder();
+
+    this.embedder = embedder ?? new NativeEmbedder();
   }
 
   #appendContext(contextTexts = []) {

server/utils/AiProviders/gemini/index.js

@@ -1,3 +1,4 @@
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 const {
   writeResponseChunk,
   clientAbortedHandler,
@@ -26,12 +27,9 @@ class GeminiLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      throw new Error(
-        "INVALID GEMINI LLM SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use Gemini as your LLM."
-      );
-    this.embedder = embedder;
     this.defaultTemp = 0.7; // not used for Gemini
+    this.safetyThreshold = this.#fetchSafetyThreshold();
   }
 
   #appendContext(contextTexts = []) {
@@ -46,6 +44,41 @@ class GeminiLLM {
     );
   }
 
+  // BLOCK_NONE can be a special candidate for some fields
+  // https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/configure-safety-attributes#how_to_remove_automated_response_blocking_for_select_safety_attributes
+  // so if you are wondering why BLOCK_NONE still failed, the link above will explain why.
+  #fetchSafetyThreshold() {
+    const threshold =
+      process.env.GEMINI_SAFETY_SETTING ?? "BLOCK_MEDIUM_AND_ABOVE";
+    const safetyThresholds = [
+      "BLOCK_NONE",
+      "BLOCK_ONLY_HIGH",
+      "BLOCK_MEDIUM_AND_ABOVE",
+      "BLOCK_LOW_AND_ABOVE",
+    ];
+    return safetyThresholds.includes(threshold)
+      ? threshold
+      : "BLOCK_MEDIUM_AND_ABOVE";
+  }
+
+  #safetySettings() {
+    return [
+      {
+        category: "HARM_CATEGORY_HATE_SPEECH",
+        threshold: this.safetyThreshold,
+      },
+      {
+        category: "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+        threshold: this.safetyThreshold,
+      },
+      { category: "HARM_CATEGORY_HARASSMENT", threshold: this.safetyThreshold },
+      {
+        category: "HARM_CATEGORY_DANGEROUS_CONTENT",
+        threshold: this.safetyThreshold,
+      },
+    ];
+  }
+
   streamingEnabled() {
     return "streamGetChatCompletion" in this;
   }
@@ -146,6 +179,7 @@ class GeminiLLM {
     )?.content;
     const chatThread = this.gemini.startChat({
       history: this.formatMessages(messages),
+      safetySettings: this.#safetySettings(),
     });
     const result = await chatThread.sendMessage(prompt);
     const response = result.response;
@@ -167,6 +201,7 @@ class GeminiLLM {
     )?.content;
     const chatThread = this.gemini.startChat({
       history: this.formatMessages(messages),
+      safetySettings: this.#safetySettings(),
     });
     const responseStream = await chatThread.sendMessageStream(prompt);
     if (!responseStream.stream)

server/utils/AiProviders/genericOpenAi/index.js

@@ -2,6 +2,7 @@ const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 const {
   handleDefaultStreamResponseV2,
 } = require("../../helpers/chat/responses");
+const { toValidNumber } = require("../../http");
 
 class GenericOpenAiLLM {
   constructor(embedder = null, modelPreference = null) {
@@ -18,7 +19,9 @@ class GenericOpenAiLLM {
     });
     this.model =
       modelPreference ?? process.env.GENERIC_OPEN_AI_MODEL_PREF ?? null;
-    this.maxTokens = process.env.GENERIC_OPEN_AI_MAX_TOKENS ?? 1024;
+    this.maxTokens = process.env.GENERIC_OPEN_AI_MAX_TOKENS
+      ? toValidNumber(process.env.GENERIC_OPEN_AI_MAX_TOKENS, 1024)
+      : 1024;
     if (!this.model)
       throw new Error("GenericOpenAI must have a valid model set.");
     this.limits = {
@@ -27,11 +30,7 @@ class GenericOpenAiLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      console.warn(
-        "No embedding provider defined for GenericOpenAiLLM - falling back to NativeEmbedder for embedding!"
-      );
-    this.embedder = !embedder ? new NativeEmbedder() : embedder;
     this.defaultTemp = 0.7;
     this.log(`Inference API: ${this.basePath} Model: ${this.model}`);
   }

server/utils/AiProviders/groq/index.js

@@ -20,7 +20,7 @@ class GroqLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    this.embedder = !embedder ? new NativeEmbedder() : embedder;
+    this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
   }
 

server/utils/AiProviders/huggingface/index.js

@@ -1,5 +1,4 @@
 const { NativeEmbedder } = require("../../EmbeddingEngines/native");
-const { OpenAiEmbedder } = require("../../EmbeddingEngines/openAi");
 const {
   handleDefaultStreamResponseV2,
 } = require("../../helpers/chat/responses");
@@ -26,11 +25,7 @@ class HuggingFaceLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      console.warn(
-        "No embedding provider defined for HuggingFaceLLM - falling back to Native for embedding!"
-      );
-    this.embedder = !embedder ? new OpenAiEmbedder() : new NativeEmbedder();
     this.defaultTemp = 0.2;
   }
 

server/utils/AiProviders/koboldCPP/index.js

@@ -26,11 +26,7 @@ class KoboldCPPLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      console.warn(
-        "No embedding provider defined for KoboldCPPLLM - falling back to NativeEmbedder for embedding!"
-      );
-    this.embedder = !embedder ? new NativeEmbedder() : embedder;
     this.defaultTemp = 0.7;
     this.log(`Inference API: ${this.basePath} Model: ${this.model}`);
   }

server/utils/AiProviders/liteLLM/index.js

@@ -0,0 +1,174 @@
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
+const {
+  writeResponseChunk,
+  clientAbortedHandler,
+} = require("../../helpers/chat/responses");
+
+class LiteLLM {
+  constructor(embedder = null, modelPreference = null) {
+    const { OpenAI: OpenAIApi } = require("openai");
+    if (!process.env.LITE_LLM_BASE_PATH)
+      throw new Error(
+        "LiteLLM must have a valid base path to use for the api."
+      );
+
+    this.basePath = process.env.LITE_LLM_BASE_PATH;
+    this.openai = new OpenAIApi({
+      baseURL: this.basePath,
+      apiKey: process.env.LITE_LLM_API_KEY ?? null,
+    });
+    this.model = modelPreference ?? process.env.LITE_LLM_MODEL_PREF ?? null;
+    this.maxTokens = process.env.LITE_LLM_MODEL_TOKEN_LIMIT ?? 1024;
+    if (!this.model) throw new Error("LiteLLM must have a valid model set.");
+    this.limits = {
+      history: this.promptWindowLimit() * 0.15,
+      system: this.promptWindowLimit() * 0.15,
+      user: this.promptWindowLimit() * 0.7,
+    };
+
+    this.embedder = embedder ?? new NativeEmbedder();
+    this.defaultTemp = 0.7;
+    this.log(`Inference API: ${this.basePath} Model: ${this.model}`);
+  }
+
+  log(text, ...args) {
+    console.log(`\x1b[36m[${this.constructor.name}]\x1b[0m ${text}`, ...args);
+  }
+
+  #appendContext(contextTexts = []) {
+    if (!contextTexts || !contextTexts.length) return "";
+    return (
+      "\nContext:\n" +
+      contextTexts
+        .map((text, i) => {
+          return `[CONTEXT ${i}]:\n${text}\n[END CONTEXT ${i}]\n\n`;
+        })
+        .join("")
+    );
+  }
+
+  streamingEnabled() {
+    return "streamGetChatCompletion" in this;
+  }
+
+  // Ensure the user set a value for the token limit
+  // and if undefined - assume 4096 window.
+  promptWindowLimit() {
+    const limit = process.env.LITE_LLM_MODEL_TOKEN_LIMIT || 4096;
+    if (!limit || isNaN(Number(limit)))
+      throw new Error("No token context limit was set.");
+    return Number(limit);
+  }
+
+  // Short circuit since we have no idea if the model is valid or not
+  // in pre-flight for generic endpoints
+  isValidChatCompletionModel(_modelName = "") {
+    return true;
+  }
+
+  constructPrompt({
+    systemPrompt = "",
+    contextTexts = [],
+    chatHistory = [],
+    userPrompt = "",
+  }) {
+    const prompt = {
+      role: "system",
+      content: `${systemPrompt}${this.#appendContext(contextTexts)}`,
+    };
+    return [prompt, ...chatHistory, { role: "user", content: userPrompt }];
+  }
+
+  async isSafe(_input = "") {
+    // Not implemented so must be stubbed
+    return { safe: true, reasons: [] };
+  }
+
+  async getChatCompletion(messages = null, { temperature = 0.7 }) {
+    const result = await this.openai.chat.completions
+      .create({
+        model: this.model,
+        messages,
+        temperature,
+        max_tokens: parseInt(this.maxTokens), // LiteLLM requires int
+      })
+      .catch((e) => {
+        throw new Error(e.response.data.error.message);
+      });
+
+    if (!result.hasOwnProperty("choices") || result.choices.length === 0)
+      return null;
+    return result.choices[0].message.content;
+  }
+
+  async streamGetChatCompletion(messages = null, { temperature = 0.7 }) {
+    const streamRequest = await this.openai.chat.completions.create({
+      model: this.model,
+      stream: true,
+      messages,
+      temperature,
+      max_tokens: parseInt(this.maxTokens), // LiteLLM requires int
+    });
+    return streamRequest;
+  }
+
+  handleStream(response, stream, responseProps) {
+    const { uuid = uuidv4(), sources = [] } = responseProps;
+
+    return new Promise(async (resolve) => {
+      let fullText = "";
+
+      const handleAbort = () => clientAbortedHandler(resolve, fullText);
+      response.on("close", handleAbort);
+
+      for await (const chunk of stream) {
+        const message = chunk?.choices?.[0];
+        const token = message?.delta?.content;
+
+        if (token) {
+          fullText += token;
+          writeResponseChunk(response, {
+            uuid,
+            sources: [],
+            type: "textResponseChunk",
+            textResponse: token,
+            close: false,
+            error: false,
+          });
+        }
+
+        // LiteLLM does not give a finish reason in stream until the final chunk
+        if (message.finish_reason || message.finish_reason === "stop") {
+          writeResponseChunk(response, {
+            uuid,
+            sources,
+            type: "textResponseChunk",
+            textResponse: "",
+            close: true,
+            error: false,
+          });
+          response.removeListener("close", handleAbort);
+          resolve(fullText);
+        }
+      }
+    });
+  }
+
+  // Simple wrapper for dynamic embedder & normalize interface for all LLM implementations
+  async embedTextInput(textInput) {
+    return await this.embedder.embedTextInput(textInput);
+  }
+  async embedChunks(textChunks = []) {
+    return await this.embedder.embedChunks(textChunks);
+  }
+
+  async compressMessages(promptArgs = {}, rawHistory = []) {
+    const { messageArrayCompressor } = require("../../helpers/chat");
+    const messageArray = this.constructPrompt(promptArgs);
+    return await messageArrayCompressor(this, messageArray, rawHistory);
+  }
+}
+
+module.exports = {
+  LiteLLM,
+};

server/utils/AiProviders/lmStudio/index.js

@@ -1,3 +1,4 @@
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 const {
   handleDefaultStreamResponseV2,
 } = require("../../helpers/chat/responses");
@@ -27,11 +28,7 @@ class LMStudioLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      throw new Error(
-        "INVALID LM STUDIO SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use LMStudio as your LLM."
-      );
-    this.embedder = embedder;
     this.defaultTemp = 0.7;
   }
 

server/utils/AiProviders/localAi/index.js

@@ -1,3 +1,4 @@
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 const {
   handleDefaultStreamResponseV2,
 } = require("../../helpers/chat/responses");
@@ -19,11 +20,7 @@ class LocalAiLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      throw new Error(
-        "INVALID LOCAL AI SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use LocalAI as your LLM."
-      );
-    this.embedder = embedder;
     this.defaultTemp = 0.7;
   }
 

server/utils/AiProviders/mistral/index.js

@@ -1,3 +1,4 @@
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 const {
   handleDefaultStreamResponseV2,
 } = require("../../helpers/chat/responses");
@@ -20,11 +21,7 @@ class MistralLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      console.warn(
-        "No embedding provider defined for MistralLLM - falling back to OpenAiEmbedder for embedding!"
-      );
-    this.embedder = embedder;
     this.defaultTemp = 0.0;
   }
 

server/utils/AiProviders/native/index.js

@@ -23,7 +23,7 @@ class NativeLLM {
       system: this.promptWindowLimit() * 0.15,
       user: this.promptWindowLimit() * 0.7,
     };
-    this.embedder = embedder || new NativeEmbedder();
+    this.embedder = embedder ?? new NativeEmbedder();
     this.cacheDir = path.resolve(
       process.env.STORAGE_DIR
         ? path.resolve(process.env.STORAGE_DIR, "models", "downloaded")

server/utils/AiProviders/ollama/index.js

@@ -3,6 +3,7 @@ const {
   writeResponseChunk,
   clientAbortedHandler,
 } = require("../../helpers/chat/responses");
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 
 // Docs: https://github.com/jmorganca/ollama/blob/main/docs/api.md
 class OllamaAILLM {
@@ -18,11 +19,7 @@ class OllamaAILLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      throw new Error(
-        "INVALID OLLAMA SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use Ollama as your LLM."
-      );
-    this.embedder = embedder;
     this.defaultTemp = 0.7;
   }
 

server/utils/AiProviders/openAi/index.js

@@ -1,4 +1,4 @@
-const { OpenAiEmbedder } = require("../../EmbeddingEngines/openAi");
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 const {
   handleDefaultStreamResponseV2,
 } = require("../../helpers/chat/responses");
@@ -18,11 +18,7 @@ class OpenAiLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = embedder ?? new NativeEmbedder();
-      console.warn(
-        "No embedding provider defined for OpenAiLLM - falling back to OpenAiEmbedder for embedding!"
-      );
-    this.embedder = !embedder ? new OpenAiEmbedder() : embedder;
     this.defaultTemp = 0.7;
   }
 

server/utils/AiProviders/openRouter/index.js

@@ -36,7 +36,7 @@ class OpenRouterLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    this.embedder = !embedder ? new NativeEmbedder() : embedder;
+    this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
 
     if (!fs.existsSync(cacheFolder))

server/utils/AiProviders/perplexity/index.js

@@ -28,7 +28,7 @@ class PerplexityLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    this.embedder = !embedder ? new NativeEmbedder() : embedder;
+    this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
   }
 

server/utils/AiProviders/textGenWebUI/index.js

@@ -23,7 +23,7 @@ class TextGenWebUILLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    this.embedder = !embedder ? new NativeEmbedder() : embedder;
+    this.embedder = embedder ?? new NativeEmbedder();
     this.defaultTemp = 0.7;
     this.log(`Inference API: ${this.basePath} Model: ${this.model}`);
   }

server/utils/AiProviders/togetherAi/index.js

@@ -1,3 +1,4 @@
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
 const {
   handleDefaultStreamResponseV2,
 } = require("../../helpers/chat/responses");
@@ -23,11 +24,7 @@ class TogetherAiLLM {
       user: this.promptWindowLimit() * 0.7,
     };
 
-    if (!embedder)
+    this.embedder = !embedder ? new NativeEmbedder() : embedder;
-      throw new Error(
-        "INVALID TOGETHER AI SETUP. No embedding engine has been set. Go to instance settings and set up an embedding interface to use Together AI as your LLM."
-      );
-    this.embedder = embedder;
     this.defaultTemp = 0.7;
   }
 

server/utils/EmbeddingEngines/voyageAi/index.js

@@ -0,0 +1,45 @@
+class VoyageAiEmbedder {
+  constructor() {
+    if (!process.env.VOYAGEAI_API_KEY)
+      throw new Error("No Voyage AI API key was set.");
+
+    const {
+      VoyageEmbeddings,
+    } = require("@langchain/community/embeddings/voyage");
+    const voyage = new VoyageEmbeddings({
+      apiKey: process.env.VOYAGEAI_API_KEY,
+    });
+
+    this.voyage = voyage;
+    this.model = process.env.EMBEDDING_MODEL_PREF || "voyage-large-2-instruct";
+
+    // Limit of how many strings we can process in a single pass to stay with resource or network limits
+    this.batchSize = 128; // Voyage AI's limit per request is 128 https://docs.voyageai.com/docs/rate-limits#use-larger-batches
+    this.embeddingMaxChunkLength = 4000; // https://docs.voyageai.com/docs/embeddings - assume a token is roughly 4 letters with some padding
+  }
+
+  async embedTextInput(textInput) {
+    const result = await this.voyage.embedDocuments(
+      Array.isArray(textInput) ? textInput : [textInput],
+      { modelName: this.model }
+    );
+    return result || [];
+  }
+
+  async embedChunks(textChunks = []) {
+    try {
+      const embeddings = await this.voyage.embedDocuments(textChunks, {
+        modelName: this.model,
+        batchSize: this.batchSize,
+      });
+      return embeddings;
+    } catch (error) {
+      console.error("Voyage AI Failed to embed:", error);
+      throw error;
+    }
+  }
+}
+
+module.exports = {
+  VoyageAiEmbedder,
+};

server/utils/TextSplitter/index.js

@@ -17,6 +17,7 @@ class TextSplitter {
       Config: {
         chunkSize: number,
         chunkOverlap: number,
+        chunkHeaderMeta: object | null, // Gets appended to top of each chunk as metadata
       }
       ------
     */
@@ -44,6 +45,18 @@ class TextSplitter {
     return prefValue > limit ? limit : prefValue;
   }
 
+  stringifyHeader() {
+    if (!this.config.chunkHeaderMeta) return null;
+    let content = "";
+    Object.entries(this.config.chunkHeaderMeta).map(([key, value]) => {
+      if (!key || !value) return;
+      content += `${key}: ${value}\n`;
+    });
+
+    if (!content) return null;
+    return `<document_metadata>\n${content}</document_metadata>\n\n`;
+  }
+
   #setSplitter(config = {}) {
     // if (!config?.splitByFilename) {// TODO do something when specific extension is present? }
     return new RecursiveSplitter({
@@ -51,6 +64,7 @@ class TextSplitter {
       chunkOverlap: isNaN(config?.chunkOverlap)
         ? 20
         : Number(config?.chunkOverlap),
+      chunkHeader: this.stringifyHeader(),
     });
   }
 
@@ -61,11 +75,12 @@ class TextSplitter {
 
 // Wrapper for Langchain default RecursiveCharacterTextSplitter class.
 class RecursiveSplitter {
-  constructor({ chunkSize, chunkOverlap }) {
+  constructor({ chunkSize, chunkOverlap, chunkHeader = null }) {
     const {
       RecursiveCharacterTextSplitter,
     } = require("@langchain/textsplitters");
     this.log(`Will split with`, { chunkSize, chunkOverlap });
+    this.chunkHeader = chunkHeader;
     this.engine = new RecursiveCharacterTextSplitter({
       chunkSize,
       chunkOverlap,
@@ -77,7 +92,14 @@ class RecursiveSplitter {
   }
 
   async _splitText(documentText) {
-    return this.engine.splitText(documentText);
+    if (!this.chunkHeader) return this.engine.splitText(documentText);
+    const strings = await this.engine.splitText(documentText);
+    const documents = await this.engine.createDocuments(strings, [], {
+      chunkHeader: this.chunkHeader,
+    });
+    return documents
+      .filter((doc) => !!doc.pageContent)
+      .map((doc) => doc.pageContent);
   }
 }
 

server/utils/agents/aibitat/index.js

@@ -498,6 +498,17 @@ Only return the role.
     return availableNodes[Math.floor(Math.random() * availableNodes.length)];
   }
 
+  /**
+   *
+   * @param {string} pluginName this name of the plugin being called
+   * @returns string of the plugin to be called compensating for children denoted by # in the string.
+   * eg: sql-agent:list-database-connections
+   */
+  #parseFunctionName(pluginName = "") {
+    if (!pluginName.includes("#")) return pluginName;
+    return pluginName.split("#")[1];
+  }
+
   /**
    * Check if the chat has reached the maximum number of rounds.
    */
@@ -550,7 +561,7 @@ ${this.getHistory({ to: route.to })
 
     // get the functions that the node can call
     const functions = fromConfig.functions
-      ?.map((name) => this.functions.get(name))
+      ?.map((name) => this.functions.get(this.#parseFunctionName(name)))
       .filter((a) => !!a);
 
     const provider = this.getProviderForConfig({

server/utils/agents/aibitat/plugins/index.js

@@ -6,6 +6,7 @@ const { saveFileInBrowser } = require("./save-file-browser.js");
 const { chatHistory } = require("./chat-history.js");
 const { memory } = require("./memory.js");
 const { rechart } = require("./rechart.js");
+const { sqlAgent } = require("./sql-agent/index.js");
 
 module.exports = {
   webScraping,
@@ -16,6 +17,7 @@ module.exports = {
   chatHistory,
   memory,
   rechart,
+  sqlAgent,
 
   // Plugin name aliases so they can be pulled by slug as well.
   [webScraping.name]: webScraping,
@@ -26,4 +28,5 @@ module.exports = {
   [chatHistory.name]: chatHistory,
   [memory.name]: memory,
   [rechart.name]: rechart,
+  [sqlAgent.name]: sqlAgent,
 };

server/utils/agents/aibitat/plugins/sql-agent/SQLConnectors/MSSQL.js

@@ -0,0 +1,89 @@
+const mssql = require("mssql");
+const UrlPattern = require("url-pattern");
+
+class MSSQLConnector {
+  #connected = false;
+  database_id = "";
+  connectionConfig = {
+    user: null,
+    password: null,
+    database: null,
+    server: null,
+    port: null,
+    pool: {
+      max: 10,
+      min: 0,
+      idleTimeoutMillis: 30000,
+    },
+    options: {
+      encrypt: false,
+      trustServerCertificate: true,
+    },
+  };
+
+  constructor(
+    config = {
+      // we will force into RFC-3986 from DB
+      // eg: mssql://user:password@server:port/database?{...opts}
+      connectionString: null, // we will force into RFC-3986
+    }
+  ) {
+    this.connectionString = config.connectionString;
+    this._client = null;
+    this.#parseDatabase();
+  }
+
+  #parseDatabase() {
+    const connectionPattern = new UrlPattern(
+      "mssql\\://:username\\::password@*\\::port/:database*"
+    );
+    const match = connectionPattern.match(this.connectionString);
+    this.database_id = match?.database;
+    this.connectionConfig = {
+      ...this.connectionConfig,
+      user: match?.username,
+      password: match?.password,
+      database: match?.database,
+      server: match?._[0],
+      port: match?.port ? Number(match.port) : null,
+    };
+  }
+
+  async connect() {
+    this._client = await mssql.connect(this.connectionConfig);
+    this.#connected = true;
+    return this._client;
+  }
+
+  /**
+   *
+   * @param {string} queryString the SQL query to be run
+   * @returns {import(".").QueryResult}
+   */
+  async runQuery(queryString = "") {
+    const result = { rows: [], count: 0, error: null };
+    try {
+      if (!this.#connected) await this.connect();
+
+      const query = await this._client.query(queryString);
+      result.rows = query.recordset;
+      result.count = query.rowsAffected.reduce((sum, a) => sum + a, 0);
+    } catch (err) {
+      console.log(this.constructor.name, err);
+      result.error = err.message;
+    } finally {
+      await this._client.close();
+      this.#connected = false;
+    }
+    return result;
+  }
+
+  getTablesSql() {
+    return `SELECT name FROM sysobjects WHERE xtype='U';`;
+  }
+  getTableSchemaSql(table_name) {
+    return `SELECT COLUMN_NAME,COLUMN_DEFAULT,IS_NULLABLE,DATA_TYPE FROM INFORMATION_SCHEMA.COLUMNS WHERE TABLE_NAME='${table_name}'`;
+  }
+}
+
+module.exports.MSSQLConnector = MSSQLConnector;

server/utils/agents/aibitat/plugins/sql-agent/SQLConnectors/MySQL.js

@@ -0,0 +1,59 @@
+const mysql = require("mysql2/promise");
+const UrlPattern = require("url-pattern");
+
+class MySQLConnector {
+  #connected = false;
+  database_id = "";
+  constructor(
+    config = {
+      connectionString: null,
+    }
+  ) {
+    this.connectionString = config.connectionString;
+    this._client = null;
+    this.database_id = this.#parseDatabase();
+  }
+
+  #parseDatabase() {
+    const connectionPattern = new UrlPattern("mysql\\://*@*/:database*");
+    const match = connectionPattern.match(this.connectionString);
+    return match?.database;
+  }
+
+  async connect() {
+    this._client = await mysql.createConnection({ uri: this.connectionString });
+    this.#connected = true;
+    return this._client;
+  }
+
+  /**
+   *
+   * @param {string} queryString the SQL query to be run
+   * @returns {import(".").QueryResult}
+   */
+  async runQuery(queryString = "") {
+    const result = { rows: [], count: 0, error: null };
+    try {
+      if (!this.#connected) await this.connect();
+      const [query] = await this._client.query(queryString);
+      result.rows = query;
+      result.count = query?.length;
+    } catch (err) {
+      console.log(this.constructor.name, err);
+      result.error = err.message;
+    } finally {
+      await this._client.end();
+      this.#connected = false;
+    }
+    return result;
+  }
+
+  getTablesSql() {
+    return `SELECT table_name FROM information_schema.tables WHERE table_schema = '${this.database_id}'`;
+  }
+  getTableSchemaSql(table_name) {
+    return `SHOW COLUMNS FROM ${this.database_id}.${table_name};`;
+  }
+}
+
+module.exports.MySQLConnector = MySQLConnector;

server/utils/agents/aibitat/plugins/sql-agent/SQLConnectors/Postgresql.js

@@ -0,0 +1,52 @@
+const pgSql = require("pg");
+
+class PostgresSQLConnector {
+  #connected = false;
+  constructor(
+    config = {
+      connectionString: null,
+    }
+  ) {
+    this.connectionString = config.connectionString;
+    this._client = new pgSql.Client({
+      connectionString: this.connectionString,
+    });
+  }
+
+  async connect() {
+    await this._client.connect();
+    this.#connected = true;
+    return this._client;
+  }
+
+  /**
+   *
+   * @param {string} queryString the SQL query to be run
+   * @returns {import(".").QueryResult}
+   */
+  async runQuery(queryString = "") {
+    const result = { rows: [], count: 0, error: null };
+    try {
+      if (!this.#connected) await this.connect();
+      const query = await this._client.query(queryString);
+      result.rows = query.rows;
+      result.count = query.rowCount;
+    } catch (err) {
+      console.log(this.constructor.name, err);
+      result.error = err.message;
+    } finally {
+      await this._client.end();
+      this.#connected = false;
+    }
+    return result;
+  }
+
+  getTablesSql() {
+    return `SELECT * FROM pg_catalog.pg_tables WHERE schemaname = 'public'`;
+  }
+  getTableSchemaSql(table_name) {
+    return ` select column_name, data_type, character_maximum_length, column_default, is_nullable from INFORMATION_SCHEMA.COLUMNS where table_name = '${table_name}'`;
+  }
+}
+
+module.exports.PostgresSQLConnector = PostgresSQLConnector;

server/utils/agents/aibitat/plugins/sql-agent/SQLConnectors/index.js

@@ -0,0 +1,60 @@
+const { SystemSettings } = require("../../../../../../models/systemSettings");
+const { safeJsonParse } = require("../../../../../http");
+
+/**
+ * @typedef {('postgresql'|'mysql'|'sql-server')} SQLEngine
+ */
+
+/**
+ * @typedef {Object} QueryResult
+ * @property {[number]} rows - The query result rows
+ * @property {number} count - Number of rows the query returned/changed
+ * @property {string|null} error - Error string if there was an issue
+ */
+
+/**
+ * A valid database SQL connection object
+ * @typedef {Object} SQLConnection
+ * @property {string} database_id - Unique identifier of the database connection
+ * @property {SQLEngine} engine - Engine used by connection
+ * @property {string} connectionString - RFC connection string for db
+ */
+
+/**
+ * @param {SQLEngine} identifier
+ * @param {object} connectionConfig
+ * @returns Database Connection Engine Class for SQLAgent or throws error
+ */
+function getDBClient(identifier = "", connectionConfig = {}) {
+  switch (identifier) {
+    case "mysql":
+      const { MySQLConnector } = require("./MySQL");
+      return new MySQLConnector(connectionConfig);
+    case "postgresql":
+      const { PostgresSQLConnector } = require("./Postgresql");
+      return new PostgresSQLConnector(connectionConfig);
+    case "sql-server":
+      const { MSSQLConnector } = require("./MSSQL");
+      return new MSSQLConnector(connectionConfig);
+    default:
+      throw new Error(
+        `There is no supported database connector for ${identifier}`
+      );
+  }
+}
+
+/**
+ * Lists all of the known database connection that can be used by the agent.
+ * @returns {Promise<[SQLConnection]>}
+ */
+async function listSQLConnections() {
+  return safeJsonParse(
+    (await SystemSettings.get({ label: "agent_sql_connections" }))?.value,
+    []
+  );
+}
+
+module.exports = {
+  getDBClient,
+  listSQLConnections,
+};

server/utils/agents/aibitat/plugins/sql-agent/get-table-schema.js

@@ -0,0 +1,98 @@
+module.exports.SqlAgentGetTableSchema = {
+  name: "sql-get-table-schema",
+  plugin: function () {
+    const {
+      listSQLConnections,
+      getDBClient,
+    } = require("./SQLConnectors/index.js");
+
+    return {
+      name: "sql-get-table-schema",
+      setup(aibitat) {
+        aibitat.function({
+          super: aibitat,
+          name: this.name,
+          description:
+            "Gets the table schema in SQL for a given `table` and `database_id`",
+          examples: [
+            {
+              prompt: "What does the customers table in access-logs look like?",
+              call: JSON.stringify({
+                database_id: "access-logs",
+                table_name: "customers",
+              }),
+            },
+            {
+              prompt:
+                "Get me the full name of a company in records-main, the table should be call comps",
+              call: JSON.stringify({
+                database_id: "records-main",
+                table_name: "comps",
+              }),
+            },
+          ],
+          parameters: {
+            $schema: "http://json-schema.org/draft-07/schema#",
+            type: "object",
+            properties: {
+              database_id: {
+                type: "string",
+                description:
+                  "The database identifier for which we will connect to to query the table schema. This is a required field.",
+              },
+              table_name: {
+                type: "string",
+                description:
+                  "The database identifier for the table name we want the schema for. This is a required field.",
+              },
+            },
+            additionalProperties: false,
+          },
+          required: ["database_id", "table_name"],
+          handler: async function ({ database_id = "", table_name = "" }) {
+            this.super.handlerProps.log(`Using the sql-get-table-schema tool.`);
+            try {
+              const databaseConfig = (await listSQLConnections()).find(
+                (db) => db.database_id === database_id
+              );
+              if (!databaseConfig) {
+                this.super.handlerProps.log(
+                  `sql-get-table-schema to find config!.`,
+                  database_id
+                );
+                return `No database connection for ${database_id} was found!`;
+              }
+
+              const db = getDBClient(databaseConfig.engine, databaseConfig);
+              this.super.introspect(
+                `${this.caller}: Querying the table schema for ${table_name} in the ${databaseConfig.database_id} database.`
+              );
+              this.super.introspect(
+                `Running SQL: ${db.getTableSchemaSql(table_name)}`
+              );
+              const result = await db.runQuery(
+                db.getTableSchemaSql(table_name)
+              );
+
+              if (result.error) {
+                this.super.handlerProps.log(
+                  `sql-get-table-schema tool reported error`,
+                  result.error
+                );
+                this.super.introspect(`Error: ${result.error}`);
+                return `There was an error running the query: ${result.error}`;
+              }
+
+              return JSON.stringify(result);
+            } catch (e) {
+              this.super.handlerProps.log(
+                `sql-get-table-schema raised an error. ${e.message}`
+              );
+              return e.message;
+            }
+          },
+        });
+      },
+    };
+  },
+};

server/utils/agents/aibitat/plugins/sql-agent/index.js

@@ -0,0 +1,21 @@
+const { SqlAgentGetTableSchema } = require("./get-table-schema");
+const { SqlAgentListDatabase } = require("./list-database");
+const { SqlAgentListTables } = require("./list-table");
+const { SqlAgentQuery } = require("./query");
+
+const sqlAgent = {
+  name: "sql-agent",
+  startupConfig: {
+    params: {},
+  },
+  plugin: [
+    SqlAgentListDatabase,
+    SqlAgentListTables,
+    SqlAgentGetTableSchema,
+    SqlAgentQuery,
+  ],
+};
+
+module.exports = {
+  sqlAgent,
+};

server/utils/agents/aibitat/plugins/sql-agent/list-database.js

@@ -0,0 +1,49 @@
+module.exports.SqlAgentListDatabase = {
+  name: "sql-list-databases",
+  plugin: function () {
+    const { listSQLConnections } = require("./SQLConnectors");
+    return {
+      name: "sql-list-databases",
+      setup(aibitat) {
+        aibitat.function({
+          super: aibitat,
+          name: this.name,
+          description:
+            "List all available databases via `list_databases` you currently have access to. Returns a unique string identifier `database_id` that can be used for future calls.",
+          examples: [
+            {
+              prompt: "What databases can you access?",
+              call: JSON.stringify({}),
+            },
+            {
+              prompt: "What databases can you tell me about?",
+              call: JSON.stringify({}),
+            },
+            {
+              prompt: "Is there a database named erp-logs you can access?",
+              call: JSON.stringify({}),
+            },
+          ],
+          parameters: {
+            $schema: "http://json-schema.org/draft-07/schema#",
+            type: "object",
+            properties: {},
+            additionalProperties: false,
+          },
+          handler: async function () {
+            this.super.handlerProps.log(`Using the sql-list-databases tool.`);
+            this.super.introspect(
+              `${this.caller}: Checking what are the available databases.`
+            );
+
+            const connections = (await listSQLConnections()).map((conn) => {
+              const { connectionString, ...rest } = conn;
+              return rest;
+            });
+            return JSON.stringify(connections);
+          },
+        });
+      },
+    };
+  },
+};

server/utils/agents/aibitat/plugins/sql-agent/list-table.js

@@ -0,0 +1,85 @@
+module.exports.SqlAgentListTables = {
+  name: "sql-list-tables",
+  plugin: function () {
+    const {
+      listSQLConnections,
+      getDBClient,
+    } = require("./SQLConnectors/index.js");
+
+    return {
+      name: "sql-list-tables",
+      setup(aibitat) {
+        aibitat.function({
+          super: aibitat,
+          name: this.name,
+          description:
+            "List all available tables in a database via its `database_id`.",
+          examples: [
+            {
+              prompt: "What tables are there in the `access-logs` database?",
+              call: JSON.stringify({ database_id: "access-logs" }),
+            },
+            {
+              prompt:
+                "What information can you access in the customer_accts postgres db?",
+              call: JSON.stringify({ database_id: "customer_accts" }),
+            },
+            {
+              prompt: "Can you tell me what is in the primary-logs db?",
+              call: JSON.stringify({ database_id: "primary-logs" }),
+            },
+          ],
+          parameters: {
+            $schema: "http://json-schema.org/draft-07/schema#",
+            type: "object",
+            properties: {
+              database_id: {
+                type: "string",
+                description:
+                  "The database identifier for which we will list all tables for. This is a required parameter",
+              },
+            },
+            additionalProperties: false,
+          },
+          required: ["database_id"],
+          handler: async function ({ database_id = "" }) {
+            try {
+              this.super.handlerProps.log(`Using the sql-list-tables tool.`);
+              const databaseConfig = (await listSQLConnections()).find(
+                (db) => db.database_id === database_id
+              );
+              if (!databaseConfig) {
+                this.super.handlerProps.log(
+                  `sql-list-tables failed to find config!.`,
+                  database_id
+                );
+                return `No database connection for ${database_id} was found!`;
+              }
+
+              const db = getDBClient(databaseConfig.engine, databaseConfig);
+              this.super.introspect(
+                `${this.caller}: Checking what are the available tables in the ${databaseConfig.database_id} database.`
+              );
+
+              this.super.introspect(`Running SQL: ${db.getTablesSql()}`);
+              const result = await db.runQuery(db.getTablesSql(database_id));
+              if (result.error) {
+                this.super.handlerProps.log(
+                  `sql-list-tables tool reported error`,
+                  result.error
+                );
+                this.super.introspect(`Error: ${result.error}`);
+                return `There was an error running the query: ${result.error}`;
+              }
+
+              return JSON.stringify(result);
+            } catch (e) {
+              console.error(e);
+              return e.message;
+            }
+          },
+        });
+      },
+    };
+  },
+};

server/utils/agents/aibitat/plugins/sql-agent/query.js

@@ -0,0 +1,101 @@
+module.exports.SqlAgentQuery = {
+  name: "sql-query",
+  plugin: function () {
+    const {
+      getDBClient,
+      listSQLConnections,
+    } = require("./SQLConnectors/index.js");
+
+    return {
+      name: "sql-query",
+      setup(aibitat) {
+        aibitat.function({
+          super: aibitat,
+          name: this.name,
+          description:
+            "Run a read-only SQL query on a `database_id` which will return up rows of data related to the query. The query must only be SELECT statements which do not modify the table data. There should be a reasonable LIMIT on the return quantity to prevent long-running or queries which crash the db.",
+          examples: [
+            {
+              prompt: "How many customers are in dvd-rentals?",
+              call: JSON.stringify({
+                database_id: "dvd-rentals",
+                sql_query: "SELECT * FROM customers",
+              }),
+            },
+            {
+              prompt: "Can you tell me the total volume of sales last month?",
+              call: JSON.stringify({
+                database_id: "sales-db",
+                sql_query:
+                  "SELECT SUM(sale_amount) AS total_sales FROM sales WHERE sale_date >= DATEADD(month, -1, DATEFROMPARTS(YEAR(GETDATE()), MONTH(GETDATE()), 1)) AND sale_date < DATEFROMPARTS(YEAR(GETDATE()), MONTH(GETDATE()), 1)",
+              }),
+            },
+            {
+              prompt:
+                "Do we have anyone in the staff table for our production db named 'sam'? ",
+              call: JSON.stringify({
+                database_id: "production",
+                sql_query:
+                  "SElECT * FROM staff WHERE first_name='sam%' OR last_name='sam%'",
+              }),
+            },
+          ],
+          parameters: {
+            $schema: "http://json-schema.org/draft-07/schema#",
+            type: "object",
+            properties: {
+              database_id: {
+                type: "string",
+                description:
+                  "The database identifier for which we will connect to to query the table schema. This is required to run the SQL query.",
+              },
+              sql_query: {
+                type: "string",
+                description:
+                  "The raw SQL query to run. Should be a query which does not modify the table and will return results.",
+              },
+            },
+            additionalProperties: false,
+          },
+          required: ["database_id", "table_name"],
+          handler: async function ({ database_id = "", sql_query = "" }) {
+            this.super.handlerProps.log(`Using the sql-query tool.`);
+            try {
+              const databaseConfig = (await listSQLConnections()).find(
+                (db) => db.database_id === database_id
+              );
+              if (!databaseConfig) {
+                this.super.handlerProps.log(
+                  `sql-query failed to find config!.`,
+                  database_id
+                );
+                return `No database connection for ${database_id} was found!`;
+              }
+
+              this.super.introspect(
+                `${this.caller}: Im going to run a query on the ${database_id} to get an answer.`
+              );
+              const db = getDBClient(databaseConfig.engine, databaseConfig);
+
+              this.super.introspect(`Running SQL: ${sql_query}`);
+              const result = await db.runQuery(sql_query);
+              if (result.error) {
+                this.super.handlerProps.log(
+                  `sql-query tool reported error`,
+                  result.error
+                );
+                this.super.introspect(`Error: ${result.error}`);
+                return `There was an error running the query: ${result.error}`;
+              }
+
+              return JSON.stringify(result);
+            } catch (e) {
+              console.error(e);
+              return e.message;
+            }
+          },
+        });
+      },
+    };
+  },
+};

server/utils/agents/aibitat/plugins/summarize.js

@@ -1,6 +1,5 @@
 const { Document } = require("../../../../models/documents");
 const { safeJsonParse } = require("../../../http");
-const { validate } = require("uuid");
 const { summarizeContent } = require("../utils/summarize");
 const Provider = require("../providers/ai-provider");
 

server/utils/agents/aibitat/plugins/websocket.js

@@ -49,7 +49,7 @@ const websocket = {
       setup(aibitat) {
         aibitat.onError(async (error) => {
           if (!!error?.message) {
-            console.error(chalk.red(`   error: ${error.message}`));
+            console.error(chalk.red(`   error: ${error.message}`), error);
             aibitat.introspect(
               `Error encountered while running: ${error.message}`
             );

server/utils/agents/aibitat/providers/anthropic.js

@@ -3,7 +3,7 @@ const { RetryError } = require("../error.js");
 const Provider = require("./ai-provider.js");
 
 /**
- * The provider for the Anthropic API.
+ * The agent provider for the Anthropic API.
  * By default, the model is set to 'claude-2'.
  */
 class AnthropicProvider extends Provider {

server/utils/agents/aibitat/providers/azure.js

@@ -4,7 +4,7 @@ const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
 
 /**
- * The provider for the Azure OpenAI API.
+ * The agent provider for the Azure OpenAI API.
  */
 class AzureOpenAiProvider extends InheritMultiple([Provider, UnTooled]) {
   model;
@@ -84,6 +84,11 @@ class AzureOpenAiProvider extends InheritMultiple([Provider, UnTooled]) {
         );
         completion = response.choices[0].message;
       }
+
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return { result: completion.content, cost: 0 };
     } catch (error) {
       throw error;

server/utils/agents/aibitat/providers/genericOpenAi.js

@@ -2,9 +2,10 @@ const OpenAI = require("openai");
 const Provider = require("./ai-provider.js");
 const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
+const { toValidNumber } = require("../../../http/index.js");
 
 /**
- * The provider for the Generic OpenAI provider.
+ * The agent provider for the Generic OpenAI provider.
  * Since we cannot promise the generic provider even supports tool calling
  * which is nearly 100% likely it does not, we can just wrap it in untooled
  * which often is far better anyway.
@@ -24,7 +25,9 @@ class GenericOpenAiProvider extends InheritMultiple([Provider, UnTooled]) {
     this._client = client;
     this.model = model;
     this.verbose = true;
-    this.maxTokens = process.env.GENERIC_OPEN_AI_MAX_TOKENS ?? 1024;
+    this.maxTokens = process.env.GENERIC_OPEN_AI_MAX_TOKENS
+      ? toValidNumber(process.env.GENERIC_OPEN_AI_MAX_TOKENS, 1024)
+      : 1024;
   }
 
   get client() {
@@ -94,6 +97,10 @@ class GenericOpenAiProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.choices[0].message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return {
         result: completion.content,
         cost: 0,

server/utils/agents/aibitat/providers/groq.js

@@ -3,7 +3,7 @@ const Provider = require("./ai-provider.js");
 const { RetryError } = require("../error.js");
 
 /**
- * The provider for the Groq provider.
+ * The agent provider for the Groq provider.
  * Using OpenAI tool calling with groq really sucks right now
  * its just fast and bad. We should probably migrate this to Untooled to improve
  * coherence.

server/utils/agents/aibitat/providers/koboldcpp.js

@@ -4,7 +4,7 @@ const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
 
 /**
- * The provider for the KoboldCPP provider.
+ * The agent provider for the KoboldCPP provider.
  */
 class KoboldCPPProvider extends InheritMultiple([Provider, UnTooled]) {
   model;
@@ -89,6 +89,10 @@ class KoboldCPPProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.choices[0].message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return {
         result: completion.content,
         cost: 0,

server/utils/agents/aibitat/providers/lmstudio.js

@@ -4,7 +4,7 @@ const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
 
 /**
- * The provider for the LMStudio provider.
+ * The agent provider for the LMStudio.
  */
 class LMStudioProvider extends InheritMultiple([Provider, UnTooled]) {
   model;
@@ -89,6 +89,10 @@ class LMStudioProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.choices[0].message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return {
         result: completion.content,
         cost: 0,

server/utils/agents/aibitat/providers/localai.js

@@ -4,7 +4,7 @@ const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
 
 /**
- * The provider for the LocalAI provider.
+ * The agent provider for the LocalAI provider.
  */
 class LocalAiProvider extends InheritMultiple([Provider, UnTooled]) {
   model;
@@ -93,6 +93,10 @@ class LocalAiProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.choices[0].message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return { result: completion.content, cost: 0 };
     } catch (error) {
       throw error;

server/utils/agents/aibitat/providers/mistral.js

@@ -4,7 +4,7 @@ const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
 
 /**
- * The provider for the Mistral provider.
+ * The agent provider for the Mistral provider.
  * Mistral limits what models can call tools and even when using those
  * the model names change and dont match docs. When you do have the right model
  * it still fails and is not truly OpenAI compatible so its easier to just wrap
@@ -93,6 +93,10 @@ class MistralProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.choices[0].message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return {
         result: completion.content,
         cost: 0,

server/utils/agents/aibitat/providers/ollama.js

@@ -4,7 +4,7 @@ const UnTooled = require("./helpers/untooled.js");
 const { Ollama } = require("ollama");
 
 /**
- * The provider for the Ollama provider.
+ * The agent provider for the Ollama provider.
  */
 class OllamaProvider extends InheritMultiple([Provider, UnTooled]) {
   model;
@@ -83,6 +83,10 @@ class OllamaProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return {
         result: completion.content,
         cost: 0,

server/utils/agents/aibitat/providers/openai.js

@@ -3,7 +3,7 @@ const Provider = require("./ai-provider.js");
 const { RetryError } = require("../error.js");
 
 /**
- * The provider for the OpenAI API.
+ * The agent provider for the OpenAI API.
  * By default, the model is set to 'gpt-3.5-turbo'.
  */
 class OpenAIProvider extends Provider {

server/utils/agents/aibitat/providers/openrouter.js

@@ -4,7 +4,7 @@ const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
 
 /**
- * The provider for the OpenRouter provider.
+ * The agent provider for the OpenRouter provider.
  */
 class OpenRouterProvider extends InheritMultiple([Provider, UnTooled]) {
   model;
@@ -93,6 +93,10 @@ class OpenRouterProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.choices[0].message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return {
         result: completion.content,
         cost: 0,

server/utils/agents/aibitat/providers/perplexity.js

@@ -4,7 +4,7 @@ const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
 
 /**
- * The provider for the Perplexity provider.
+ * The agent provider for the Perplexity provider.
  */
 class PerplexityProvider extends InheritMultiple([Provider, UnTooled]) {
   model;
@@ -89,6 +89,10 @@ class PerplexityProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.choices[0].message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return {
         result: completion.content,
         cost: 0,

server/utils/agents/aibitat/providers/textgenwebui.js

@@ -4,7 +4,7 @@ const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
 
 /**
- * The provider for the Oobabooga provider.
+ * The agent provider for the Oobabooga provider.
  */
 class TextWebGenUiProvider extends InheritMultiple([Provider, UnTooled]) {
   model;
@@ -88,6 +88,10 @@ class TextWebGenUiProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.choices[0].message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return {
         result: completion.content,
         cost: 0,

server/utils/agents/aibitat/providers/togetherai.js

@@ -4,7 +4,7 @@ const InheritMultiple = require("./helpers/classes.js");
 const UnTooled = require("./helpers/untooled.js");
 
 /**
- * The provider for the TogetherAI provider.
+ * The agent provider for the TogetherAI provider.
  */
 class TogetherAIProvider extends InheritMultiple([Provider, UnTooled]) {
   model;
@@ -89,6 +89,10 @@ class TogetherAIProvider extends InheritMultiple([Provider, UnTooled]) {
         completion = response.choices[0].message;
       }
 
+      // The UnTooled class inherited Deduplicator is mostly useful to prevent the agent
+      // from calling the exact same function over and over in a loop within a single chat exchange
+      // _but_ we should enable it to call previously used tools in a new chat interaction.
+      this.deduplicator.reset("runs");
       return {
         result: completion.content,
         cost: 0,

server/utils/agents/aibitat/utils/dedupe.js

@@ -38,6 +38,25 @@ class Deduplicator {
     return this.#hashes.hasOwnProperty(newSig);
   }
 
+  /**
+   * Resets the object property for this instance of the Deduplicator class
+   * @param {('runs'|'cooldowns'|'uniques')} type - The type of prop to reset
+   */
+  reset(type = "runs") {
+    switch (type) {
+      case "runs":
+        this.#hashes = {};
+        break;
+      case "cooldowns":
+        this.#cooldowns = {};
+        break;
+      case "uniques":
+        this.#uniques = {};
+        break;
+    }
+    return;
+  }
+
   startCooldown(
     key,
     parameters = {