merge with master

.vscode/settings.json

@@ -5,6 +5,7 @@
     "AIbitat",
     "allm",
     "anythingllm",
+    "Apipie",
     "Astra",
     "Chartable",
     "cleancss",
@@ -18,6 +19,7 @@
     "elevenlabs",
     "Embeddable",
     "epub",
+    "fireworksai",
     "GROQ",
     "hljs",
     "huggingface",
@@ -40,14 +42,13 @@
     "pagerender",
     "Qdrant",
     "royalblue",
-    "searxng",
     "SearchApi",
+    "searxng",
     "Serper",
     "Serply",
     "streamable",
     "textgenwebui",
     "togetherai",
-    "fireworksai",
     "Unembed",
     "vectordbs",
     "Weaviate",

docker/.env.example

@@ -105,6 +105,10 @@ GID='1000'
 # FIREWORKS_AI_LLM_API_KEY='my-fireworks-ai-key'
 # FIREWORKS_AI_LLM_MODEL_PREF='accounts/fireworks/models/llama-v3p1-8b-instruct'
 
+# LLM_PROVIDER='apipie'
+# APIPIE_LLM_API_KEY='sk-123abc'
+# APIPIE_LLM_MODEL_PREF='openrouter/llama-3.1-8b-instruct'
+
 ###########################################
 ######## Embedding API SElECTION ##########
 ###########################################

frontend/src/App.jsx

@@ -23,7 +23,6 @@ const WorkspaceChat = lazy(() => import("@/pages/WorkspaceChat"));
 const AdminUsers = lazy(() => import("@/pages/Admin/Users"));
 const AdminInvites = lazy(() => import("@/pages/Admin/Invitations"));
 const AdminWorkspaces = lazy(() => import("@/pages/Admin/Workspaces"));
-const AdminSystem = lazy(() => import("@/pages/Admin/System"));
 const AdminLogs = lazy(() => import("@/pages/Admin/Logging"));
 const AdminAgents = lazy(() => import("@/pages/Admin/Agents"));
 const GeneralChats = lazy(() => import("@/pages/GeneralSettings/Chats"));
@@ -172,10 +171,6 @@ export default function App() {
                     path="/settings/workspace-chats"
                     element={<ManagerRoute Component={GeneralChats} />}
                   />
-                  <Route
-                    path="/settings/system-preferences"
-                    element={<ManagerRoute Component={AdminSystem} />}
-                  />
                   <Route
                     path="/settings/invites"
                     element={<ManagerRoute Component={AdminInvites} />}

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

@@ -0,0 +1,101 @@
+import System from "@/models/system";
+import { useState, useEffect } from "react";
+
+export default function ApiPieLLMOptions({ settings }) {
+  return (
+    <div className="flex flex-col gap-y-4 mt-1.5">
+      <div className="flex gap-[36px]">
+        <div className="flex flex-col w-60">
+          <label className="text-white text-sm font-semibold block mb-3">
+            APIpie API Key
+          </label>
+          <input
+            type="password"
+            name="ApipieLLMApiKey"
+            className="bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:outline-primary-button active:outline-primary-button outline-none block w-full p-2.5"
+            placeholder="APIpie API Key"
+            defaultValue={settings?.ApipieLLMApiKey ? "*".repeat(20) : ""}
+            required={true}
+            autoComplete="off"
+            spellCheck={false}
+          />
+        </div>
+        {!settings?.credentialsOnly && (
+          <APIPieModelSelection settings={settings} />
+        )}
+      </div>
+    </div>
+  );
+}
+
+function APIPieModelSelection({ settings }) {
+  const [groupedModels, setGroupedModels] = useState({});
+  const [loading, setLoading] = useState(true);
+
+  useEffect(() => {
+    async function findCustomModels() {
+      setLoading(true);
+      const { models } = await System.customModels("apipie");
+      if (models?.length > 0) {
+        const modelsByOrganization = models.reduce((acc, model) => {
+          acc[model.organization] = acc[model.organization] || [];
+          acc[model.organization].push(model);
+          return acc;
+        }, {});
+
+        setGroupedModels(modelsByOrganization);
+      }
+
+      setLoading(false);
+    }
+    findCustomModels();
+  }, []);
+
+  if (loading || Object.keys(groupedModels).length === 0) {
+    return (
+      <div className="flex flex-col w-60">
+        <label className="text-white text-sm font-semibold block mb-3">
+          Chat Model Selection
+        </label>
+        <select
+          name="ApipieLLMModelPref"
+          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}>
+            -- loading available models --
+          </option>
+        </select>
+      </div>
+    );
+  }
+
+  return (
+    <div className="flex flex-col w-60">
+      <label className="text-white text-sm font-semibold block mb-3">
+        Chat Model Selection
+      </label>
+      <select
+        name="ApipieLLMModelPref"
+        required={true}
+        className="bg-zinc-900 border-gray-500 text-white text-sm rounded-lg block w-full p-2.5"
+      >
+        {Object.keys(groupedModels)
+          .sort()
+          .map((organization) => (
+            <optgroup key={organization} label={organization}>
+              {groupedModels[organization].map((model) => (
+                <option
+                  key={model.id}
+                  value={model.id}
+                  selected={settings?.ApipieLLMModelPref === model.id}
+                >
+                  {model.name}
+                </option>
+              ))}
+            </optgroup>
+          ))}
+      </select>
+    </div>
+  );
+}

frontend/src/components/Modals/ManageWorkspace/Documents/Directory/FileRow/index.jsx

@@ -31,7 +31,7 @@ export default function FileRow({ item, selected, toggleSelection }) {
           className="shrink-0 text-base font-bold w-4 h-4 mr-[3px]"
           weight="fill"
         />
-        <p className="whitespace-nowrap overflow-hidden text-ellipsis">
+        <p className="whitespace-nowrap overflow-hidden text-ellipsis max-w-[400px]">
           {middleTruncate(item.title, 55)}
         </p>
       </div>

frontend/src/components/Modals/ManageWorkspace/Documents/Directory/FolderRow/index.jsx

@@ -51,7 +51,7 @@ export default function FolderRow({
             className="shrink-0 text-base font-bold w-4 h-4 mr-[3px]"
             weight="fill"
           />
-          <p className="whitespace-nowrap overflow-show">
+          <p className="whitespace-nowrap overflow-show max-w-[400px]">
             {middleTruncate(item.name, 35)}
           </p>
         </div>

frontend/src/components/Modals/ManageWorkspace/Documents/WorkspaceDirectory/WorkspaceFileRow/index.jsx

@@ -83,7 +83,7 @@ export default function WorkspaceFileRow({
           className="shrink-0 text-base font-bold w-4 h-4 mr-[3px] ml-1"
           weight="fill"
         />
-        <p className="whitespace-nowrap overflow-hidden text-ellipsis">
+        <p className="whitespace-nowrap overflow-hidden text-ellipsis max-w-[400px]">
           {middleTruncate(item.title, 50)}
         </p>
       </div>

frontend/src/components/SettingsButton/index.jsx

@@ -29,10 +29,8 @@ export default function SettingsButton() {
   return (
     <ToolTipWrapper id="open-settings">
       <Link
-        to={
+        to={paths.settings.appearance()}
-          !!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"
-        }
-        className="transition-all duration-300 p-2 rounded-full text-white bg-theme-sidebar-footer-icon hover:bg-theme-sidebar-footer-icon-hover"
         aria-label="Settings"
         data-tooltip-id="open-settings"
         data-tooltip-content="Open settings"

frontend/src/components/SettingsSidebar/index.jsx

@@ -277,11 +277,6 @@ const SidebarOptions = ({ user = null, t }) => (
           href: paths.settings.invites(),
           roles: ["admin", "manager"],
         },
-        {
-          btnText: t("settings.system"),
-          href: paths.settings.system(),
-          roles: ["admin", "manager"],
-        },
       ]}
     />
     <Option

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

@@ -122,9 +122,22 @@ export default function PromptInput({
 
     const pasteText = e.clipboardData.getData("text/plain");
     if (pasteText) {
-      const newPromptInput = promptInput + pasteText.trim();
+      const textarea = textareaRef.current;
+      const start = textarea.selectionStart;
+      const end = textarea.selectionEnd;
+      const newPromptInput =
+        promptInput.substring(0, start) +
+        pasteText +
+        promptInput.substring(end);
       setPromptInput(newPromptInput);
       onChange({ target: { value: newPromptInput } });
+
+      // Set the cursor position after the pasted text
+      // we need to use setTimeout to prevent the cursor from being set to the end of the text
+      setTimeout(() => {
+        textarea.selectionStart = textarea.selectionEnd =
+          start + pasteText.length;
+      }, 0);
     }
     return;
   };

frontend/src/pages/Admin/System/index.jsx

@@ -1,128 +0,0 @@
-import { useEffect, useState } from "react";
-import Sidebar from "@/components/SettingsSidebar";
-import { isMobile } from "react-device-detect";
-import Admin from "@/models/admin";
-import showToast from "@/utils/toast";
-import CTAButton from "@/components/lib/CTAButton";
-
-export default function AdminSystem() {
-  const [saving, setSaving] = useState(false);
-  const [hasChanges, setHasChanges] = useState(false);
-  const [messageLimit, setMessageLimit] = useState({
-    enabled: false,
-    limit: 10,
-  });
-
-  const handleSubmit = async (e) => {
-    e.preventDefault();
-    setSaving(true);
-    await Admin.updateSystemPreferences({
-      limit_user_messages: messageLimit.enabled,
-      message_limit: messageLimit.limit,
-    });
-    setSaving(false);
-    setHasChanges(false);
-    showToast("System preferences updated successfully.", "success");
-  };
-
-  useEffect(() => {
-    async function fetchSettings() {
-      const settings = (await Admin.systemPreferences())?.settings;
-      if (!settings) return;
-      setMessageLimit({
-        enabled: settings.limit_user_messages,
-        limit: settings.message_limit,
-      });
-    }
-    fetchSettings();
-  }, []);
-
-  return (
-    <div className="w-screen h-screen overflow-hidden bg-theme-bg-container flex">
-      <Sidebar />
-      <div
-        style={{ height: isMobile ? "100%" : "calc(100% - 32px)" }}
-        className="relative md:ml-[2px] md:mr-[16px] md:my-[16px] md:rounded-[16px] bg-theme-bg-secondary w-full h-full overflow-y-scroll p-4 md:p-0"
-      >
-        <form
-          onSubmit={handleSubmit}
-          onChange={() => setHasChanges(true)}
-          className="flex flex-col w-full px-1 md:pl-6 md:pr-[50px] md:py-6 py-16"
-        >
-          <div className="w-full flex flex-col gap-y-1 pb-6 border-white/10 border-b-2">
-            <div className="items-center">
-              <p className="text-lg leading-6 font-bold text-white">
-                System Preferences
-              </p>
-            </div>
-            <p className="text-xs leading-[18px] font-base text-white text-opacity-60">
-              These are the overall settings and configurations of your
-              instance.
-            </p>
-          </div>
-          {hasChanges && (
-            <div className="flex justify-end">
-              <CTAButton onClick={handleSubmit} className="mt-3 mr-0">
-                {saving ? "Saving..." : "Save changes"}
-              </CTAButton>
-            </div>
-          )}
-          <div className="mt-4 mb-8">
-            <div className="flex flex-col gap-y-1">
-              <h2 className="text-base leading-6 font-bold text-white">
-                Limit messages per user per day
-              </h2>
-              <p className="text-xs leading-[18px] font-base text-white/60">
-                Restrict non-admin users to a number of successful queries or
-                chats within a 24 hour window. Enable this to prevent users from
-                running up OpenAI costs.
-              </p>
-              <div className="mt-2">
-                <label className="relative inline-flex cursor-pointer items-center">
-                  <input
-                    type="checkbox"
-                    name="limit_user_messages"
-                    value="yes"
-                    checked={messageLimit.enabled}
-                    onChange={(e) => {
-                      setMessageLimit({
-                        ...messageLimit,
-                        enabled: e.target.checked,
-                      });
-                    }}
-                    className="peer sr-only"
-                  />
-                  <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>
-            </div>
-            {messageLimit.enabled && (
-              <div className="mt-4">
-                <label className="text-white text-sm font-semibold block mb-4">
-                  Message limit per day
-                </label>
-                <div className="relative mt-2">
-                  <input
-                    type="number"
-                    name="message_limit"
-                    onScroll={(e) => e.target.blur()}
-                    onChange={(e) => {
-                      setMessageLimit({
-                        enabled: true,
-                        limit: Number(e?.target?.value || 0),
-                      });
-                    }}
-                    value={messageLimit.limit}
-                    min={1}
-                    className="bg-theme-settings-input-bg text-white placeholder:text-theme-settings-input-placeholder text-sm rounded-lg focus:outline-primary-button active:outline-primary-button outline-none block w-60 p-2.5"
-                  />
-                </div>
-              </div>
-            )}
-          </div>
-        </form>
-      </div>
-    </div>
-  );
-}

frontend/src/pages/Admin/Users/NewUserModal/index.jsx

@@ -2,11 +2,15 @@ import React, { useState } from "react";
 import { X } from "@phosphor-icons/react";
 import Admin from "@/models/admin";
 import { userFromStorage } from "@/utils/request";
-import { RoleHintDisplay } from "..";
+import { MessageLimitInput, RoleHintDisplay } from "..";
 
 export default function NewUserModal({ closeModal }) {
   const [error, setError] = useState(null);
   const [role, setRole] = useState("default");
+  const [messageLimit, setMessageLimit] = useState({
+    enabled: false,
+    limit: 10,
+  });
 
   const handleCreate = async (e) => {
     setError(null);
@@ -14,6 +18,8 @@ export default function NewUserModal({ closeModal }) {
     const data = {};
     const form = new FormData(e.target);
     for (var [key, value] of form.entries()) data[key] = value;
+    data.dailyMessageLimit = messageLimit.enabled ? messageLimit.limit : null;
+
     const { user, error } = await Admin.newUser(data);
     if (!!user) window.location.reload();
     setError(error);
@@ -59,13 +65,13 @@ export default function NewUserModal({ closeModal }) {
                   pattern="^[a-z0-9_-]+$"
                   onInvalid={(e) =>
                     e.target.setCustomValidity(
-                      "Username must be only contain lowercase letters, numbers, underscores, and hyphens with no spaces"
+                      "Username must only contain lowercase letters, numbers, underscores, and hyphens with no spaces"
                     )
                   }
                   onChange={(e) => e.target.setCustomValidity("")}
                 />
                 <p className="mt-2 text-xs text-white/60">
-                  Username must be only contain lowercase letters, numbers,
+                  Username must only contain lowercase letters, numbers,
                   underscores, and hyphens with no spaces
                 </p>
               </div>
@@ -111,6 +117,12 @@ export default function NewUserModal({ closeModal }) {
                 </select>
                 <RoleHintDisplay role={role} />
               </div>
+              <MessageLimitInput
+                role={role}
+                enabled={messageLimit.enabled}
+                limit={messageLimit.limit}
+                updateState={setMessageLimit}
+              />
               {error && <p className="text-red-400 text-sm">Error: {error}</p>}
               <p className="text-white text-xs md:text-sm">
                 After creating a user they will need to login with their initial

frontend/src/pages/Admin/Users/UserRow/EditUserModal/index.jsx

@@ -1,11 +1,15 @@
 import React, { useState } from "react";
 import { X } from "@phosphor-icons/react";
 import Admin from "@/models/admin";
-import { RoleHintDisplay } from "../..";
+import { MessageLimitInput, RoleHintDisplay } from "../..";
 
 export default function EditUserModal({ currentUser, user, closeModal }) {
   const [role, setRole] = useState(user.role);
   const [error, setError] = useState(null);
+  const [messageLimit, setMessageLimit] = useState({
+    enabled: user.dailyMessageLimit !== null,
+    limit: user.dailyMessageLimit || 10,
+  });
 
   const handleUpdate = async (e) => {
     setError(null);
@@ -16,6 +20,12 @@ export default function EditUserModal({ currentUser, user, closeModal }) {
       if (!value || value === null) continue;
       data[key] = value;
     }
+    if (messageLimit.enabled) {
+      data.dailyMessageLimit = messageLimit.limit;
+    } else {
+      data.dailyMessageLimit = null;
+    }
+
     const { success, error } = await Admin.updateUser(user.id, data);
     if (success) window.location.reload();
     setError(error);
@@ -59,7 +69,7 @@ export default function EditUserModal({ currentUser, user, closeModal }) {
                   autoComplete="off"
                 />
                 <p className="mt-2 text-xs text-white/60">
-                  Username must be only contain lowercase letters, numbers,
+                  Username must only contain lowercase letters, numbers,
                   underscores, and hyphens with no spaces
                 </p>
               </div>
@@ -104,6 +114,12 @@ export default function EditUserModal({ currentUser, user, closeModal }) {
                 </select>
                 <RoleHintDisplay role={role} />
               </div>
+              <MessageLimitInput
+                role={role}
+                enabled={messageLimit.enabled}
+                limit={messageLimit.limit}
+                updateState={setMessageLimit}
+              />
               {error && <p className="text-red-400 text-sm">Error: {error}</p>}
             </div>
             <div className="flex justify-between items-center mt-6 pt-6 border-t border-theme-modal-border">

frontend/src/pages/Admin/Users/index.jsx

@@ -142,3 +142,58 @@ export function RoleHintDisplay({ role }) {
     </div>
   );
 }
+
+export function MessageLimitInput({ enabled, limit, updateState, role }) {
+  if (role === "admin") return null;
+  return (
+    <div className="mt-4 mb-8">
+      <div className="flex flex-col gap-y-1">
+        <div className="flex items-center gap-x-2">
+          <h2 className="text-base leading-6 font-bold text-white">
+            Limit messages per day
+          </h2>
+          <label className="relative inline-flex cursor-pointer items-center">
+            <input
+              type="checkbox"
+              checked={enabled}
+              onChange={(e) => {
+                updateState((prev) => ({
+                  ...prev,
+                  enabled: e.target.checked,
+                }));
+              }}
+              className="peer sr-only"
+            />
+            <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>
+          </label>
+        </div>
+        <p className="text-xs leading-[18px] font-base text-white/60">
+          Restrict this user to a number of successful queries or chats within a
+          24 hour window.
+        </p>
+      </div>
+      {enabled && (
+        <div className="mt-4">
+          <label className="text-white text-sm font-semibold block mb-4">
+            Message limit per day
+          </label>
+          <div className="relative mt-2">
+            <input
+              type="number"
+              onScroll={(e) => e.target.blur()}
+              onChange={(e) => {
+                updateState({
+                  enabled: true,
+                  limit: Number(e?.target?.value || 0),
+                });
+              }}
+              value={limit}
+              min={1}
+              className="bg-zinc-900 text-white placeholder:text-white/20 text-sm rounded-lg focus:border-white block w-60 p-2.5"
+            />
+          </div>
+        </div>
+      )}
+    </div>
+  );
+}

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

@@ -26,6 +26,7 @@ import CohereLogo from "@/media/llmprovider/cohere.png";
 import LiteLLMLogo from "@/media/llmprovider/litellm.png";
 import AWSBedrockLogo from "@/media/llmprovider/bedrock.png";
 import DeepSeekLogo from "@/media/llmprovider/deepseek.png";
+import APIPieLogo from "@/media/llmprovider/apipie.png";
 
 import PreLoader from "@/components/Preloader";
 import OpenAiOptions from "@/components/LLMSelection/OpenAiOptions";
@@ -50,6 +51,7 @@ import TextGenWebUIOptions from "@/components/LLMSelection/TextGenWebUIOptions";
 import LiteLLMOptions from "@/components/LLMSelection/LiteLLMOptions";
 import AWSBedrockLLMOptions from "@/components/LLMSelection/AwsBedrockLLMOptions";
 import DeepSeekOptions from "@/components/LLMSelection/DeepSeekOptions";
+import ApiPieLLMOptions from "@/components/LLMSelection/ApiPieOptions";
 
 import LLMItem from "@/components/LLMSelection/LLMItem";
 import { CaretUpDown, MagnifyingGlass, X } from "@phosphor-icons/react";
@@ -221,6 +223,27 @@ export const AVAILABLE_LLM_PROVIDERS = [
     description: "Run DeepSeek's powerful LLMs.",
     requiredConfig: ["DeepSeekApiKey"],
   },
+  {
+    name: "AWS Bedrock",
+    value: "bedrock",
+    logo: AWSBedrockLogo,
+    options: (settings) => <AWSBedrockLLMOptions settings={settings} />,
+    description: "Run powerful foundation models privately with AWS Bedrock.",
+    requiredConfig: [
+      "AwsBedrockLLMAccessKeyId",
+      "AwsBedrockLLMAccessKey",
+      "AwsBedrockLLMRegion",
+      "AwsBedrockLLMModel",
+    ],
+  },
+  {
+    name: "APIpie",
+    value: "apipie",
+    logo: APIPieLogo,
+    options: (settings) => <ApiPieLLMOptions settings={settings} />,
+    description: "A unified API of AI services from leading providers",
+    requiredConfig: ["ApipieLLMApiKey", "ApipieLLMModelPref"],
+  },
   {
     name: "Generic OpenAI",
     value: "generic-openai",
@@ -235,19 +258,6 @@ export const AVAILABLE_LLM_PROVIDERS = [
       "GenericOpenAiKey",
     ],
   },
-  {
-    name: "AWS Bedrock",
-    value: "bedrock",
-    logo: AWSBedrockLogo,
-    options: (settings) => <AWSBedrockLLMOptions settings={settings} />,
-    description: "Run powerful foundation models privately with AWS Bedrock.",
-    requiredConfig: [
-      "AwsBedrockLLMAccessKeyId",
-      "AwsBedrockLLMAccessKey",
-      "AwsBedrockLLMRegion",
-      "AwsBedrockLLMModel",
-    ],
-  },
   {
     name: "Native",
     value: "native",

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

@@ -21,6 +21,7 @@ import TextGenWebUILogo from "@/media/llmprovider/text-generation-webui.png";
 import LiteLLMLogo from "@/media/llmprovider/litellm.png";
 import AWSBedrockLogo from "@/media/llmprovider/bedrock.png";
 import DeepSeekLogo from "@/media/llmprovider/deepseek.png";
+import APIPieLogo from "@/media/llmprovider/apipie.png";
 
 import CohereLogo from "@/media/llmprovider/cohere.png";
 import ZillizLogo from "@/media/vectordbs/zilliz.png";
@@ -202,6 +203,13 @@ export const LLM_SELECTION_PRIVACY = {
     description: ["Your model and chat contents are visible to DeepSeek"],
     logo: DeepSeekLogo,
   },
+  apipie: {
+    name: "APIpie.AI",
+    description: [
+      "Your model and chat contents are visible to APIpie in accordance with their terms of service.",
+    ],
+    logo: APIPieLogo,
+  },
 };
 
 export const VECTOR_DB_PRIVACY = {

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

@@ -21,6 +21,7 @@ import TextGenWebUILogo from "@/media/llmprovider/text-generation-webui.png";
 import LiteLLMLogo from "@/media/llmprovider/litellm.png";
 import AWSBedrockLogo from "@/media/llmprovider/bedrock.png";
 import DeepSeekLogo from "@/media/llmprovider/deepseek.png";
+import APIPieLogo from "@/media/llmprovider/apipie.png";
 
 import CohereLogo from "@/media/llmprovider/cohere.png";
 import OpenAiOptions from "@/components/LLMSelection/OpenAiOptions";
@@ -45,6 +46,7 @@ import TextGenWebUIOptions from "@/components/LLMSelection/TextGenWebUIOptions";
 import LiteLLMOptions from "@/components/LLMSelection/LiteLLMOptions";
 import AWSBedrockLLMOptions from "@/components/LLMSelection/AwsBedrockLLMOptions";
 import DeepSeekOptions from "@/components/LLMSelection/DeepSeekOptions";
+import ApiPieLLMOptions from "@/components/LLMSelection/ApiPieOptions";
 
 import LLMItem from "@/components/LLMSelection/LLMItem";
 import System from "@/models/system";
@@ -195,6 +197,13 @@ const LLMS = [
     options: (settings) => <DeepSeekOptions settings={settings} />,
     description: "Run DeepSeek's powerful LLMs.",
   },
+  {
+    name: "APIpie",
+    value: "apipie",
+    logo: APIPieLogo,
+    options: (settings) => <ApiPieLLMOptions settings={settings} />,
+    description: "A unified API of AI services from leading providers",
+  },
   {
     name: "Generic OpenAI",
     value: "generic-openai",

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

@@ -24,6 +24,8 @@ const ENABLED_PROVIDERS = [
   "bedrock",
   "fireworksai",
   "deepseek",
+  "litellm",
+  "apipie",
   // TODO: More agent support.
   // "cohere",         // Has tool calling and will need to build explicit support
   // "huggingface"     // Can be done but already has issues with no-chat templated. Needs to be tested.

frontend/src/utils/paths.js

@@ -80,9 +80,6 @@ export default {
     return `/fine-tuning`;
   },
   settings: {
-    system: () => {
-      return `/settings/system-preferences`;
-    },
     users: () => {
       return `/settings/users`;
     },

server/.env.example

@@ -95,6 +95,10 @@ SIG_SALT='salt' # Please generate random string at least 32 chars long.
 # COHERE_API_KEY=
 # COHERE_MODEL_PREF='command-r'
 
+# LLM_PROVIDER='apipie'
+# APIPIE_LLM_API_KEY='sk-123abc'
+# APIPIE_LLM_MODEL_PREF='openrouter/llama-3.1-8b-instruct'
+
 ###########################################
 ######## Embedding API SElECTION ##########
 ###########################################

server/endpoints/admin.js

@@ -347,14 +347,6 @@ function adminEndpoints(app) {
             : await SystemSettings.get({ label });
 
           switch (label) {
-            case "limit_user_messages":
-              requestedSettings[label] = setting?.value === "true";
-              break;
-            case "message_limit":
-              requestedSettings[label] = setting?.value
-                ? Number(setting.value)
-                : 10;
-              break;
             case "footer_data":
               requestedSettings[label] = setting?.value ?? JSON.stringify([]);
               break;
@@ -422,13 +414,6 @@ function adminEndpoints(app) {
       try {
         const embedder = getEmbeddingEngineSelection();
         const settings = {
-          limit_user_messages:
-            (await SystemSettings.get({ label: "limit_user_messages" }))
-              ?.value === "true",
-          message_limit:
-            Number(
-              (await SystemSettings.get({ label: "message_limit" }))?.value
-            ) || 10,
           footer_data:
             (await SystemSettings.get({ label: "footer_data" }))?.value ||
             JSON.stringify([]),

server/endpoints/api/admin/index.js

@@ -595,56 +595,6 @@ function apiAdminEndpoints(app) {
     }
   );
 
-  app.get("/v1/admin/preferences", [validApiKey], async (request, response) => {
-    /*
-    #swagger.tags = ['Admin']
-    #swagger.description = 'Show all multi-user preferences for instance. Methods are disabled until multi user mode is enabled via the UI.'
-    #swagger.responses[200] = {
-      content: {
-        "application/json": {
-          schema: {
-            type: 'object',
-            example: {
-              settings: {
-                limit_user_messages: false,
-                message_limit: 10,
-              }
-            }
-          }
-        }
-      }
-    }
-    #swagger.responses[403] = {
-      schema: {
-        "$ref": "#/definitions/InvalidAPIKey"
-      }
-    }
-     #swagger.responses[401] = {
-      description: "Instance is not in Multi-User mode. Method denied",
-    }
-    */
-    try {
-      if (!multiUserMode(response)) {
-        response.sendStatus(401).end();
-        return;
-      }
-
-      const settings = {
-        limit_user_messages:
-          (await SystemSettings.get({ label: "limit_user_messages" }))
-            ?.value === "true",
-        message_limit:
-          Number(
-            (await SystemSettings.get({ label: "message_limit" }))?.value
-          ) || 10,
-      };
-      response.status(200).json({ settings });
-    } catch (e) {
-      console.error(e);
-      response.sendStatus(500).end();
-    }
-  });
-
   app.post(
     "/v1/admin/preferences",
     [validApiKey],
@@ -658,8 +608,7 @@ function apiAdminEndpoints(app) {
       content: {
         "application/json": {
           example: {
-            limit_user_messages: true,
+            support_email: "support@example.com",
-            message_limit: 5,
           }
         }
       }

server/endpoints/chat.js

@@ -1,8 +1,6 @@
 const { v4: uuidv4 } = require("uuid");
 const { reqBody, userFromSession, multiUserMode } = require("../utils/http");
 const { validatedRequest } = require("../utils/middleware/validatedRequest");
-const { WorkspaceChats } = require("../models/workspaceChats");
-const { SystemSettings } = require("../models/systemSettings");
 const { Telemetry } = require("../models/telemetry");
 const { streamChatWithWorkspace } = require("../utils/chats/stream");
 const {
@@ -16,6 +14,7 @@ const {
 } = require("../utils/middleware/validWorkspace");
 const { writeResponseChunk } = require("../utils/helpers/chat/responses");
 const { WorkspaceThread } = require("../models/workspaceThread");
+const { User } = require("../models/user");
 const truncate = require("truncate");
 
 function chatEndpoints(app) {
@@ -48,40 +47,17 @@ function chatEndpoints(app) {
         response.setHeader("Connection", "keep-alive");
         response.flushHeaders();
 
-        if (multiUserMode(response) && user.role !== ROLES.admin) {
+        if (multiUserMode(response) && !(await User.canSendChat(user))) {
-          const limitMessagesSetting = await SystemSettings.get({
-            label: "limit_user_messages",
-          });
-          const limitMessages = limitMessagesSetting?.value === "true";
-
-          if (limitMessages) {
-            const messageLimitSetting = await SystemSettings.get({
-              label: "message_limit",
-            });
-            const systemLimit = Number(messageLimitSetting?.value);
-
-            if (!!systemLimit) {
-              const currentChatCount = await WorkspaceChats.count({
-                user_id: user.id,
-                createdAt: {
-                  gte: new Date(new Date() - 24 * 60 * 60 * 1000),
-                },
-              });
-
-              if (currentChatCount >= systemLimit) {
           writeResponseChunk(response, {
             id: uuidv4(),
             type: "abort",
             textResponse: null,
             sources: [],
             close: true,
-                  error: `You have met your maximum 24 hour chat quota of ${systemLimit} chats set by the instance administrators. Try again later.`,
+            error: `You have met your maximum 24 hour chat quota of ${user.dailyMessageLimit} chats. Try again later.`,
           });
           return;
         }
-            }
-          }
-        }
 
         await streamChatWithWorkspace(
           response,
@@ -157,42 +133,17 @@ function chatEndpoints(app) {
         response.setHeader("Connection", "keep-alive");
         response.flushHeaders();
 
-        if (multiUserMode(response) && user.role !== ROLES.admin) {
+        if (multiUserMode(response) && !(await User.canSendChat(user))) {
-          const limitMessagesSetting = await SystemSettings.get({
-            label: "limit_user_messages",
-          });
-          const limitMessages = limitMessagesSetting?.value === "true";
-
-          if (limitMessages) {
-            const messageLimitSetting = await SystemSettings.get({
-              label: "message_limit",
-            });
-            const systemLimit = Number(messageLimitSetting?.value);
-
-            if (!!systemLimit) {
-              // Chat qty includes all threads because any user can freely
-              // create threads and would bypass this rule.
-              const currentChatCount = await WorkspaceChats.count({
-                user_id: user.id,
-                createdAt: {
-                  gte: new Date(new Date() - 24 * 60 * 60 * 1000),
-                },
-              });
-
-              if (currentChatCount >= systemLimit) {
           writeResponseChunk(response, {
             id: uuidv4(),
             type: "abort",
             textResponse: null,
             sources: [],
             close: true,
-                  error: `You have met your maximum 24 hour chat quota of ${systemLimit} chats set by the instance administrators. Try again later.`,
+            error: `You have met your maximum 24 hour chat quota of ${user.dailyMessageLimit} chats. Try again later.`,
           });
           return;
         }
-            }
-          }
-        }
 
         await streamChatWithWorkspace(
           response,

server/endpoints/embed/index.js

@@ -56,6 +56,7 @@ function embeddedEndpoints(app) {
         writeResponseChunk(response, {
           id: uuidv4(),
           type: "abort",
+          sources: [],
           textResponse: null,
           close: true,
           error: e.message,
@@ -72,11 +73,15 @@ function embeddedEndpoints(app) {
       try {
         const { sessionId } = request.params;
         const embed = response.locals.embedConfig;
+        const history = await EmbedChats.forEmbedByUser(
+          embed.id,
+          sessionId,
+          null,
+          null,
+          true
+        );
 
-        const history = await EmbedChats.forEmbedByUser(embed.id, sessionId);
+        response.status(200).json({ history: convertToChatHistory(history) });
-        response.status(200).json({
-          history: convertToChatHistory(history),
-        });
       } catch (e) {
         console.error(e.message, e);
         response.sendStatus(500).end();

server/endpoints/system.js

@@ -490,8 +490,6 @@ function systemEndpoints(app) {
 
         await SystemSettings._updateSettings({
           multi_user_mode: true,
-          limit_user_messages: false,
-          message_limit: 25,
         });
         await BrowserExtensionApiKey.migrateApiKeysToMultiUser(user.id);
 

server/models/embedChats.js

@@ -1,5 +1,17 @@
+const { safeJsonParse } = require("../utils/http");
 const prisma = require("../utils/prisma");
 
+/**
+ * @typedef {Object} EmbedChat
+ * @property {number} id
+ * @property {number} embed_id
+ * @property {string} prompt
+ * @property {string} response
+ * @property {string} connection_information
+ * @property {string} session_id
+ * @property {boolean} include
+ */
+
 const EmbedChats = {
   new: async function ({
     embedId,
@@ -25,11 +37,36 @@ const EmbedChats = {
     }
   },
 
+  /**
+   * Loops through each chat and filters out the sources from the response object.
+   * We do this when returning /history of an embed to the frontend to prevent inadvertent leaking
+   * of private sources the user may not have intended to share with users.
+   * @param {EmbedChat[]} chats
+   * @returns {EmbedChat[]} Returns a new array of chats with the sources filtered out of responses
+   */
+  filterSources: function (chats) {
+    return chats.map((chat) => {
+      const { response, ...rest } = chat;
+      const { sources, ...responseRest } = safeJsonParse(response);
+      return { ...rest, response: JSON.stringify(responseRest) };
+    });
+  },
+
+  /**
+   * Fetches chats for a given embed and session id.
+   * @param {number} embedId the id of the embed to fetch chats for
+   * @param {string} sessionId the id of the session to fetch chats for
+   * @param {number|null} limit the maximum number of chats to fetch
+   * @param {string|null} orderBy the order to fetch chats in
+   * @param {boolean} filterSources whether to filter out the sources from the response (default: false)
+   * @returns {Promise<EmbedChat[]>} Returns an array of chats for the given embed and session
+   */
   forEmbedByUser: async function (
     embedId = null,
     sessionId = null,
     limit = null,
-    orderBy = null
+    orderBy = null,
+    filterSources = false
   ) {
     if (!embedId || !sessionId) return [];
 
@@ -43,7 +80,7 @@ const EmbedChats = {
         ...(limit !== null ? { take: limit } : {}),
         ...(orderBy !== null ? { orderBy } : { orderBy: { id: "asc" } }),
       });
-      return chats;
+      return filterSources ? this.filterSources(chats) : chats;
     } catch (error) {
       console.error(error.message);
       return [];

server/models/systemSettings.js

@@ -16,8 +16,6 @@ function isNullOrNaN(value) {
 const SystemSettings = {
   protectedFields: ["multi_user_mode"],
   publicFields: [
-    "limit_user_messages",
-    "message_limit",
     "footer_data",
     "support_email",
     "text_splitter_chunk_size",
@@ -33,8 +31,6 @@ const SystemSettings = {
     "meta_page_favicon",
   ],
   supportedFields: [
-    "limit_user_messages",
-    "message_limit",
     "logo_filename",
     "telemetry_id",
     "footer_data",
@@ -512,6 +508,10 @@ const SystemSettings = {
       // DeepSeek API Keys
       DeepSeekApiKey: !!process.env.DEEPSEEK_API_KEY,
       DeepSeekModelPref: process.env.DEEPSEEK_MODEL_PREF,
+
+      // APIPie LLM API Keys
+      ApipieLLMApiKey: !!process.env.APIPIE_LLM_API_KEY,
+      ApipieLLMModelPref: process.env.APIPIE_LLM_MODEL_PREF,
     };
   },
 

server/models/user.js

@@ -1,6 +1,17 @@
 const prisma = require("../utils/prisma");
 const { EventLogs } = require("./eventLogs");
 
+/**
+ * @typedef {Object} User
+ * @property {number} id
+ * @property {string} username
+ * @property {string} password
+ * @property {string} pfpFilename
+ * @property {string} role
+ * @property {boolean} suspended
+ * @property {number|null} dailyMessageLimit
+ */
+
 const User = {
   usernameRegex: new RegExp(/^[a-z0-9_-]+$/),
   writable: [
@@ -10,6 +21,7 @@ const User = {
     "pfpFilename",
     "role",
     "suspended",
+    "dailyMessageLimit",
   ],
   validations: {
     username: (newValue = "") => {
@@ -32,12 +44,24 @@ const User = {
       }
       return String(role);
     },
+    dailyMessageLimit: (dailyMessageLimit = null) => {
+      if (dailyMessageLimit === null) return null;
+      const limit = Number(dailyMessageLimit);
+      if (isNaN(limit) || limit < 1) {
+        throw new Error(
+          "Daily message limit must be null or a number greater than or equal to 1"
+        );
+      }
+      return limit;
+    },
   },
   // validations for the above writable fields.
   castColumnValue: function (key, value) {
     switch (key) {
       case "suspended":
         return Number(Boolean(value));
+      case "dailyMessageLimit":
+        return value === null ? null : Number(value);
       default:
         return String(value);
     }
@@ -48,7 +72,12 @@ const User = {
     return { ...rest };
   },
 
-  create: async function ({ username, password, role = "default" }) {
+  create: async function ({
+    username,
+    password,
+    role = "default",
+    dailyMessageLimit = null,
+  }) {
     const passwordCheck = this.checkPasswordComplexity(password);
     if (!passwordCheck.checkedOK) {
       return { user: null, error: passwordCheck.error };
@@ -58,7 +87,7 @@ const User = {
       // Do not allow new users to bypass validation
       if (!this.usernameRegex.test(username))
         throw new Error(
-          "Username must be only contain lowercase letters, numbers, underscores, and hyphens with no spaces"
+          "Username must only contain lowercase letters, numbers, underscores, and hyphens with no spaces"
         );
 
       const bcrypt = require("bcrypt");
@@ -68,6 +97,8 @@ const User = {
           username: this.validations.username(username),
           password: hashedPassword,
           role: this.validations.role(role),
+          dailyMessageLimit:
+            this.validations.dailyMessageLimit(dailyMessageLimit),
         },
       });
       return { user: this.filterFields(user), error: null };
@@ -135,7 +166,7 @@ const User = {
         return {
           success: false,
           error:
-            "Username must be only contain lowercase letters, numbers, underscores, and hyphens with no spaces",
+            "Username must only contain lowercase letters, numbers, underscores, and hyphens with no spaces",
         };
 
       const user = await prisma.users.update({
@@ -260,6 +291,29 @@ const User = {
 
     return { checkedOK: true, error: "No error." };
   },
+
+  /**
+   * Check if a user can send a chat based on their daily message limit.
+   * This limit is system wide and not per workspace and only applies to
+   * multi-user mode AND non-admin users.
+   * @param {User} user The user object record.
+   * @returns {Promise<boolean>} True if the user can send a chat, false otherwise.
+   */
+  canSendChat: async function (user) {
+    const { ROLES } = require("../utils/middleware/multiUserProtected");
+    if (!user || user.dailyMessageLimit === null || user.role === ROLES.admin)
+      return true;
+
+    const { WorkspaceChats } = require("./workspaceChats");
+    const currentChatCount = await WorkspaceChats.count({
+      user_id: user.id,
+      createdAt: {
+        gte: new Date(new Date() - 24 * 60 * 60 * 1000), // 24 hours
+      },
+    });
+
+    return currentChatCount < user.dailyMessageLimit;
+  },
 };
 
 module.exports = { User };

server/prisma/migrations/20241003192954_init/migration.sql

@@ -0,0 +1,2 @@
+-- AlterTable
+ALTER TABLE "users" ADD COLUMN "dailyMessageLimit" INTEGER;

server/prisma/schema.prisma

@@ -67,6 +67,7 @@ model users {
   seen_recovery_codes         Boolean?                      @default(false)
   createdAt                   DateTime                      @default(now())
   lastUpdatedAt               DateTime                      @default(now())
+  dailyMessageLimit           Int?
   workspace_chats             workspace_chats[]
   workspace_users             workspace_users[]
   embed_configs               embed_configs[]

server/prisma/seed.js

@@ -4,8 +4,6 @@ const prisma = new PrismaClient();
 async function main() {
   const settings = [
     { label: "multi_user_mode", value: "false" },
-    { label: "limit_user_messages", value: "false" },
-    { label: "message_limit", value: "25" },
     { label: "logo_filename", value: "anything-llm.png" },
   ];
 

server/storage/models/.gitignore

@@ -2,3 +2,4 @@ Xenova
 downloaded/*
 !downloaded/.placeholder
 openrouter
+apipie

server/swagger/openapi.json

@@ -693,52 +693,6 @@
       }
     },
     "/v1/admin/preferences": {
-      "get": {
-        "tags": [
-          "Admin"
-        ],
-        "description": "Show all multi-user preferences for instance. Methods are disabled until multi user mode is enabled via the UI.",
-        "parameters": [],
-        "responses": {
-          "200": {
-            "description": "OK",
-            "content": {
-              "application/json": {
-                "schema": {
-                  "type": "object",
-                  "example": {
-                    "settings": {
-                      "limit_user_messages": false,
-                      "message_limit": 10
-                    }
-                  }
-                }
-              }
-            }
-          },
-          "401": {
-            "description": "Instance is not in Multi-User mode. Method denied"
-          },
-          "403": {
-            "description": "Forbidden",
-            "content": {
-              "application/json": {
-                "schema": {
-                  "$ref": "#/components/schemas/InvalidAPIKey"
-                }
-              },
-              "application/xml": {
-                "schema": {
-                  "$ref": "#/components/schemas/InvalidAPIKey"
-                }
-              }
-            }
-          },
-          "500": {
-            "description": "Internal Server Error"
-          }
-        }
-      },
       "post": {
         "tags": [
           "Admin"
@@ -788,8 +742,7 @@
           "content": {
             "application/json": {
               "example": {
-                "limit_user_messages": true,
+                "support_email": "support@example.com"
-                "message_limit": 5
               }
             }
           }

server/utils/AiProviders/apipie/index.js

@@ -0,0 +1,336 @@
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
+const {
+  handleDefaultStreamResponseV2,
+} = require("../../helpers/chat/responses");
+
+const { v4: uuidv4 } = require("uuid");
+const {
+  writeResponseChunk,
+  clientAbortedHandler,
+} = require("../../helpers/chat/responses");
+
+const fs = require("fs");
+const path = require("path");
+const { safeJsonParse } = require("../../http");
+const cacheFolder = path.resolve(
+  process.env.STORAGE_DIR
+    ? path.resolve(process.env.STORAGE_DIR, "models", "apipie")
+    : path.resolve(__dirname, `../../../storage/models/apipie`)
+);
+
+class ApiPieLLM {
+  constructor(embedder = null, modelPreference = null) {
+    if (!process.env.APIPIE_LLM_API_KEY)
+      throw new Error("No ApiPie LLM API key was set.");
+
+    const { OpenAI: OpenAIApi } = require("openai");
+    this.basePath = "https://apipie.ai/v1";
+    this.openai = new OpenAIApi({
+      baseURL: this.basePath,
+      apiKey: process.env.APIPIE_LLM_API_KEY ?? null,
+    });
+    this.model =
+      modelPreference ||
+      process.env.APIPIE_LLM_MODEL_PREF ||
+      "openrouter/mistral-7b-instruct";
+    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;
+
+    if (!fs.existsSync(cacheFolder))
+      fs.mkdirSync(cacheFolder, { recursive: true });
+    this.cacheModelPath = path.resolve(cacheFolder, "models.json");
+    this.cacheAtPath = path.resolve(cacheFolder, ".cached_at");
+  }
+
+  log(text, ...args) {
+    console.log(`\x1b[36m[${this.constructor.name}]\x1b[0m ${text}`, ...args);
+  }
+
+  // This checks if the .cached_at file has a timestamp that is more than 1Week (in millis)
+  // from the current date. If it is, then we will refetch the API so that all the models are up
+  // to date.
+  #cacheIsStale() {
+    const MAX_STALE = 6.048e8; // 1 Week in MS
+    if (!fs.existsSync(this.cacheAtPath)) return true;
+    const now = Number(new Date());
+    const timestampMs = Number(fs.readFileSync(this.cacheAtPath));
+    return now - timestampMs > MAX_STALE;
+  }
+
+  // This function fetches the models from the ApiPie API and caches them locally.
+  // We do this because the ApiPie API has a lot of models, and we need to get the proper token context window
+  // for each model and this is a constructor property - so we can really only get it if this cache exists.
+  // We used to have this as a chore, but given there is an API to get the info - this makes little sense.
+  // This might slow down the first request, but we need the proper token context window
+  // for each model and this is a constructor property - so we can really only get it if this cache exists.
+  async #syncModels() {
+    if (fs.existsSync(this.cacheModelPath) && !this.#cacheIsStale())
+      return false;
+
+    this.log("Model cache is not present or stale. Fetching from ApiPie API.");
+    await fetchApiPieModels();
+    return;
+  }
+
+  #appendContext(contextTexts = []) {
+    if (!contextTexts || !contextTexts.length) return "";
+    return (
+      "\nContext:\n" +
+      contextTexts
+        .map((text, i) => {
+          return `[CONTEXT ${i}]:\n${text}\n[END CONTEXT ${i}]\n\n`;
+        })
+        .join("")
+    );
+  }
+
+  models() {
+    if (!fs.existsSync(this.cacheModelPath)) return {};
+    return safeJsonParse(
+      fs.readFileSync(this.cacheModelPath, { encoding: "utf-8" }),
+      {}
+    );
+  }
+
+  streamingEnabled() {
+    return "streamGetChatCompletion" in this;
+  }
+
+  static promptWindowLimit(modelName) {
+    const cacheModelPath = path.resolve(cacheFolder, "models.json");
+    const availableModels = fs.existsSync(cacheModelPath)
+      ? safeJsonParse(
+          fs.readFileSync(cacheModelPath, { encoding: "utf-8" }),
+          {}
+        )
+      : {};
+    return availableModels[modelName]?.maxLength || 4096;
+  }
+
+  promptWindowLimit() {
+    const availableModels = this.models();
+    return availableModels[this.model]?.maxLength || 4096;
+  }
+
+  async isValidChatCompletionModel(model = "") {
+    await this.#syncModels();
+    const availableModels = this.models();
+    return availableModels.hasOwnProperty(model);
+  }
+
+  /**
+   * Generates appropriate content array for a message + attachments.
+   * @param {{userPrompt:string, attachments: import("../../helpers").Attachment[]}}
+   * @returns {string|object[]}
+   */
+  #generateContent({ userPrompt, attachments = [] }) {
+    if (!attachments.length) {
+      return userPrompt;
+    }
+
+    const content = [{ type: "text", text: userPrompt }];
+    for (let attachment of attachments) {
+      content.push({
+        type: "image_url",
+        image_url: {
+          url: attachment.contentString,
+          detail: "auto",
+        },
+      });
+    }
+    return content.flat();
+  }
+
+  constructPrompt({
+    systemPrompt = "",
+    contextTexts = [],
+    chatHistory = [],
+    userPrompt = "",
+    attachments = [],
+  }) {
+    const prompt = {
+      role: "system",
+      content: `${systemPrompt}${this.#appendContext(contextTexts)}`,
+    };
+    return [
+      prompt,
+      ...chatHistory,
+      {
+        role: "user",
+        content: this.#generateContent({ userPrompt, attachments }),
+      },
+    ];
+  }
+
+  async getChatCompletion(messages = null, { temperature = 0.7 }) {
+    if (!(await this.isValidChatCompletionModel(this.model)))
+      throw new Error(
+        `ApiPie chat: ${this.model} is not valid for chat completion!`
+      );
+
+    const result = await this.openai.chat.completions
+      .create({
+        model: this.model,
+        messages,
+        temperature,
+      })
+      .catch((e) => {
+        throw new Error(e.message);
+      });
+
+    if (!result.hasOwnProperty("choices") || result.choices.length === 0)
+      return null;
+    return result.choices[0].message.content;
+  }
+
+  // APIPie says it supports streaming, but it does not work across all models and providers.
+  // Notably, it is not working for OpenRouter models at all.
+  // async streamGetChatCompletion(messages = null, { temperature = 0.7 }) {
+  //   if (!(await this.isValidChatCompletionModel(this.model)))
+  //     throw new Error(
+  //       `ApiPie chat: ${this.model} is not valid for chat completion!`
+  //     );
+
+  //   const streamRequest = await this.openai.chat.completions.create({
+  //     model: this.model,
+  //     stream: true,
+  //     messages,
+  //     temperature,
+  //   });
+  //   return streamRequest;
+  // }
+
+  handleStream(response, stream, responseProps) {
+    const { uuid = uuidv4(), sources = [] } = responseProps;
+
+    return new Promise(async (resolve) => {
+      let fullText = "";
+
+      // Establish listener to early-abort a streaming response
+      // in case things go sideways or the user does not like the response.
+      // We preserve the generated text but continue as if chat was completed
+      // to preserve previously generated content.
+      const handleAbort = () => clientAbortedHandler(resolve, fullText);
+      response.on("close", handleAbort);
+
+      try {
+        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,
+            });
+          }
+
+          if (message === undefined || message.finish_reason !== null) {
+            writeResponseChunk(response, {
+              uuid,
+              sources,
+              type: "textResponseChunk",
+              textResponse: "",
+              close: true,
+              error: false,
+            });
+            response.removeListener("close", handleAbort);
+            resolve(fullText);
+          }
+        }
+      } catch (e) {
+        writeResponseChunk(response, {
+          uuid,
+          sources,
+          type: "abort",
+          textResponse: null,
+          close: true,
+          error: e.message,
+        });
+        response.removeListener("close", handleAbort);
+        resolve(fullText);
+      }
+    });
+  }
+
+  // handleStream(response, stream, responseProps) {
+  //   return handleDefaultStreamResponseV2(response, stream, responseProps);
+  // }
+
+  // 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);
+  }
+}
+
+async function fetchApiPieModels(providedApiKey = null) {
+  const apiKey = providedApiKey || process.env.APIPIE_LLM_API_KEY || null;
+  return await fetch(`https://apipie.ai/v1/models`, {
+    method: "GET",
+    headers: {
+      "Content-Type": "application/json",
+      ...(apiKey ? { Authorization: `Bearer ${apiKey}` } : {}),
+    },
+  })
+    .then((res) => res.json())
+    .then(({ data = [] }) => {
+      const models = {};
+      data.forEach((model) => {
+        models[`${model.provider}/${model.model}`] = {
+          id: `${model.provider}/${model.model}`,
+          name: `${model.provider}/${model.model}`,
+          organization: model.provider,
+          maxLength: model.max_tokens,
+        };
+      });
+
+      // Cache all response information
+      if (!fs.existsSync(cacheFolder))
+        fs.mkdirSync(cacheFolder, { recursive: true });
+      fs.writeFileSync(
+        path.resolve(cacheFolder, "models.json"),
+        JSON.stringify(models),
+        {
+          encoding: "utf-8",
+        }
+      );
+      fs.writeFileSync(
+        path.resolve(cacheFolder, ".cached_at"),
+        String(Number(new Date())),
+        {
+          encoding: "utf-8",
+        }
+      );
+
+      return models;
+    })
+    .catch((e) => {
+      console.error(e);
+      return {};
+    });
+}
+
+module.exports = {
+  ApiPieLLM,
+  fetchApiPieModels,
+};

server/utils/agents/aibitat/index.js

@@ -785,6 +785,10 @@ ${this.getHistory({ to: route.to })
         return new Providers.FireworksAIProvider({ model: config.model });
       case "deepseek":
         return new Providers.DeepSeekProvider({ model: config.model });
+      case "litellm":
+        return new Providers.LiteLLMProvider({ model: config.model });
+      case "apipie":
+        return new Providers.ApiPieProvider({ model: config.model });
 
       default:
         throw new Error(

server/utils/agents/aibitat/providers/ai-provider.js

@@ -130,6 +130,22 @@ class Provider {
           apiKey: process.env.FIREWORKS_AI_LLM_API_KEY,
           ...config,
         });
+      case "apipie":
+        return new ChatOpenAI({
+          configuration: {
+            baseURL: "https://apipie.ai/v1",
+          },
+          apiKey: process.env.APIPIE_LLM_API_KEY ?? null,
+          ...config,
+        });
+      case "deepseek":
+        return new ChatOpenAI({
+          configuration: {
+            baseURL: "https://api.deepseek.com/v1",
+          },
+          apiKey: process.env.DEEPSEEK_API_KEY ?? null,
+          ...config,
+        });
 
       // OSS Model Runners
       // case "anythingllm_ollama":
@@ -174,14 +190,15 @@ class Provider {
           apiKey: process.env.TEXT_GEN_WEB_UI_API_KEY ?? "not-used",
           ...config,
         });
-      case "deepseek":
+      case "litellm":
         return new ChatOpenAI({
           configuration: {
-            baseURL: "https://api.deepseek.com/v1",
+            baseURL: process.env.LITE_LLM_BASE_PATH,
           },
-          apiKey: process.env.DEEPSEEK_API_KEY ?? null,
+          apiKey: process.env.LITE_LLM_API_KEY ?? null,
           ...config,
         });
+
       default:
         throw new Error(`Unsupported provider ${provider} for this task.`);
     }

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

@@ -0,0 +1,116 @@
+const OpenAI = require("openai");
+const Provider = require("./ai-provider.js");
+const InheritMultiple = require("./helpers/classes.js");
+const UnTooled = require("./helpers/untooled.js");
+
+/**
+ * The agent provider for the OpenRouter provider.
+ */
+class ApiPieProvider extends InheritMultiple([Provider, UnTooled]) {
+  model;
+
+  constructor(config = {}) {
+    const { model = "openrouter/llama-3.1-8b-instruct" } = config;
+    super();
+    const client = new OpenAI({
+      baseURL: "https://apipie.ai/v1",
+      apiKey: process.env.APIPIE_LLM_API_KEY,
+      maxRetries: 3,
+    });
+
+    this._client = client;
+    this.model = model;
+    this.verbose = true;
+  }
+
+  get client() {
+    return this._client;
+  }
+
+  async #handleFunctionCallChat({ messages = [] }) {
+    return await this.client.chat.completions
+      .create({
+        model: this.model,
+        temperature: 0,
+        messages,
+      })
+      .then((result) => {
+        if (!result.hasOwnProperty("choices"))
+          throw new Error("ApiPie chat: No results!");
+        if (result.choices.length === 0)
+          throw new Error("ApiPie chat: No results length!");
+        return result.choices[0].message.content;
+      })
+      .catch((_) => {
+        return null;
+      });
+  }
+
+  /**
+   * Create a completion based on the received messages.
+   *
+   * @param messages A list of messages to send to the API.
+   * @param functions
+   * @returns The completion.
+   */
+  async complete(messages, functions = null) {
+    try {
+      let completion;
+      if (functions.length > 0) {
+        const { toolCall, text } = await this.functionCall(
+          messages,
+          functions,
+          this.#handleFunctionCallChat.bind(this)
+        );
+
+        if (toolCall !== null) {
+          this.providerLog(`Valid tool call found - running ${toolCall.name}.`);
+          this.deduplicator.trackRun(toolCall.name, toolCall.arguments);
+          return {
+            result: null,
+            functionCall: {
+              name: toolCall.name,
+              arguments: toolCall.arguments,
+            },
+            cost: 0,
+          };
+        }
+        completion = { content: text };
+      }
+
+      if (!completion?.content) {
+        this.providerLog(
+          "Will assume chat completion without tool call inputs."
+        );
+        const response = await this.client.chat.completions.create({
+          model: this.model,
+          messages: this.cleanMsgs(messages),
+        });
+        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;
+    }
+  }
+
+  /**
+   * Get the cost of the completion.
+   *
+   * @param _usage The completion to get the cost for.
+   * @returns The cost of the completion.
+   */
+  getCost(_usage) {
+    return 0;
+  }
+}
+
+module.exports = ApiPieProvider;

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

@@ -15,6 +15,8 @@ const TextWebGenUiProvider = require("./textgenwebui.js");
 const AWSBedrockProvider = require("./bedrock.js");
 const FireworksAIProvider = require("./fireworksai.js");
 const DeepSeekProvider = require("./deepseek.js");
+const LiteLLMProvider = require("./litellm.js");
+const ApiPieProvider = require("./apipie.js");
 
 module.exports = {
   OpenAIProvider,
@@ -34,4 +36,6 @@ module.exports = {
   TextWebGenUiProvider,
   AWSBedrockProvider,
   FireworksAIProvider,
+  LiteLLMProvider,
+  ApiPieProvider,
 };

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

@@ -0,0 +1,110 @@
+const OpenAI = require("openai");
+const Provider = require("./ai-provider.js");
+const InheritMultiple = require("./helpers/classes.js");
+const UnTooled = require("./helpers/untooled.js");
+
+/**
+ * The agent provider for LiteLLM.
+ */
+class LiteLLMProvider extends InheritMultiple([Provider, UnTooled]) {
+  model;
+
+  constructor(config = {}) {
+    super();
+    const { model = null } = config;
+    const client = new OpenAI({
+      baseURL: process.env.LITE_LLM_BASE_PATH,
+      apiKey: process.env.LITE_LLM_API_KEY ?? null,
+      maxRetries: 3,
+    });
+
+    this._client = client;
+    this.model = model || process.env.LITE_LLM_MODEL_PREF;
+    this.verbose = true;
+  }
+
+  get client() {
+    return this._client;
+  }
+
+  async #handleFunctionCallChat({ messages = [] }) {
+    return await this.client.chat.completions
+      .create({
+        model: this.model,
+        temperature: 0,
+        messages,
+      })
+      .then((result) => {
+        if (!result.hasOwnProperty("choices"))
+          throw new Error("LiteLLM chat: No results!");
+        if (result.choices.length === 0)
+          throw new Error("LiteLLM chat: No results length!");
+        return result.choices[0].message.content;
+      })
+      .catch((_) => {
+        return null;
+      });
+  }
+
+  /**
+   * Create a completion based on the received messages.
+   *
+   * @param messages A list of messages to send to the API.
+   * @param functions
+   * @returns The completion.
+   */
+  async complete(messages, functions = null) {
+    try {
+      let completion;
+      if (functions.length > 0) {
+        const { toolCall, text } = await this.functionCall(
+          messages,
+          functions,
+          this.#handleFunctionCallChat.bind(this)
+        );
+
+        if (toolCall !== null) {
+          this.providerLog(`Valid tool call found - running ${toolCall.name}.`);
+          this.deduplicator.trackRun(toolCall.name, toolCall.arguments);
+          return {
+            result: null,
+            functionCall: {
+              name: toolCall.name,
+              arguments: toolCall.arguments,
+            },
+            cost: 0,
+          };
+        }
+        completion = { content: text };
+      }
+
+      if (!completion?.content) {
+        this.providerLog(
+          "Will assume chat completion without tool call inputs."
+        );
+        const response = await this.client.chat.completions.create({
+          model: this.model,
+          messages: this.cleanMsgs(messages),
+        });
+        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;
+    }
+  }
+
+  getCost(_usage) {
+    return 0;
+  }
+}
+
+module.exports = LiteLLMProvider;

server/utils/agents/index.js

@@ -166,6 +166,16 @@ class AgentHandler {
         if (!process.env.DEEPSEEK_API_KEY)
           throw new Error("DeepSeek API Key must be provided to use agents.");
         break;
+      case "litellm":
+        if (!process.env.LITE_LLM_BASE_PATH)
+          throw new Error(
+            "LiteLLM API base path and key must be provided to use agents."
+          );
+        break;
+      case "apipie":
+        if (!process.env.APIPIE_LLM_API_KEY)
+          throw new Error("ApiPie API Key must be provided to use agents.");
+        break;
 
       default:
         throw new Error(
@@ -212,6 +222,10 @@ class AgentHandler {
         return null;
       case "deepseek":
         return "deepseek-chat";
+      case "litellm":
+        return null;
+      case "apipie":
+        return null;
       default:
         return "unknown";
     }

server/utils/chats/embed.js

@@ -60,8 +60,7 @@ async function streamChatWithForEmbed(
   const { rawHistory, chatHistory } = await recentEmbedChatHistory(
     sessionId,
     embed,
-    messageLimit,
+    messageLimit
-    chatMode
   );
 
   // See stream.js comment for more information on this implementation.
@@ -113,16 +112,27 @@ async function streamChatWithForEmbed(
     return;
   }
 
-  contextTexts = [...contextTexts, ...vectorSearchResults.contextTexts];
+  const { fillSourceWindow } = require("../helpers/chat");
+  const filledSources = fillSourceWindow({
+    nDocs: embed.workspace?.topN || 4,
+    searchResults: vectorSearchResults.sources,
+    history: rawHistory,
+    filterIdentifiers: pinnedDocIdentifiers,
+  });
+
+  // Why does contextTexts get all the info, but sources only get current search?
+  // This is to give the ability of the LLM to "comprehend" a contextual response without
+  // populating the Citations under a response with documents the user "thinks" are irrelevant
+  // due to how we manage backfilling of the context to keep chats with the LLM more correct in responses.
+  // If a past citation was used to answer the question - that is visible in the history so it logically makes sense
+  // and does not appear to the user that a new response used information that is otherwise irrelevant for a given prompt.
+  // TLDR; reduces GitHub issues for "LLM citing document that has no answer in it" while keep answers highly accurate.
+  contextTexts = [...contextTexts, ...filledSources.contextTexts];
   sources = [...sources, ...vectorSearchResults.sources];
 
-  // If in query mode and no sources are found, do not
+  // If in query mode and no sources are found in current search or backfilled from history, do not
   // let the LLM try to hallucinate a response or use general knowledge
-  if (
+  if (chatMode === "query" && contextTexts.length === 0) {
-    chatMode === "query" &&
-    sources.length === 0 &&
-    pinnedDocIdentifiers.length === 0
-  ) {
     writeResponseChunk(response, {
       id: uuid,
       type: "textResponse",
@@ -178,7 +188,7 @@ async function streamChatWithForEmbed(
   await EmbedChats.new({
     embedId: embed.id,
     prompt: message,
-    response: { text: completeText, type: chatMode },
+    response: { text: completeText, type: chatMode, sources },
     connection_information: response.locals.connection
       ? {
           ...response.locals.connection,
@@ -190,15 +200,13 @@ async function streamChatWithForEmbed(
   return;
 }
 
-// On query we don't return message history. All other chat modes and when chatting
+/**
-// with no embeddings we return history.
+ * @param {string} sessionId the session id of the user from embed widget
-async function recentEmbedChatHistory(
+ * @param {Object} embed the embed config object
-  sessionId,
+ * @param {Number} messageLimit the number of messages to return
-  embed,
+ * @returns {Promise<{rawHistory: import("@prisma/client").embed_chats[], chatHistory: {role: string, content: string}[]}>
-  messageLimit = 20,
+ */
-  chatMode = null
+async function recentEmbedChatHistory(sessionId, embed, messageLimit = 20) {
-) {
-  if (chatMode === "query") return { rawHistory: [], chatHistory: [] };
   const rawHistory = (
     await EmbedChats.forEmbedByUser(embed.id, sessionId, messageLimit, {
       id: "desc",

server/utils/helpers/customModels.js

@@ -1,4 +1,5 @@
 const { fetchOpenRouterModels } = require("../AiProviders/openRouter");
+const { fetchApiPieModels } = require("../AiProviders/apipie");
 const { perplexityModels } = require("../AiProviders/perplexity");
 const { togetherAiModels } = require("../AiProviders/togetherAi");
 const { fireworksAiModels } = require("../AiProviders/fireworksAi");
@@ -19,6 +20,7 @@ const SUPPORT_CUSTOM_MODELS = [
   "elevenlabs-tts",
   "groq",
   "deepseek",
+  "apipie",
 ];
 
 async function getCustomModels(provider = "", apiKey = null, basePath = null) {
@@ -56,6 +58,8 @@ async function getCustomModels(provider = "", apiKey = null, basePath = null) {
       return await getGroqAiModels(apiKey);
     case "deepseek":
       return await getDeepSeekModels(apiKey);
+    case "apipie":
+      return await getAPIPieModels(apiKey);
     default:
       return { models: [], error: "Invalid provider for custom models" };
   }
@@ -355,6 +359,21 @@ async function getOpenRouterModels() {
   return { models, error: null };
 }
 
+async function getAPIPieModels(apiKey = null) {
+  const knownModels = await fetchApiPieModels(apiKey);
+  if (!Object.keys(knownModels).length === 0)
+    return { models: [], error: null };
+
+  const models = Object.values(knownModels).map((model) => {
+    return {
+      id: model.id,
+      organization: model.organization,
+      name: model.name,
+    };
+  });
+  return { models, error: null };
+}
+
 async function getMistralModels(apiKey = null) {
   const { OpenAI: OpenAIApi } = require("openai");
   const openai = new OpenAIApi({

server/utils/helpers/index.js

@@ -162,6 +162,9 @@ function getLLMProvider({ provider = null, model = null } = {}) {
     case "deepseek":
       const { DeepSeekLLM } = require("../AiProviders/deepseek");
       return new DeepSeekLLM(embedder, model);
+    case "apipie":
+      const { ApiPieLLM } = require("../AiProviders/apipie");
+      return new ApiPieLLM(embedder, model);
     default:
       throw new Error(
         `ENV: No valid LLM_PROVIDER value found in environment! Using ${process.env.LLM_PROVIDER}`
@@ -285,6 +288,12 @@ function getLLMProviderClass({ provider = null } = {}) {
     case "bedrock":
       const { AWSBedrockLLM } = require("../AiProviders/bedrock");
       return AWSBedrockLLM;
+    case "deepseek":
+      const { DeepSeekLLM } = require("../AiProviders/deepseek");
+      return DeepSeekLLM;
+    case "apipie":
+      const { ApiPieLLM } = require("../AiProviders/apipie");
+      return ApiPieLLM;
     default:
       return null;
   }

server/utils/helpers/updateENV.js

@@ -515,6 +515,16 @@ const KEY_MAPPING = {
     envKey: "DEEPSEEK_MODEL_PREF",
     checks: [isNotEmpty],
   },
+
+  // APIPie Options
+  ApipieLLMApiKey: {
+    envKey: "APIPIE_LLM_API_KEY",
+    checks: [isNotEmpty],
+  },
+  ApipieLLMModelPref: {
+    envKey: "APIPIE_LLM_MODEL_PREF",
+    checks: [isNotEmpty],
+  },
 };
 
 function isNotEmpty(input = "") {
@@ -617,6 +627,7 @@ function supportedLLM(input = "") {
     "generic-openai",
     "bedrock",
     "deepseek",
+    "apipie",
   ].includes(input);
   return validSelection ? null : `${input} is not a valid LLM provider.`;
 }