Z.ai LLM & agent provider (#4573)

* wip zai llm provider

* cleanup + add zai agent provider

* lint

* change how caching works for failed models

---------

Co-authored-by: Timothy Carambat <rambat1010@gmail.com>

README.md

@@ -99,6 +99,7 @@ AnythingLLM divides your documents into objects called `workspaces`. A Workspace
 - [Text Generation Web UI](https://github.com/oobabooga/text-generation-webui)
 - [Apipie](https://apipie.ai/)
 - [xAI](https://x.ai/)
+- [Z.AI (chat models)](https://z.ai/model-api)
 - [Novita AI (chat models)](https://novita.ai/model-api/product/llm-api?utm_source=github_anything-llm&utm_medium=github_readme&utm_campaign=link)
 - [PPIO](https://ppinfra.com?utm_source=github_anything-llm)
 - [Moonshot AI](https://www.moonshot.ai/)

docker/.env.example

@@ -131,6 +131,10 @@ GID='1000'
 # XAI_LLM_API_KEY='xai-your-api-key-here'
 # XAI_LLM_MODEL_PREF='grok-beta'
 
+# LLM_PROVIDER='zai'
+# ZAI_API_KEY="your-zai-api-key-here"
+# ZAI_MODEL_PREF="glm-4.5"
+
 # LLM_PROVIDER='nvidia-nim'
 # NVIDIA_NIM_LLM_BASE_PATH='http://127.0.0.1:8000'
 # NVIDIA_NIM_LLM_MODEL_PREF='meta/llama-3.2-3b-instruct'

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

@@ -0,0 +1,114 @@
+import { useState, useEffect } from "react";
+import System from "@/models/system";
+
+export default function ZAiLLMOptions({ settings }) {
+  const [inputValue, setInputValue] = useState(settings?.ZAiApiKey);
+  const [apiKey, setApiKey] = useState(settings?.ZAiApiKey);
+
+  return (
+    <div className="flex gap-[36px] mt-1.5">
+      <div className="flex flex-col w-60">
+        <label className="text-white text-sm font-semibold block mb-3">
+          Z.AI API Key
+        </label>
+        <input
+          type="password"
+          name="ZAiApiKey"
+          className="border-none 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-full p-2.5"
+          placeholder="Z.AI API Key"
+          defaultValue={settings?.ZAiApiKey ? "*".repeat(20) : ""}
+          required={true}
+          autoComplete="off"
+          spellCheck={false}
+          onChange={(e) => setInputValue(e.target.value)}
+          onBlur={() => setApiKey(inputValue)}
+        />
+      </div>
+
+      {!settings?.credentialsOnly && (
+        <ZAiModelSelection settings={settings} apiKey={apiKey} />
+      )}
+    </div>
+  );
+}
+
+function ZAiModelSelection({ apiKey, settings }) {
+  const [customModels, setCustomModels] = useState([]);
+  const [loading, setLoading] = useState(true);
+
+  useEffect(() => {
+    async function findCustomModels() {
+      if (!apiKey) {
+        setCustomModels([]);
+        setLoading(true);
+        return;
+      }
+
+      try {
+        setLoading(true);
+        const { models } = await System.customModels("zai", apiKey);
+        setCustomModels(models || []);
+      } catch (error) {
+        console.error("Failed to fetch custom models:", error);
+        setCustomModels([]);
+      } finally {
+        setLoading(false);
+      }
+    }
+    findCustomModels();
+  }, [apiKey]);
+
+  if (loading) {
+    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="ZAiModelPref"
+          disabled={true}
+          className="border-none bg-theme-settings-input-bg 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>
+        <p className="text-xs leading-[18px] font-base text-white text-opacity-60 mt-2">
+          Enter a valid API key to view all available models for your account.
+        </p>
+      </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="ZAiModelPref"
+        required={true}
+        className="border-none bg-theme-settings-input-bg border-gray-500 text-white text-sm rounded-lg block w-full p-2.5"
+      >
+        {customModels.length > 0 && (
+          <optgroup label="Available models">
+            {customModels.map((model) => {
+              return (
+                <option
+                  key={model.id}
+                  value={model.id}
+                  selected={settings?.ZAiModelPref === model.id}
+                >
+                  {model.id}
+                </option>
+              );
+            })}
+          </optgroup>
+        )}
+      </select>
+      <p className="text-xs leading-[18px] font-base text-white text-opacity-60 mt-2">
+        Select the Z.AI model you want to use for your conversations.
+      </p>
+    </div>
+  );
+}

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

@@ -29,6 +29,7 @@ import AWSBedrockLogo from "@/media/llmprovider/bedrock.png";
 import DeepSeekLogo from "@/media/llmprovider/deepseek.png";
 import APIPieLogo from "@/media/llmprovider/apipie.png";
 import XAILogo from "@/media/llmprovider/xai.png";
+import ZAiLogo from "@/media/llmprovider/zai.png";
 import NvidiaNimLogo from "@/media/llmprovider/nvidia-nim.png";
 import PPIOLogo from "@/media/llmprovider/ppio.png";
 import DellProAiStudioLogo from "@/media/llmprovider/dpais.png";
@@ -62,6 +63,7 @@ import AWSBedrockLLMOptions from "@/components/LLMSelection/AwsBedrockLLMOptions
 import DeepSeekOptions from "@/components/LLMSelection/DeepSeekOptions";
 import ApiPieLLMOptions from "@/components/LLMSelection/ApiPieOptions";
 import XAILLMOptions from "@/components/LLMSelection/XAiLLMOptions";
+import ZAiLLMOptions from "@/components/LLMSelection/ZAiLLMOptions";
 import NvidiaNimOptions from "@/components/LLMSelection/NvidiaNimOptions";
 import PPIOLLMOptions from "@/components/LLMSelection/PPIOLLMOptions";
 import DellProAiStudioOptions from "@/components/LLMSelection/DPAISOptions";
@@ -335,6 +337,14 @@ export const AVAILABLE_LLM_PROVIDERS = [
     description: "Run xAI's powerful LLMs like Grok-2 and more.",
     requiredConfig: ["XAIApiKey", "XAIModelPref"],
   },
+  {
+    name: "Z.AI",
+    value: "zai",
+    logo: ZAiLogo,
+    options: (settings) => <ZAiLLMOptions settings={settings} />,
+    description: "Run Z.AI's powerful GLM models.",
+    requiredConfig: ["ZAiApiKey"],
+  },
   {
     name: "Generic OpenAI",
     value: "generic-openai",

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

@@ -25,6 +25,7 @@ import AWSBedrockLogo from "@/media/llmprovider/bedrock.png";
 import DeepSeekLogo from "@/media/llmprovider/deepseek.png";
 import APIPieLogo from "@/media/llmprovider/apipie.png";
 import XAILogo from "@/media/llmprovider/xai.png";
+import ZAiLogo from "@/media/llmprovider/zai.png";
 import CohereLogo from "@/media/llmprovider/cohere.png";
 import ZillizLogo from "@/media/vectordbs/zilliz.png";
 import AstraDBLogo from "@/media/vectordbs/astraDB.png";
@@ -231,6 +232,15 @@ export const LLM_SELECTION_PRIVACY = {
     ],
     logo: XAILogo,
   },
+  zai: {
+    name: "Z.AI",
+    description: [
+      "Your content is processed in real-time and not stored on Z.AI servers",
+      "Your prompts and document text are visible to Z.AI during processing",
+      "Data is processed in accordance with Z.AI's API Services terms",
+    ],
+    logo: ZAiLogo,
+  },
   ppio: {
     name: "PPIO",
     description: [

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

@@ -23,6 +23,7 @@ import DeepSeekLogo from "@/media/llmprovider/deepseek.png";
 import APIPieLogo from "@/media/llmprovider/apipie.png";
 import NovitaLogo from "@/media/llmprovider/novita.png";
 import XAILogo from "@/media/llmprovider/xai.png";
+import ZAiLogo from "@/media/llmprovider/zai.png";
 import NvidiaNimLogo from "@/media/llmprovider/nvidia-nim.png";
 import CohereLogo from "@/media/llmprovider/cohere.png";
 import PPIOLogo from "@/media/llmprovider/ppio.png";
@@ -54,6 +55,7 @@ import DeepSeekOptions from "@/components/LLMSelection/DeepSeekOptions";
 import ApiPieLLMOptions from "@/components/LLMSelection/ApiPieOptions";
 import NovitaLLMOptions from "@/components/LLMSelection/NovitaLLMOptions";
 import XAILLMOptions from "@/components/LLMSelection/XAiLLMOptions";
+import ZAiLLMOptions from "@/components/LLMSelection/ZAiLLMOptions";
 import NvidiaNimOptions from "@/components/LLMSelection/NvidiaNimOptions";
 import PPIOLLMOptions from "@/components/LLMSelection/PPIOLLMOptions";
 import DellProAiStudioOptions from "@/components/LLMSelection/DPAISOptions";
@@ -267,6 +269,13 @@ const LLMS = [
     options: (settings) => <XAILLMOptions settings={settings} />,
     description: "Run xAI's powerful LLMs like Grok-2 and more.",
   },
+  {
+    name: "Z.AI",
+    value: "zai",
+    logo: ZAiLogo,
+    options: (settings) => <ZAiLLMOptions settings={settings} />,
+    description: "Run Z.AI's powerful GLM models.",
+  },
   {
     name: "Moonshot AI",
     value: "moonshotai",

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

@@ -34,6 +34,7 @@ const ENABLED_PROVIDERS = [
   "moonshotai",
   "cometapi",
   "foundry",
+  "zai",
   // 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.

locales/README.fa-IR.md

@@ -102,6 +102,7 @@ AnythingLLM اسناد شما را به اشیایی به نام `workspaces` ت
 - [Text Generation Web UI](https://github.com/oobabooga/text-generation-webui)
 - [Apipie](https://apipie.ai/)
 - [xAI](https://x.ai/)
+- [Z.AI (chat models)](https://z.ai/model-api)
 - [Novita AI (chat models)](https://novita.ai/model-api/product/llm-api?utm_source=github_anything-llm&utm_medium=github_readme&utm_campaign=link)
 - [PPIO](https://ppinfra.com?utm_source=github_anything-llm)
 

locales/README.ja-JP.md

@@ -90,6 +90,8 @@ AnythingLLMは、ドキュメントを`ワークスペース`と呼ばれるオ
 - [Groq](https://groq.com/)
 - [Cohere](https://cohere.com/)
 - [KoboldCPP](https://github.com/LostRuins/koboldcpp)
+- [xAI](https://x.ai/)
+- [Z.AI (チャットモデル)](https://z.ai/model-api)
 - [PPIO](https://ppinfra.com?utm_source=github_anything-llm)
 - [CometAPI (チャットモデル)](https://api.cometapi.com/)
 

locales/README.tr-TR.md

@@ -99,6 +99,7 @@ AnythingLLM, belgelerinizi **"çalışma alanları" (workspaces)** adı verilen
 - [Text Generation Web UI](https://github.com/oobabooga/text-generation-webui)
 - [Apipie](https://apipie.ai/)
 - [xAI](https://x.ai/)
+- [Z.AI (chat models)](https://z.ai/model-api)
 - [Novita AI (chat models)](https://novita.ai/model-api/product/llm-api?utm_source=github_anything-llm&utm_medium=github_readme&utm_campaign=link)
 - [PPIO](https://ppinfra.com?utm_source=github_anything-llm)
 

locales/README.zh-CN.md

@@ -98,6 +98,7 @@ AnythingLLM将您的文档划分为称为`workspaces` (工作区)的对象。工
 - [Text Generation Web UI](https://github.com/oobabooga/text-generation-webui)
 - [Apipie](https://apipie.ai/)
 - [xAI](https://x.ai/)
+- [Z.AI (聊天模型)](https://z.ai/model-api)
 - [Novita AI (聊天模型)](https://novita.ai/model-api/product/llm-api?utm_source=github_anything-llm&utm_medium=github_readme&utm_campaign=link)
 - [PPIO (聊天模型)](https://ppinfra.com?utm_source=github_anything-llm)
 - [CometAPI (聊天模型)](https://api.cometapi.com/)

server/.env.example

@@ -134,6 +134,10 @@ SIG_SALT='salt' # Please generate random string at least 32 chars long.
 # XAI_LLM_API_KEY='xai-your-api-key-here'
 # XAI_LLM_MODEL_PREF='grok-beta'
 
+# LLM_PROVIDER='zai'
+# ZAI_API_KEY="your-zai-api-key-here"
+# ZAI_MODEL_PREF="glm-4.5"
+
 # LLM_PROVIDER='nvidia-nim'
 # NVIDIA_NIM_LLM_BASE_PATH='http://127.0.0.1:8000'
 # NVIDIA_NIM_LLM_MODEL_PREF='meta/llama-3.2-3b-instruct'

server/endpoints/utils.js

@@ -145,6 +145,9 @@ function getModelTag() {
     case "moonshotai":
       model = process.env.MOONSHOT_AI_MODEL_PREF;
       break;
+    case "zai":
+      model = process.env.ZAI_MODEL_PREF;
+      break;
     default:
       model = "--";
       break;

server/models/systemSettings.js

@@ -625,6 +625,10 @@ const SystemSettings = {
       CometApiLLMApiKey: !!process.env.COMETAPI_LLM_API_KEY,
       CometApiLLMModelPref: process.env.COMETAPI_LLM_MODEL_PREF,
       CometApiLLMTimeout: process.env.COMETAPI_LLM_TIMEOUT_MS,
+
+      // Z.AI Keys
+      ZAiApiKey: !!process.env.ZAI_API_KEY,
+      ZAiModelPref: process.env.ZAI_MODEL_PREF,
     };
   },
 

server/utils/AiProviders/modelMap/index.js

@@ -19,6 +19,7 @@ class ContextWindowFinder {
     xai: "xai",
     deepseek: "deepseek",
     moonshot: "moonshot",
+    zai: "vercel_ai_gateway", // Vercel has correct context windows for Z.AI models
   };
   static expiryMs = 1000 * 60 * 60 * 24 * 3; // 3 days
   static remoteUrl =
@@ -116,8 +117,9 @@ You can fix this by restarting AnythingLLM so the model map is re-pulled.
         });
       if (!remoteContexWindowMap) return null;
 
-      const modelMap = this.#formatModelMap(remoteContexWindowMap);
+      const modelMap = this.#validateModelMap(
-      this.#validateModelMap(modelMap);
+        this.#formatModelMap(remoteContexWindowMap)
+      );
       fs.writeFileSync(this.cacheFilePath, JSON.stringify(modelMap, null, 2));
       fs.writeFileSync(this.cacheFileExpiryPath, Date.now().toString());
       return modelMap;
@@ -139,13 +141,17 @@ You can fix this by restarting AnythingLLM so the model map is re-pulled.
 
       // Validate that the context window is a number
       for (const [model, contextWindow] of Object.entries(models)) {
-        if (isNaN(contextWindow) || contextWindow <= 0)
+        if (isNaN(contextWindow) || contextWindow <= 0) {
-          throw new Error(
+          this.log(
-            `Invalid model map for ${provider} - context window is not a positive number for model ${model}`
+            `${provider}:${model} - context window is not a positive number. Got ${contextWindow}.`
           );
+          delete models[model];
+          continue;
         }
       }
     }
+    return modelMap;
+  }
 
   /**
    * Formats the remote model map to a format that is compatible with how we store the model map

server/utils/AiProviders/ollama/index.js

@@ -435,7 +435,8 @@ class OllamaAILLM {
           type: "textResponseChunk",
           textResponse: "",
           close: true,
-          error: `Ollama:streaming - could not stream chat. ${error?.cause ?? error.message
+          error: `Ollama:streaming - could not stream chat. ${
+            error?.cause ?? error.message
           }`,
         });
         response.removeListener("close", handleAbort);

server/utils/AiProviders/zai/index.js

@@ -0,0 +1,179 @@
+const { NativeEmbedder } = require("../../EmbeddingEngines/native");
+const {
+  LLMPerformanceMonitor,
+} = require("../../helpers/chat/LLMPerformanceMonitor");
+const {
+  handleDefaultStreamResponseV2,
+} = require("../../helpers/chat/responses");
+const { MODEL_MAP } = require("../modelMap");
+
+class ZAiLLM {
+  constructor(embedder = null, modelPreference = null) {
+    if (!process.env.ZAI_API_KEY) throw new Error("No Z.AI API key was set.");
+    this.className = "ZAiLLM";
+    const { OpenAI: OpenAIApi } = require("openai");
+
+    this.openai = new OpenAIApi({
+      baseURL: "https://api.z.ai/api/paas/v4",
+      apiKey: process.env.ZAI_API_KEY,
+    });
+    this.model = modelPreference || process.env.ZAI_MODEL_PREF || "glm-4.5";
+    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(
+      `Initialized ${this.model} with context window ${this.promptWindowLimit()}`
+    );
+  }
+
+  log(text, ...args) {
+    console.log(`\x1b[36m[${this.className}]\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 ${i}]\n\n`;
+        })
+        .join("")
+    );
+  }
+
+  streamingEnabled() {
+    return "streamGetChatCompletion" in this;
+  }
+
+  static promptWindowLimit(modelName) {
+    return MODEL_MAP.get("zai", modelName) ?? 131072;
+  }
+
+  promptWindowLimit() {
+    return MODEL_MAP.get("zai", this.model) ?? 131072;
+  }
+
+  async isValidChatCompletionModel(modelName = "") {
+    return !!modelName; // name just needs to exist
+  }
+
+  /**
+   * Generates appropriate content array for a message + attachments.
+   * @param {{userPrompt:string, attachments: import("../../helpers").Attachment[]}}
+   * @returns {string|object[]}
+   */
+  #generateContent({ userPrompt, attachments = [] }) {
+    if (!attachments.length) return userPrompt;
+
+    const content = [{ type: "text", text: userPrompt }];
+    for (let attachment of attachments) {
+      content.push({
+        type: "image_url",
+        image_url: {
+          url: attachment.contentString,
+        },
+      });
+    }
+    return content.flat();
+  }
+
+  /**
+   * Construct the user prompt for this model.
+   * @param {{attachments: import("../../helpers").Attachment[]}} param0
+   * @returns
+   */
+  constructPrompt({
+    systemPrompt = "",
+    contextTexts = [],
+    chatHistory = [],
+    userPrompt = "",
+    attachments = [],
+  }) {
+    const prompt = {
+      role: "system",
+      content: `${systemPrompt}${this.#appendContext(contextTexts)}`,
+    };
+    return [
+      prompt,
+      ...chatHistory,
+      {
+        role: "user",
+        content: this.#generateContent({ userPrompt, attachments }),
+      },
+    ];
+  }
+
+  async getChatCompletion(messages = null, { temperature = 0.7 }) {
+    const result = await LLMPerformanceMonitor.measureAsyncFunction(
+      this.openai.chat.completions
+        .create({
+          model: this.model,
+          messages,
+          temperature,
+        })
+        .catch((e) => {
+          throw new Error(e.message);
+        })
+    );
+
+    if (
+      !result.output.hasOwnProperty("choices") ||
+      result.output.choices.length === 0
+    )
+      return null;
+
+    return {
+      textResponse: result.output.choices[0].message.content,
+      metrics: {
+        prompt_tokens: result.output.usage?.prompt_tokens || 0,
+        completion_tokens: result.output.usage?.completion_tokens || 0,
+        total_tokens: result.output.usage?.total_tokens || 0,
+        outputTps: result.output.usage?.completion_tokens / result.duration,
+        duration: result.duration,
+      },
+    };
+  }
+
+  async streamGetChatCompletion(messages = null, { temperature = 0.7 }) {
+    const measuredStreamRequest = await LLMPerformanceMonitor.measureStream(
+      this.openai.chat.completions.create({
+        model: this.model,
+        stream: true,
+        messages,
+        temperature,
+      }),
+      messages,
+      false
+    );
+
+    return measuredStreamRequest;
+  }
+
+  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);
+  }
+}
+
+module.exports = {
+  ZAiLLM,
+};

server/utils/agents/aibitat/index.js

@@ -964,6 +964,8 @@ ${this.getHistory({ to: route.to })
         return new Providers.ApiPieProvider({ model: config.model });
       case "xai":
         return new Providers.XAIProvider({ model: config.model });
+      case "zai":
+        return new Providers.ZAIProvider({ model: config.model });
       case "novita":
         return new Providers.NovitaProvider({ model: config.model });
       case "ppio":

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

@@ -183,6 +183,14 @@ class Provider {
           apiKey: process.env.XAI_LLM_API_KEY ?? null,
           ...config,
         });
+      case "zai":
+        return new ChatOpenAI({
+          configuration: {
+            baseURL: "https://api.z.ai/api/paas/v4",
+          },
+          apiKey: process.env.ZAI_API_KEY ?? null,
+          ...config,
+        });
       case "novita":
         return new ChatOpenAI({
           configuration: {

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

@@ -18,6 +18,7 @@ const DeepSeekProvider = require("./deepseek.js");
 const LiteLLMProvider = require("./litellm.js");
 const ApiPieProvider = require("./apipie.js");
 const XAIProvider = require("./xai.js");
+const ZAIProvider = require("./zai.js");
 const NovitaProvider = require("./novita.js");
 const NvidiaNimProvider = require("./nvidiaNim.js");
 const PPIOProvider = require("./ppio.js");
@@ -48,6 +49,7 @@ module.exports = {
   LiteLLMProvider,
   ApiPieProvider,
   XAIProvider,
+  ZAIProvider,
   NovitaProvider,
   CometApiProvider,
   NvidiaNimProvider,

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

@@ -0,0 +1,88 @@
+const OpenAI = require("openai");
+const Provider = require("./ai-provider.js");
+const InheritMultiple = require("./helpers/classes.js");
+const UnTooled = require("./helpers/untooled.js");
+
+class ZAIProvider extends InheritMultiple([Provider, UnTooled]) {
+  model;
+
+  constructor(config = {}) {
+    const { model = "glm-4.5" } = config;
+    super();
+    const client = new OpenAI({
+      baseURL: "https://api.z.ai/api/paas/v4",
+      apiKey: process.env.ZAI_API_KEY,
+      maxRetries: 3,
+    });
+
+    this._client = client;
+    this.model = model;
+    this.verbose = true;
+  }
+
+  /**
+   * Create a completion based on the received messages.
+   *
+   * @param messages A list of messages to send to the API.
+   * @param functions
+   * @returns The completion.
+   */
+  get client() {
+    return this._client;
+  }
+
+  get supportsAgentStreaming() {
+    return true;
+  }
+
+  async #handleFunctionCallChat({ messages = [] }) {
+    return await this.client.chat.completions
+      .create({
+        model: this.model,
+        messages,
+      })
+      .then((result) => {
+        if (!result.hasOwnProperty("choices"))
+          throw new Error("Z.AI chat: No results!");
+        if (result.choices.length === 0)
+          throw new Error("Z.AI chat: No results length!");
+        return result.choices[0].message.content;
+      })
+      .catch((_) => {
+        return null;
+      });
+  }
+
+  async #handleFunctionCallStream({ messages = [] }) {
+    return await this.client.chat.completions.create({
+      model: this.model,
+      stream: true,
+      messages,
+    });
+  }
+
+  async stream(messages, functions = [], eventHandler = null) {
+    return await UnTooled.prototype.stream.call(
+      this,
+      messages,
+      functions,
+      this.#handleFunctionCallStream.bind(this),
+      eventHandler
+    );
+  }
+
+  async complete(messages, functions = []) {
+    return await UnTooled.prototype.complete.call(
+      this,
+      messages,
+      functions,
+      this.#handleFunctionCallChat.bind(this)
+    );
+  }
+
+  getCost(_usage) {
+    return 0;
+  }
+}
+
+module.exports = ZAIProvider;

server/utils/agents/index.js

@@ -172,6 +172,10 @@ class AgentHandler {
         if (!process.env.XAI_LLM_API_KEY)
           throw new Error("xAI API Key must be provided to use agents.");
         break;
+      case "zai":
+        if (!process.env.ZAI_API_KEY)
+          throw new Error("Z.AI API Key must be provided to use agents.");
+        break;
       case "novita":
         if (!process.env.NOVITA_LLM_API_KEY)
           throw new Error("Novita API Key must be provided to use agents.");
@@ -275,6 +279,8 @@ class AgentHandler {
         return process.env.APIPIE_LLM_MODEL_PREF ?? null;
       case "xai":
         return process.env.XAI_LLM_MODEL_PREF ?? "grok-beta";
+      case "zai":
+        return process.env.ZAI_MODEL_PREF ?? "glm-4.5";
       case "novita":
         return process.env.NOVITA_LLM_MODEL_PREF ?? "deepseek/deepseek-r1";
       case "nvidia-nim":

server/utils/helpers/customModels.js

@@ -38,6 +38,7 @@ const SUPPORT_CUSTOM_MODELS = [
   "moonshotai",
   "foundry",
   "cohere",
+  "zai",
   // Embedding Engines
   "native-embedder",
   "cohere-embedder",
@@ -100,6 +101,8 @@ async function getCustomModels(provider = "", apiKey = null, basePath = null) {
       return await getFoundryModels(basePath);
     case "cohere":
       return await getCohereModels(apiKey, "chat");
+    case "zai":
+      return await getZAiModels(apiKey);
     case "native-embedder":
       return await getNativeEmbedderModels();
     case "cohere-embedder":
@@ -798,6 +801,29 @@ async function getCohereModels(_apiKey = null, type = "chat") {
   return { models, error: null };
 }
 
+async function getZAiModels(_apiKey = null) {
+  const { OpenAI: OpenAIApi } = require("openai");
+  const apiKey =
+    _apiKey === true
+      ? process.env.ZAI_API_KEY
+      : _apiKey || process.env.ZAI_API_KEY || null;
+  const openai = new OpenAIApi({
+    baseURL: "https://api.z.ai/api/paas/v4",
+    apiKey,
+  });
+  const models = await openai.models
+    .list()
+    .then((results) => results.data)
+    .catch((e) => {
+      console.error(`Z.AI:listModels`, e.message);
+      return [];
+    });
+
+  // Api Key was successful so lets save it for future uses
+  if (models.length > 0 && !!apiKey) process.env.ZAI_API_KEY = apiKey;
+  return { models, error: null };
+}
+
 module.exports = {
   getCustomModels,
   SUPPORT_CUSTOM_MODELS,

server/utils/helpers/index.js

@@ -222,6 +222,9 @@ function getLLMProvider({ provider = null, model = null } = {}) {
     case "foundry":
       const { FoundryLLM } = require("../AiProviders/foundry");
       return new FoundryLLM(embedder, model);
+    case "zai":
+      const { ZAiLLM } = require("../AiProviders/zai");
+      return new ZAiLLM(embedder, model);
     default:
       throw new Error(
         `ENV: No valid LLM_PROVIDER value found in environment! Using ${process.env.LLM_PROVIDER}`
@@ -378,6 +381,9 @@ function getLLMProviderClass({ provider = null } = {}) {
     case "foundry":
       const { FoundryLLM } = require("../AiProviders/foundry");
       return FoundryLLM;
+    case "zai":
+      const { ZAiLLM } = require("../AiProviders/zai");
+      return ZAiLLM;
     default:
       return null;
   }
@@ -450,6 +456,8 @@ function getBaseLLMProviderModel({ provider = null } = {}) {
       return process.env.COMETAPI_LLM_MODEL_PREF;
     case "foundry":
       return process.env.FOUNDRY_MODEL_PREF;
+    case "zai":
+      return process.env.ZAI_MODEL_PREF;
     default:
       return null;
   }

server/utils/helpers/updateENV.js

@@ -752,6 +752,16 @@ const KEY_MAPPING = {
     envKey: "COMETAPI_LLM_TIMEOUT_MS",
     checks: [],
   },
+
+  // Z.AI Options
+  ZAiApiKey: {
+    envKey: "ZAI_API_KEY",
+    checks: [isNotEmpty],
+  },
+  ZAiModelPref: {
+    envKey: "ZAI_MODEL_PREF",
+    checks: [isNotEmpty],
+  },
 };
 
 function isNotEmpty(input = "") {
@@ -863,6 +873,7 @@ function supportedLLM(input = "") {
     "moonshotai",
     "cometapi",
     "foundry",
+    "zai",
   ].includes(input);
   return validSelection ? null : `${input} is not a valid LLM provider.`;
 }