Mintplex-Labs/anything-llm · error · Error

No AZURE_OPENAI_MODEL_PREF ENV defined. This must the name o

Error message

No AZURE_OPENAI_MODEL_PREF ENV defined. This must the name of a deployment on your Azure account for an LLM chat model like GPT-3.5.

What it means

Thrown by AzureOpenAiLLM.getChatCompletion when `this.model` is falsy at call time. this.model is resolved in the constructor from modelPreference arg, else AZURE_OPENAI_MODEL_PREF, else OPEN_MODEL_PREF — if all are empty, this.model is undefined and the chat call is blocked because Azure needs an explicit deployment name (it cannot infer one).

Source

Thrown at server/utils/AiProviders/azureOpenAi/index.js:155

    attachments = [], // This is the specific attachment for only this prompt
  }) {
    const prompt = {
      role: this.isOTypeModel ? "user" : "system",
      content: `${systemPrompt}${this.#appendContext(contextTexts)}`,
    };
    return [
      prompt,
      ...formatChatHistory(chatHistory, this.#generateContent),
      {
        role: "user",
        content: this.#generateContent({ userPrompt, attachments }),
      },
    ];
  }

  async getChatCompletion(messages = [], { temperature = 0.7 }) {
    if (!this.model)
      throw new Error(
        "No AZURE_OPENAI_MODEL_PREF ENV defined. This must the name of a deployment on your Azure account for an LLM chat model like GPT-3.5."
      );

    const result = await LLMPerformanceMonitor.measureAsyncFunction(
      this.openai.chat.completions.create({
        messages,
        model: this.model,
        ...(this.isOTypeModel ? {} : { temperature }),
      })
    );

    if (
      !result.output.hasOwnProperty("choices") ||
      result.output.choices.length === 0
    )
      return null;

    return {

View on GitHub (pinned to 526360e320)

Solutions

1. Set `AZURE_OPENAI_MODEL_PREF=<deployment-name>` in .env to the name of a chat deployment in your Azure resource (not the base model name).
2. Pass modelPreference explicitly when constructing the provider for a per-workspace model.
3. If relying on the OPEN_MODEL_PREF fallback, ensure it is set to a valid Azure deployment name.
4. Restart the server so the constructor re-reads the env.

Example fix

// before
// .env
AZURE_OPENAI_ENDPOINT=https://my-resource.openai.azure.com
AZURE_OPENAI_KEY=...
// no model pref

// after
AZURE_OPENAI_ENDPOINT=https://my-resource.openai.azure.com
AZURE_OPENAI_KEY=...
AZURE_OPENAI_MODEL_PREF=gpt-35-turbo-deploy

// Resolve the effective model the same way the constructor does
const model =
  modelPref ||
  process.env.AZURE_OPENAI_MODEL_PREF ||
  process.env.OPEN_MODEL_PREF;
if (!model) {
  throw new Error(
    "No Azure deployment name configured. Set AZURE_OPENAI_MODEL_PREF to a chat deployment name."
  );
}
const llm = new AzureOpenAiLLM(embedder, model);

/** @param {unknown} m @returns {boolean} */
function hasChatModel(m) {
  return typeof m === "string" && m.trim().length > 0;
}

try {
  const res = await llm.getChatCompletion(messages, { temperature });
} catch (e) {
  if (/No AZURE_OPENAI_MODEL_PREF/.test(e.message)) {
    return { ok: false, reason: "missing-model", hint: "Create a deployment and set AZURE_OPENAI_MODEL_PREF." };
  }
  throw e;
}

When it happens

Trigger: Calling getChatCompletion after constructing the provider with no model preference and neither AZURE_OPENAI_MODEL_PREF nor OPEN_MODEL_PREF set in env. The constructor does not throw on missing model (only endpoint/key), so this surfaces lazily on first chat.

Common situations: Azure provider configured with endpoint+key but model/deployment field left blank; deployment created in Azure but its name not entered; env var typo (e.g. AZURE_OPENAI_MODEL instead of AZURE_OPENAI_MODEL_PREF); relying on OPEN_MODEL_PREF but that is also unset.

Related errors

• No Azure API endpoint was set.
• ApiPie chat: ${this.model} is not valid for chat completion!
• No Azure API key was set.
• "${azureOpenAiEndpoint}" is not a valid URL. Check your sett
• No Azure API endpoint was set.

AI-assisted analysis of Mintplex-Labs/anything-llm@526360e320 (2026-08-13). Data as JSON: /api/errors/b1c5fbeec8ec8422. Report an issue: GitHub.