Mintplex-Labs/anything-llm · critical · Error

No Gemini API key was set.

Error message

No Gemini API key was set.

What it means

Thrown by the GeminiLLM constructor when process.env.GEMINI_API_KEY is unset. Gemini is reached through Google's OpenAI-compatible endpoint (v1beta/openai/), which still requires the Gemini API key as the bearer token. Without it the OpenAI client cannot authenticate and the provider refuses to construct.

Source

Thrown at server/utils/AiProviders/gemini/index.js:30

const { defaultGeminiModels, v1BetaModels } = require("./defaultModels");
const { safeJsonParse } = require("../../http");
const cacheFolder = path.resolve(
  process.env.STORAGE_DIR
    ? path.resolve(process.env.STORAGE_DIR, "models", "gemini")
    : path.resolve(__dirname, `../../../storage/models/gemini`)
);

const NO_SYSTEM_PROMPT_MODELS = [
  "gemma-3-1b-it",
  "gemma-3-4b-it",
  "gemma-3-12b-it",
  "gemma-3-27b-it",
];

class GeminiLLM {
  constructor(embedder = null, modelPreference = null) {
    if (!process.env.GEMINI_API_KEY)
      throw new Error("No Gemini API key was set.");

    this.className = "GeminiLLM";
    const { OpenAI: OpenAIApi } = require("openai");
    this.model =
      modelPreference ||
      process.env.GEMINI_LLM_MODEL_PREF ||
      "gemini-2.0-flash-lite";

    const isExperimental = this.isExperimentalModel(this.model);
    this.openai = new OpenAIApi({
      apiKey: process.env.GEMINI_API_KEY,
      // Even models that are v1 in gemini API can be used with v1beta/openai/ endpoint and nobody knows why.
      baseURL: "https://generativelanguage.googleapis.com/v1beta/openai/",
    });

    this.limits = {
      history: this.promptWindowLimit() * 0.15,
      system: this.promptWindowLimit() * 0.15,

View on GitHub (pinned to 526360e320)

Solutions

1. Create a key at Google AI Studio (aistudio.google.com) and put GEMINI_API_KEY=... in .env
2. Restart AnythingLLM so process.env picks up the new key
3. Confirm the key works with a direct curl to the Gemini OpenAI endpoint before configuring AnythingLLM
4. If running in Docker, pass -e GEMINI_API_KEY or use an env file

Example fix

// before
// .env: (no GEMINI vars beyond maybe a model pref)

// after
// .env:
//   GEMINI_API_KEY=AIzaSy...
//   GEMINI_LLM_MODEL_PREF=gemini-2.0-flash-lite

function hasGeminiKey() {
  return Boolean(process.env.GEMINI_API_KEY);
}
if (!hasGeminiKey()) {
  throw new ConfigError('Create a key in Google AI Studio and set GEMINI_API_KEY.');
}

When it happens

Trigger: Constructing `new GeminiLLM(embedder, modelPreference)` while GEMINI_API_KEY is absent from process.env — fires at provider selection when the user chose Gemini but supplied no key.

Common situations: .env missing the GEMINI_API_KEY line; key revoked or never created in Google AI Studio; key stored in a different env source not loaded by AnythingLLM; Docker deployment without the env passthrough.

Related errors

• No Gitee AI API key was set.
• No Groq API key was set.
• No Foundry Base Path was set.
• GenericOpenAI must have a valid base path to use for the api
• KoboldCPP must have a valid base path to use for the api.

AI-assisted analysis of Mintplex-Labs/anything-llm@526360e320 (2026-08-13). Data as JSON: /api/errors/01cacab46a8b5d79. Report an issue: GitHub.