Mintplex-Labs/anything-llm · critical · Error

No embedding model was set.

Error message

No embedding model was set.

What it means

Thrown by the LocalAiEmbedder constructor (line 12) when process.env.EMBEDDING_MODEL_PREF is falsy, right after the base-path guard. The model name is sent as the `model` field on every /embeddings request; LocalAI needs it to pick the loaded model. The guard prevents queuing requests that LocalAI would reject.

Source

Thrown at server/utils/EmbeddingEngines/localAi/index.js:12

const {
  toChunks,
  maximumChunkLength,
  reportEmbeddingProgress,
} = require("../../helpers");

class LocalAiEmbedder {
  constructor() {
    if (!process.env.EMBEDDING_BASE_PATH)
      throw new Error("No embedding base path was set.");
    if (!process.env.EMBEDDING_MODEL_PREF)
      throw new Error("No embedding model was set.");

    this.className = "LocalAiEmbedder";
    const { OpenAI: OpenAIApi } = require("openai");
    this.model = process.env.EMBEDDING_MODEL_PREF;
    this.openai = new OpenAIApi({
      baseURL: process.env.EMBEDDING_BASE_PATH,
      apiKey: process.env.LOCAL_AI_API_KEY ?? null,
    });

    // Limit of how many strings we can process in a single pass to stay with resource or network limits
    this.maxConcurrentChunks = 50;
    this.embeddingMaxChunkLength = maximumChunkLength();

    this.log(
      `Initialized with ${this.model} - Max Size: ${this.embeddingMaxChunkLength}` +
        (this.outputDimensions
          ? ` - Output Dimensions: ${this.outputDimensions}`
          : " Assuming default output dimensions")

View on GitHub (pinned to 526360e320)

Solutions

1. Set EMBEDDING_MODEL_PREF to the model id exactly as configured in LocalAI's models.yaml
2. Confirm the model is actually loaded/available in LocalAI (check its /v1/models endpoint)
3. Reload AnythingLLM env

Example fix

// before
EMBEDDING_BASE_PATH=http://localhost:8080/v1
// EMBEDDING_MODEL_PREF unset

// after
EMBEDDING_BASE_PATH=http://localhost:8080/v1
EMBEDDING_MODEL_PREF=bge-small-en

function hasEmbeddingModelPref(env = process.env) {
  return typeof env.EMBEDDING_MODEL_PREF === 'string' &&
         env.EMBEDDING_MODEL_PREF.trim().length > 0;
}
if (!hasEmbeddingModelPref()) {
  throw new Error('Missing EMBEDDING_MODEL_PREF for LocalAI.');
}

function isNonEmptyModel(v) {
  return typeof v === 'string' && v.trim().length > 0;
}

When it happens

Trigger: Constructing `new LocalAiEmbedder()` with EMBEDDING_MODEL_PREF unset/empty. Fires at construction time before any embedding work, mirroring the LMStudio model guard.

Common situations: LocalAI model installed but EMBEDDING_MODEL_PREF not set to match; model name typo; switching engines cleared the previous model env; models.yaml in LocalAI uses a different id than what was typed.

Related errors

• No embedding base path was set.
• LiteLLM must have a valid base path to use for the api.
• No embedding base path was set.
• No embedding model was set.
• No Mistral API key was set.

AI-assisted analysis of Mintplex-Labs/anything-llm@526360e320 (2026-08-13). Data as JSON: /api/errors/8141341434a9b824. Report an issue: GitHub.