Mintplex-Labs/anything-llm · error · Error

GenericOpenAI Failed to embed: ${error.message}

Error message

GenericOpenAI Failed to embed: ${error.message}

What it means

Thrown inside the embedChunks loop when a batch returns an error. Unlike the Azure/Cohere/Gemini embedders, this is a single per-batch error (error.message), not an aggregated set, and the loop aborts on the first batch that fails. The message is prefixed 'GenericOpenAI Failed to embed:'.

Source

Thrown at server/utils/EmbeddingEngines/genericOpenAi/index.js:156

          .create({
            model: this.model,
            input: chunk,
          })
          .then((result) => resolve({ data: result?.data, error: null }))
          .catch((e) => {
            e.type =
              e?.response?.data?.error?.code ||
              e?.response?.status ||
              "failed_to_embed";
            e.message = e?.response?.data?.error?.message || e.message;
            resolve({ data: [], error: e });
          });
      });

      // If any errors were returned from OpenAI abort the entire sequence because the embeddings
      // will be incomplete.
      if (error)
        throw new Error(`GenericOpenAI Failed to embed: ${error.message}`);
      allResults.push(...(data || []));
      reportEmbeddingProgress(allResults.length, textChunks.length);
      if (this.apiRequestDelay) await this.runDelay();
    }

    return allResults.length > 0 &&
      allResults.every((embd) => embd.hasOwnProperty("embedding"))
      ? allResults.map((embd) => embd.embedding)
      : null;
  }
}

module.exports = {
  GenericOpenAiEmbedder,
};

View on GitHub (pinned to 526360e320)

Solutions

1. Read error.message: 'model not found' -> pull/load the model and set EMBEDDING_MODEL_PREF correctly; 'unauthorized' -> set GENERIC_OPEN_AI_EMBEDDING_API_KEY; connection errors -> confirm the server is up at EMBEDDING_BASE_PATH.
2. GET <EMBEDDING_BASE_PATH>/models to confirm the model id is exposed before embedding.
3. Shorten chunks to fit the local model's context; reduce concurrency if the server is resource-limited.
4. Retry transient local-server errors after the server is healthy.

Example fix

// before
if (error) throw new Error(`GenericOpenAI Failed to embed: ${error.message}`);

// after (retry once on transient, then surface)
if (error) {
  if (isTransient(error)) { /* retry batch */ }
  else throw new Error(`GenericOpenAI Failed to embed: ${error.message}`);
}

// Pre-flight: confirm the upstream exposes the model before bulk embedding
const { OpenAI } = require('openai');
const client = new OpenAI({ baseURL: process.env.EMBEDDING_BASE_PATH, apiKey: process.env.GENERIC_OPEN_AI_EMBEDDING_API_KEY ?? null });
const list = await client.models.list();
if (!list.body.some((m) => m.id === process.env.EMBEDDING_MODEL_PREF)) {
  throw new Error(`Generic upstream does not expose model ${process.env.EMBEDDING_MODEL_PREF}`);
}

try {
  await embedder.embedChunks(chunks);
} catch (e) {
  const msg = e.message;
  if (/model.*not|404/i.test(msg)) loadModel();
  else if (/unauthorized|401/i.test(msg)) setApiKey();
  else if (/econnrefused|timeout|socket/i.test(msg)) waitForServer();
  else throw e;
}

When it happens

Trigger: Upstream at EMBEDDING_BASE_PATH returns non-200 (auth required but key unset/invalid, model not loaded, internal error); EMBEDDING_MODEL_PREF names a model the server does not expose; chunks exceed the server's max tokens; network/timeout to the local or remote generic endpoint.

Common situations: Local LLM server (Ollama/LM Studio/vLLM) not running or model not pulled; GENERIC_OPEN_AI_EMBEDDING_API_KEY required by the proxy but unset; mistyped model id; proxy rate limiting.

Related errors

• Lemonade Failed to embed: [${error.type}]: ${error.message}
• GenericOpenAI must have a valid base path to use for the api
• Could not find indexes.
• Failed to fetch documents by paths.
• e.message

AI-assisted analysis of Mintplex-Labs/anything-llm@526360e320 (2026-08-13). Data as JSON: /api/errors/360ff021173fe66d. Report an issue: GitHub.