microsoft/semantic-kernel · error · ServiceResponseException

{type(self)} service failed to complete the embedding reques

Error message

{type(self)} service failed to complete the embedding request.

What it means

Raised as ServiceResponseException (chaining `ex`) when `self.async_client.embeddings.create_async(model=..., inputs=texts)` throws any exception during `generate_raw_embeddings`. Like the chat variant it is a catch-all; the actual HTTP/SDK error is in the cause. Unlike initialization errors, this means the client built fine and the request was attempted.

Source

Thrown at python/semantic_kernel/connectors/ai/mistral_ai/services/mistral_ai_text_embedding.py:102

        texts: list[str],
        settings: "PromptExecutionSettings | None" = None,
        **kwargs: Any,
    ) -> ndarray:
        embedding_response = await self.generate_raw_embeddings(texts, settings, **kwargs)
        return array(embedding_response)

    @override
    async def generate_raw_embeddings(
        self,
        texts: list[str],
        settings: "PromptExecutionSettings | None" = None,
        **kwargs: Any,
    ) -> Any:
        """Generate embeddings from the Mistral AI service."""
        try:
            embedding_response = await self.async_client.embeddings.create_async(model=self.ai_model_id, inputs=texts)
        except Exception as ex:
            raise ServiceResponseException(
                f"{type(self)} service failed to complete the embedding request.",
                ex,
            ) from ex
        if isinstance(embedding_response, EmbeddingResponse):
            return [item.embedding for item in embedding_response.data]
        return []

View on GitHub (pinned to c028a0c7dc)

Solutions

  1. Inspect `e.__cause__` for the upstream status/message.
  2. Ensure ai_model_id is an embedding model (e.g. 'mistral-embed'), not a chat model.
  3. Filter empty/whitespace strings from `texts` before calling; chunk very long inputs.
  4. Retry with backoff on 429/timeout.

Example fix

# before
vecs = await svc.generate_embeddings(texts)

# after
texts = [t for t in texts if t and t.strip()]
try:
    vecs = await svc.generate_embeddings(texts)
except ServiceResponseException as e:
    raise RuntimeError(f"Mistral embedding upstream error: {e.__cause__!r}") from e
Defensive patterns

Strategy: retry

Validate before calling

texts = [t for t in texts if t and t.strip()]
assert texts, 'no non-empty texts to embed'
assert svc.ai_model_id and 'embed' in svc.ai_model_id.lower(), 'use an embedding model id'

Type guard

from semantic_kernel.exceptions import ServiceResponseException

def is_mistral_embed_error(e: BaseException) -> bool:
    return isinstance(e, ServiceResponseException) and 'failed to complete the embedding request' in str(e)

Try / catch

import asyncio
from semantic_kernel.exceptions import ServiceResponseException
async def embed_with_retry(svc, texts, attempts=3):
    last = None
    for i in range(attempts):
        try:
            return await svc.generate_embeddings(texts)
        except ServiceResponseException as e:
            last = e
            if e.__cause__ and getattr(e.__cause__, 'status_code', None) == 429:
                await asyncio.sleep(2 ** i)
                continue
            raise
    raise last

Prevention

When it happens

Trigger: Calling `generate_embeddings`/`generate_raw_embeddings` on MistralAITextEmbedding where the Mistral embeddings endpoint returns an error: 401 auth, 404 unknown embedding model, 422 bad inputs (empty list, too-long text), 429 rate limit, or network/timeout.

Common situations: ai_model_id points to a chat model instead of an embedding model (e.g. 'mistral-large-latest' passed to embeddings), inputs contain empty strings or exceed Mistral's token cap, transient rate limits during batch embedding.

Related errors

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/21572aef65ceb75f. Report an issue: GitHub.