HuggingFaceAPIDocumentEmbedder Bad Request Error: list indices must be integers or slices, not str

[hodge-jai]

Hello,

I am encountering an issue when trying to use a dedicated inference endpoint on HuggingFace. I'm not sure where the error is, as I triple checked that I am indeed passing a list of Documents to the component. The embedder initializes just fine, but when I run it, I receive a 400 Bad Request error.

document_embedder = HuggingFaceAPIDocumentEmbedder(
    api_type="inference_endpoints",
    api_params={"url": "<my_dedicated_endpoint"},
    token=Secret.from_token(HF_TOKEN),
    batch_size=32
)
documents = [Document(content="My name is Wolfgang and I live in Berlin"),
             Document(content="I saw a black horse running"),
             Document(content="Germany has many big cities")]
document_embedder.run(documents=documents)

Returns

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/ragstackapp/fast_api/IndexingPipeline.py", line 130, in <module>
    pool.map(index_html_files, chunks)
  File "/usr/lib/python3.10/multiprocessing/pool.py", line 367, in map
    return self._map_async(func, iterable, mapstar, chunksize).get()
  File "/usr/lib/python3.10/multiprocessing/pool.py", line 774, in get
    raise self._value
huggingface_hub.utils._errors.BadRequestError:  (Request ID: HOp5U9)

Bad request:
list indices must be integers or slices, not str

Not sure if this is an issue with Haystack or HuggingFace, so I would appreciate any guidance!

[julian-risch]

Hi @hodge-jai I tried the code example from our documentation page here: https://docs.haystack.deepset.ai/docs/huggingfaceapidocumentembedder#using-free-serverless-inference-api and it worked for me

from haystack.components.embedders import HuggingFaceAPIDocumentEmbedder
from haystack.utils import Secret
from haystack.dataclasses import Document

doc = Document(content="I love pizza!")

document_embedder = HuggingFaceAPIDocumentEmbedder(api_type="serverless_inference_api",
                                              api_params={"model": "BAAI/bge-small-en-v1.5"},
                                              token=Secret.from_token("<your-api-key>"))

result = document_embedder.run([doc])
print(result["documents"][0].embedding)

# [0.017020374536514282, -0.023255806416273117, ...]

(just had to rename doc_embedder to document_embedder, for which I created an issue here ).
So the problem with the Bad Request Error you describe seems to be limited to Paid Inference Endpoints and api_type="inference_endpoints". I'll check whether we can investigate more in one of our next sprints. It could be a problem on HuggingFace side with https://github.com/huggingface/huggingface_hub

[hodge-jai]

I only realized my issue a bit later and it was purely user error 🤦

I had to configure the Task option properly in the HuggingFace paid inference endpoint setting panel to be in line with what I was passing to the endpoint. One doesn't need to change this setting when using api_type=serverless_inference_api

Thanks for your reply!