Rag_api usage gives 'Tokenized input is not supported'

[TJKlein]

Hi,
I am trying to use rag_api for document processing using some embedder which conforms the openai API.

EMBEDDINGS_PROVIDER=openai
RAG_OPENAI_BASEURL=SERVER_URL/v1
RAG_OPENAI_API_KEY=API_KEY
EMBEDDINGS_MODEL=intfloat/e5-mistral-7b-instruct

Yet curl works:
curl '$SERVER_URL/v1/embeddings' \ -X 'POST' \ -H 'Content-Type: application/json' \ -H 'Accept: */*' \ -H "Authorization: Bearer $API_KEY" \ --data-binary '{"model":"intfloat/e5-mistral-7b-instruct","input":"This is a test."}'

But then I keep getting this error:

rag_api | /usr/local/lib/python3.10/site-packages/pypdf/_crypt_providers/_cryptography.py:32: CryptographyDeprecationWarning: ARC4 has been moved to cryptography.hazmat.decrepit.ciphers.algorithms.ARC4 and will be removed from cryptography.hazmat.primitives.ciphers.algorithms in 48.0.0.
rag_api | from cryptography.hazmat.primitives.ciphers.algorithms import AES, ARC4
rag_api | 2025-06-25 17:29:24,550 - httpx - INFO - HTTP Request: POST SERVER_URLv1/embeddings "HTTP/1.1 400 Bad Request"
rag_api | 2025-06-25 17:29:24,562 - root - ERROR - Failed to store data in vector DB | File ID: 1d743c98-1726-469b-9cd0-9d8d94a7de6c | User ID: 685993b1f2ab18b6cbdd9e43 | Error: Error code: 400 - {'detail': 'Tokenized input is not supported'} | Traceback: Traceback (most recent call last):
rag_api | File "/app/app/routes/document_routes.py", line 276, in store_data_in_vector_db
rag_api | ids = await vector_store.aadd_documents(
rag_api | File "/usr/local/lib/python3.10/site-packages/langchain_core/vectorstores/base.py", line 324, in aadd_documents
rag_api | return await run_in_executor(None, self.add_documents, documents, **kwargs)
LibreChat | 2025-06-25 17:29:24 error: Error uploading vectors An error occurred while setting up the request: File embedding failed.
rag_api | File "/usr/local/lib/python3.10/site-packages/langchain_core/runnables/config.py", line 588, in run_in_executor
LibreChat | 2025-06-25 17:29:24 error: [/files] Error processing file: File embedding failed.
rag_api | return await asyncio.get_running_loop().run_in_executor(
rag_api | File "/usr/local/lib/python3.10/concurrent/futures/thread.py", line 58, in run
rag_api | result = self.fn(*self.args, **self.kwargs)
rag_api | File "/usr/local/lib/python3.10/site-packages/langchain_core/runnables/config.py", line 579, in wrapper
rag_api | return func(*args, **kwargs)
rag_api | File "/usr/local/lib/python3.10/site-packages/langchain_core/vectorstores/base.py", line 287, in add_documents
rag_api | return self.add_texts(texts, metadatas, **kwargs)
rag_api | File "/usr/local/lib/python3.10/site-packages/langchain_community/vectorstores/pgvector.py", line 561, in add_texts
rag_api | embeddings = self.embedding_function.embed_documents(list(texts))
rag_api | File "/usr/local/lib/python3.10/site-packages/langchain_openai/embeddings/base.py", line 588, in embed_documents
rag_api | return self._get_len_safe_embeddings(texts, engine=engine)
rag_api | File "/usr/local/lib/python3.10/site-packages/langchain_openai/embeddings/base.py", line 483, in _get_len_safe_embeddings
rag_api | response = self.client.create(
rag_api | File "/usr/local/lib/python3.10/site-packages/openai/resources/embeddings.py", line 129, in create
rag_api | return self._post(
rag_api | File "/usr/local/lib/python3.10/site-packages/openai/_base_client.py", line 1249, in post
rag_api | return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))
rag_api | File "/usr/local/lib/python3.10/site-packages/openai/_base_client.py", line 1037, in request
rag_api | raise self._make_status_error_from_response(err.response) from None
rag_api | openai.BadRequestError: Error code: 400 - {'detail': 'Tokenized input is not supported'}
rag_api |
rag_api | 2025-06-25 17:29:24,572 - root - INFO - Request POST http://rag_api:8000/embed - 200

[TJKlein]

I disabled tiktoken for the OpenAIEmbeddings a bit similar to: #3046
Then it worked.