Error with Chroma.from_documents and OllamaEmbeddings: ResponseError: 404 page not found

[480284856]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

from langchain_chroma import Chroma
from langchain_ollama import OllamaEmbeddings

local_embeddings = OllamaEmbeddings(model="nomic-embed-text")

vectorstore = Chroma.from_documents(documents=all_splits, embedding=local_embeddings)

Description

I'm stuck at the code of vector store creation in the official document.

The output error information is:

{
	"name": "ResponseError",
	"message": "404 page not found",
	"stack": "---------------------------------------------------------------------------
ResponseError                             Traceback (most recent call last)
Cell In[2], line 6
      2 from langchain_ollama import OllamaEmbeddings
      4 local_embeddings = OllamaEmbeddings(model=\"nomic-embed-text\")
----> 6 vectorstore = Chroma.from_documents(documents=all_splits, embedding=local_embeddings)

File ~/miniconda3/lib/python3.10/site-packages/langchain_chroma/vectorstores.py:921, in Chroma.from_documents(cls, documents, embedding, ids, collection_name, persist_directory, client_settings, client, collection_metadata, **kwargs)
    919 texts = [doc.page_content for doc in documents]
    920 metadatas = [doc.metadata for doc in documents]
--> 921 return cls.from_texts(
    922     texts=texts,
    923     embedding=embedding,
    924     metadatas=metadatas,
    925     ids=ids,
    926     collection_name=collection_name,
    927     persist_directory=persist_directory,
    928     client_settings=client_settings,
    929     client=client,
    930     collection_metadata=collection_metadata,
    931     **kwargs,
    932 )

File ~/miniconda3/lib/python3.10/site-packages/langchain_chroma/vectorstores.py:882, in Chroma.from_texts(cls, texts, embedding, metadatas, ids, collection_name, persist_directory, client_settings, client, collection_metadata, **kwargs)
    876         chroma_collection.add_texts(
    877             texts=batch[3] if batch[3] else [],
    878             metadatas=batch[2] if batch[2] else None,  # type: ignore
    879             ids=batch[0],
    880         )
    881 else:
--> 882     chroma_collection.add_texts(texts=texts, metadatas=metadatas, ids=ids)
    883 return chroma_collection

File ~/miniconda3/lib/python3.10/site-packages/langchain_chroma/vectorstores.py:389, in Chroma.add_texts(self, texts, metadatas, ids, **kwargs)
    387 texts = list(texts)
    388 if self._embedding_function is not None:
--> 389     embeddings = self._embedding_function.embed_documents(texts)
    390 if metadatas:
    391     # fill metadatas with empty dicts if somebody
    392     # did not specify metadata for all texts
    393     length_diff = len(texts) - len(metadatas)

File ~/miniconda3/lib/python3.10/site-packages/langchain_ollama/embeddings.py:31, in OllamaEmbeddings.embed_documents(self, texts)
     29 def embed_documents(self, texts: List[str]) -> List[List[float]]:
     30     \"\"\"Embed search docs.\"\"\"
---> 31     embedded_docs = ollama.embed(self.model, texts)[\"embedding\"]
     32     return embedded_docs

File ~/miniconda3/lib/python3.10/site-packages/ollama/_client.py:261, in Client.embed(self, model, input, truncate, options, keep_alive)
    258 if not model:
    259   raise RequestError('must provide a model')
--> 261 return self._request(
    262   'POST',
    263   '/api/embed',
    264   json={
    265     'model': model,
    266     'input': input,
    267     'truncate': truncate,
    268     'options': options or {},
    269     'keep_alive': keep_alive,
    270   },
    271 ).json()

File ~/miniconda3/lib/python3.10/site-packages/ollama/_client.py:74, in Client._request(self, method, url, **kwargs)
     72   response.raise_for_status()
     73 except httpx.HTTPStatusError as e:
---> 74   raise ResponseError(e.response.text, e.response.status_code) from None
     76 return response

ResponseError: 404 page not found"
}

It seems that ollama with version of 0.3.1 can't work well with ollama.embed:

Code for reproducing:

import ollama
ollama.embed(
    model="nomic-embed-text",
    input=["hello"]
)

and the output is:

{
	"name": "ResponseError",
	"message": "404 page not found",
	"stack": "---------------------------------------------------------------------------
ResponseError                             Traceback (most recent call last)
Cell In[3], line 2
      1 import ollama
----> 2 ollama.embed(
      3     model=\"nomic-embed-text\",
      4     input=[\"hello\"]
      5 )

File ~/miniconda3/lib/python3.10/site-packages/ollama/_client.py:261, in Client.embed(self, model, input, truncate, options, keep_alive)
    258 if not model:
    259   raise RequestError('must provide a model')
--> 261 return self._request(
    262   'POST',
    263   '/api/embed',
    264   json={
    265     'model': model,
    266     'input': input,
    267     'truncate': truncate,
    268     'options': options or {},
    269     'keep_alive': keep_alive,
    270   },
    271 ).json()

File ~/miniconda3/lib/python3.10/site-packages/ollama/_client.py:74, in Client._request(self, method, url, **kwargs)
     72   response.raise_for_status()
     73 except httpx.HTTPStatusError as e:
---> 74   raise ResponseError(e.response.text, e.response.status_code) from None
     76 return response

ResponseError: 404 page not found"
}

It seems similar with the error before.

We can also experiment with the code:

curl http://localhost:11434/api/embed -d '{
  "model": "nomic-embed-text:latest",
  "prompt": "Here is an article about llamas..."
}'

and then the output is 404 page not found.

System Info

Version of packages:

langchain-ollama                         0.1.1
ollama                                   0.3.1

Extra info：The version of ollama in linux system is 0.1.45.

[dosubot[bot]]

Hello, @480284856! I'm here to assist you with any bugs, questions, or contributions. Let's tackle this together while waiting for a human maintainer.

The ResponseError: 404 page not found error occurs because the base URL http://localhost:11434 specified in the OllamaEmbeddings class is not reachable or the endpoint /api/embeddings does not exist on the server.

To resolve this issue, ensure that the server hosting the Ollama model is running and accessible at http://localhost:11434, and that the endpoint /api/embeddings is correctly set up to handle the requests.

You can verify if the Ollama model server is running and accessible by using the following curl command:

curl http://localhost:11434/api/generate -d '{
  "model": "llama3",
  "prompt": "Why is the sky blue?"
}'

If the server is running and accessible, you should receive a response from the server. Additionally, you can run ollama list to view all pulled models on your local instance and ollama run <name-of-model> to start interacting with a model via the command line directly. All of your local models are automatically served on localhost:11434 ^[1].

If the server is not running, you can start it with the following command:

OLLAMA_NUM_PARALLEL=4 OLLAMA_MAX_LOADED_MODELS=4 ollama serve

Make sure to replace nomic-embed-text with the correct model name if necessary. If the issue persists, double-check the endpoint and model name being used in your code ^[2][3].

To continue talking to Dosu, mention @dosu.

[480284856]

solved by using the latest ollama image.