github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #2416] /v1/embeddings OpenAI compatible API endpoint #47920

New Issue
Closed
opened 2026-04-28 05:52:54 -05:00 by GiteaMirror · 15 comments
GiteaMirror commented 2026-04-28 05:52:54 -05:00
Owner
Copy Link

Originally created by @pamelafox on GitHub (Feb 8, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/2416

Originally assigned to: @bmizerany on GitHub.

Your blog post mentions you're considering it. We'd love it so that we can point our RAG apps at ollama. Thanks!

Originally created by @pamelafox on GitHub (Feb 8, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/2416 Originally assigned to: @bmizerany on GitHub. Your blog post mentions you're considering it. We'd love it so that we can point our RAG apps at ollama. Thanks!
GiteaMirror added the embeddingscompatibilityfeature request labels 2026-04-28 05:52:55 -05:00
GiteaMirror commented 2026-04-28 05:53:04 -05:00
Author
Owner
Copy Link

@MuhammadHadiofficial commented on GitHub (Feb 9, 2024):

@pamelafox Its there now -> https://ollama.com/blog/openai-compatibility

<!-- gh-comment-id:1936741718 --> @MuhammadHadiofficial commented on GitHub (Feb 9, 2024): @pamelafox Its there now -> https://ollama.com/blog/openai-compatibility
GiteaMirror commented 2026-04-28 05:53:06 -05:00
Author
Owner
Copy Link

@pamelafox commented on GitHub (Feb 10, 2024):

It specifically says that embeddings API is not yet supported on that page (at the bottom).

<!-- gh-comment-id:1936755446 --> @pamelafox commented on GitHub (Feb 10, 2024): It specifically says that embeddings API is not yet supported on that page (at the bottom).
GiteaMirror commented 2026-04-28 05:53:09 -05:00
Author
Owner
Copy Link

@glorat commented on GitHub (Feb 12, 2024):

One concrete point of incompatibility is the inability to pass in a vector of embedding requests. If one looks at the ollama source code, EmbeddingRequest clearly only takes in a single item rather than an array.

Would appreciate if maintainers would triage to bug / won't fix / feature request so I can adjust expectations accordingly. (Maintainers, you guys do a fantastic job with this project, any response is acceptable).

<!-- gh-comment-id:1938767906 --> @glorat commented on GitHub (Feb 12, 2024): One concrete point of incompatibility is the inability to pass in a vector of embedding requests. If one looks at the ollama source code, EmbeddingRequest clearly only takes in a single item rather than an array. Would appreciate if maintainers would triage to bug / won't fix / feature request so I can adjust expectations accordingly. (Maintainers, you guys do a fantastic job with this project, any response is acceptable).
GiteaMirror commented 2026-04-28 05:53:12 -05:00
Author
Owner
Copy Link

@odrobnik commented on GitHub (Apr 19, 2024):

So, why is the Ollama endpoint at /api/embeddings and not at /v1/embeddings where it should be to be compatible with OpenAI or for example LM Studio?

<!-- gh-comment-id:2067082568 --> @odrobnik commented on GitHub (Apr 19, 2024): So, why is the Ollama endpoint at `/api/embeddings` and not at `/v1/embeddings` where it should be to be compatible with OpenAI or for example LM Studio?
GiteaMirror commented 2026-04-28 05:53:13 -05:00
Author
Owner
Copy Link

@john-c-kane commented on GitHub (Apr 29, 2024):

I found this issue because i was trying to use Ollama Embeddings API for the Microsoft Semantic Kernel Memory functionality using the OPENAI provider with Ollama URL but I discovered the application is sending JSON format to API as "model" and "input" but Ollama embeddings api expects "model" and "prompt". So not sure if this is being considered as part of this OPENAI interface compatibility request but wanted to make you aware.

<!-- gh-comment-id:2082426079 --> @john-c-kane commented on GitHub (Apr 29, 2024): I found this issue because i was trying to use Ollama Embeddings API for the Microsoft Semantic Kernel Memory functionality using the OPENAI provider with Ollama URL but I discovered the application is sending JSON format to API as "model" and "input" but Ollama embeddings api expects "model" and "prompt". So not sure if this is being considered as part of this OPENAI interface compatibility request but wanted to make you aware.
GiteaMirror commented 2026-04-28 05:53:14 -05:00
Author
Owner
Copy Link

@odrobnik commented on GitHub (Apr 29, 2024):

There is no v1 embeddings endpoint yet, only the standard one at api

<!-- gh-comment-id:2082555885 --> @odrobnik commented on GitHub (Apr 29, 2024): There is no `v1` embeddings endpoint yet, only the standard one at `api`
GiteaMirror commented 2026-04-28 05:53:15 -05:00
Author
Owner
Copy Link

@yuriwa commented on GitHub (May 1, 2024):

I've just noticed the pull request (2476) and wanted to express my appreciation for the work put into it. Thank you for your contribution! Looking forward to its potential merge.

<!-- gh-comment-id:2089116127 --> @yuriwa commented on GitHub (May 1, 2024): I've just noticed the pull request [(2476)](https://github.com/ollama/ollama/pull/2476) and wanted to express my appreciation for the work put into it. Thank you for your contribution! Looking forward to its potential merge.
GiteaMirror commented 2026-04-28 05:53:16 -05:00
Author
Owner
Copy Link

@jmatsushita commented on GitHub (May 2, 2024):

There seems to be an active PR for this https://github.com/ollama/ollama/pull/2925 by @tazarov I haven't tested it yet.

<!-- gh-comment-id:2091260166 --> @jmatsushita commented on GitHub (May 2, 2024): There seems to be an active PR for this https://github.com/ollama/ollama/pull/2925 by @tazarov I haven't tested it yet.
GiteaMirror commented 2026-04-28 05:53:17 -05:00
Author
Owner
Copy Link

@JpEncausse commented on GitHub (May 26, 2024):

Very curious of availability and capabilities.

My usecase is straight forward:

  • During BUILD I need to create Embeddings of a lot of content locally in order to give a try, reduce cost, and bootstrap a database.
  • During RUN I'll calling online APIs when it's needed.

When Ollama will be compatible with text-embedding-3-large, text-embedding-3-small and text-embedding-ada-002 it will open the door to run embedding both locally (with a good PC for Free) and Online

<!-- gh-comment-id:2132231369 --> @JpEncausse commented on GitHub (May 26, 2024): Very curious of availability and capabilities. My usecase is straight forward: - During BUILD I need to create Embeddings of a lot of content locally in order to give a try, reduce cost, and bootstrap a database. - During RUN I'll calling online APIs when it's needed. When Ollama will be compatible with `text-embedding-3-large`, `text-embedding-3-small` and `text-embedding-ada-002` it will open the door to run embedding both locally (with a good PC for Free) and Online
GiteaMirror commented 2026-04-28 05:53:17 -05:00
Author
Owner
Copy Link

@odrobnik commented on GitHub (May 26, 2024):

@JpEncausse the Models you mention as OpenAI proprietary AFAIK. So you will never see them on Ollama.

<!-- gh-comment-id:2132235092 --> @odrobnik commented on GitHub (May 26, 2024): @JpEncausse the Models you mention as OpenAI proprietary AFAIK. So you will never see them on Ollama.
GiteaMirror commented 2026-04-28 05:53:18 -05:00
Author
Owner
Copy Link

@cristoslc commented on GitHub (Jun 6, 2024):

I think I see two PRs related to this (https://github.com/ollama/ollama/pull/2925 and https://github.com/ollama/ollama/pull/3642). As others have said, the fact that the api/embeddings endpoint doesn't accept an array of inputs AND the difference in the request structure vs. OpenAI's structure (per https://github.com/ollama/ollama/issues/2416#issuecomment-2082426079) are both major blocks to using Ollama in a variety of RAG applications. Any word on where those PRs are in priority?

Thank you all for your hard work on the project!

<!-- gh-comment-id:2151357086 --> @cristoslc commented on GitHub (Jun 6, 2024): I think I see two PRs related to this (https://github.com/ollama/ollama/pull/2925 and https://github.com/ollama/ollama/pull/3642). As others have said, the fact that the `api/embeddings` endpoint doesn't accept an array of inputs AND the difference in the request structure vs. OpenAI's structure (per https://github.com/ollama/ollama/issues/2416#issuecomment-2082426079) are both major blocks to using Ollama in a variety of RAG applications. Any word on where those PRs are in priority? Thank you all for your hard work on the project!
GiteaMirror commented 2026-04-28 05:53:19 -05:00
Author
Owner
Copy Link

@aletfa commented on GitHub (Jun 26, 2024):

+1 here. Please implement the complete compatibility

<!-- gh-comment-id:2192228973 --> @aletfa commented on GitHub (Jun 26, 2024): +1 here. Please implement the complete compatibility
GiteaMirror commented 2026-04-28 05:53:21 -05:00
Author
Owner
Copy Link

@JPMoresmau commented on GitHub (Jul 18, 2024):

I'm trying 0.2.6 that has the PR in it that is bringing the compatibility API, and I'm getting a 400 error when /v1/embeddings is called from the python openai client : {'error': {'message': 'invalid input type', 'type': 'api_error', 'param': None, 'code': None}}", "taskName": "Task-3"}. Not sure how to debug this further.

<!-- gh-comment-id:2236938126 --> @JPMoresmau commented on GitHub (Jul 18, 2024): I'm trying 0.2.6 that has the PR in it that is bringing the compatibility API, and I'm getting a 400 error when /v1/embeddings is called from the python openai client : ` {'error': {'message': 'invalid input type', 'type': 'api_error', 'param': None, 'code': None}}", "taskName": "Task-3"}`. Not sure how to debug this further.
GiteaMirror commented 2026-04-28 05:53:21 -05:00
Author
Owner
Copy Link

@arch7tect commented on GitHub (Jul 29, 2024):

Here is a temporary solution while v1/embeddings in ollama is broken
https://github.com/severian42/GraphRAG-Local-UI/blob/main/embedding_proxy.py

<!-- gh-comment-id:2255001969 --> @arch7tect commented on GitHub (Jul 29, 2024): Here is a temporary solution while v1/embeddings in ollama is broken https://github.com/severian42/GraphRAG-Local-UI/blob/main/embedding_proxy.py
GiteaMirror commented 2026-04-28 05:53:21 -05:00
Author
Owner
Copy Link

@royjhan commented on GitHub (Jul 29, 2024):

@JPMoresmau how are you calling v1/embeddings? I'm unable to reproduce the error, this code is working as expected

from openai import OpenAI
 
client = OpenAI(
    base_url='http://localhost:11434/v1/',
    api_key='ollama'
)

response = client.embeddings.create(
  model="all-minilm",
  input=["input1", "input2"], // "input1" also works, no other input type is supported
)

print(response)
<!-- gh-comment-id:2256807748 --> @royjhan commented on GitHub (Jul 29, 2024): @JPMoresmau how are you calling v1/embeddings? I'm unable to reproduce the error, this code is working as expected ``` from openai import OpenAI client = OpenAI( base_url='http://localhost:11434/v1/', api_key='ollama' ) response = client.embeddings.create( model="all-minilm", input=["input1", "input2"], // "input1" also works, no other input type is supported ) print(response) ```
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label compatibility embeddings feature request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#47920
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?