github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 00:51:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #8395] Empty response via API #67447

New Issue
Closed
opened 2026-05-04 10:23:11 -05:00 by GiteaMirror · 12 comments
GiteaMirror commented 2026-05-04 10:23:11 -05:00
Owner
Copy Link

Originally created by @gl2007 on GitHub (Jan 12, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/8395

What is the issue?

Hosted ollama via 0.0.0.0 in my server in my LAN and "curl :11434 " returns ollama is running. Also, when I run ollama run in cmd in that machine, I am able to see proper responses.

However, when I run an API request via Postman, I get this empty response, irrespective of the model, which seems to indicate model is not loaded properly. This also happens on the server machine via Postman using localhost.

{
"model": "Mistral-Nemo-12B-Instruct-2407-Q8_0:latest",
"created_at": "2025-01-12T07:39:16.7356243Z",
"response": "",
"done": true,
"done_reason": "load"
}

But it seems model is loaded correctly as I see it in "ollama ps".

What am I doing wrong?

OS

Windows

GPU

None

CPU

Intel

Ollama version

0.5.4

Originally created by @gl2007 on GitHub (Jan 12, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/8395 ### What is the issue? Hosted ollama via 0.0.0.0 in my server in my LAN and "curl <ip>:11434 " returns ollama is running. Also, when I run ollama run<model> in cmd in that machine, I am able to see proper responses. However, when I run an API request via Postman, I get this empty response, irrespective of the model, which seems to indicate model is not loaded properly. This also happens on the server machine via Postman using localhost. { "model": "Mistral-Nemo-12B-Instruct-2407-Q8_0:latest", "created_at": "2025-01-12T07:39:16.7356243Z", "response": "", "done": true, "done_reason": "load" } But it seems model is loaded correctly as I see it in "ollama ps". What am I doing wrong? ### OS Windows ### GPU None ### CPU Intel ### Ollama version 0.5.4
GiteaMirror added the bug label 2026-05-04 10:23:11 -05:00
GiteaMirror commented 2026-05-04 10:23:18 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 12, 2025):

You need to send a prompt. This is the response when the model is loaded but hasn't been asked to do a completion: "done_reason": "load"

<!-- gh-comment-id:2585687010 --> @rick-github commented on GitHub (Jan 12, 2025): You need to send a prompt. This is the response when the model is loaded but hasn't been asked to do a completion: `"done_reason": "load"`
GiteaMirror commented 2026-05-04 10:23:21 -05:00
Author
Owner
Copy Link

@gl2007 commented on GitHub (Jan 12, 2025):

You need to send a prompt. This is the response when the model is loaded but hasn't been asked to do a completion: "done_reason": "load"

I did send a prompt:

{
"model": "Mistral-Nemo-12B-Instruct-2407-Q8_0:latest", //insert any models from Ollama that are on your local machine
"messages": [
{
"role": "system", //"system" is a prompt to define how the model should act.
"content": "you are a salty pirate" //system prompt should be written here
},
{
"role": "user", //"user" is a prompt provided by the user.
"content": "why is the sky blue" //user prompt should be written here
}
],
"stream": false //returns as a full message rather than a streamed response
}

<!-- gh-comment-id:2585906449 --> @gl2007 commented on GitHub (Jan 12, 2025): > You need to send a prompt. This is the response when the model is loaded but hasn't been asked to do a completion: `"done_reason": "load"` I did send a prompt: { "model": "Mistral-Nemo-12B-Instruct-2407-Q8_0:latest", //insert any models from Ollama that are on your local machine "messages": [ { "role": "system", //"system" is a prompt to define how the model should act. "content": "you are a salty pirate" //system prompt should be written here }, { "role": "user", //"user" is a prompt provided by the user. "content": "why is the sky blue" //user prompt should be written here } ], "stream": false //returns as a full message rather than a streamed response }
GiteaMirror commented 2026-05-04 10:23:22 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 12, 2025):

Which endpoint did you use?

<!-- gh-comment-id:2585933452 --> @rick-github commented on GitHub (Jan 12, 2025): Which endpoint did you use?
GiteaMirror commented 2026-05-04 10:23:25 -05:00
Author
Owner
Copy Link

@gl2007 commented on GitHub (Jan 12, 2025):

http://:11434/api/generate and using it via Postman, via POST. Using Postman documentation example prompts from here

<!-- gh-comment-id:2585942877 --> @gl2007 commented on GitHub (Jan 12, 2025): http://<my-server-ip>:11434/api/generate and using it via Postman, via POST. Using Postman documentation example prompts from [here](https://www.postman.com/postman-student-programs/ollama-api/documentation/suc47x8/ollama-rest-api)
GiteaMirror commented 2026-05-04 10:23:27 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 12, 2025):

For /api/generate the format is {"model":"Mistral-Nemo-12B-Instruct-2407-Q8_0:latest","prompt":"why is the sky blue"}. The format you are using above is for the /api/chat endpoint.

<!-- gh-comment-id:2585945663 --> @rick-github commented on GitHub (Jan 12, 2025): For `/api/generate` the format is `{"model":"Mistral-Nemo-12B-Instruct-2407-Q8_0:latest","prompt":"why is the sky blue"}`. The format you are using above is for the `/api/chat` endpoint.
GiteaMirror commented 2026-05-04 10:23:34 -05:00
Author
Owner
Copy Link

@gl2007 commented on GitHub (Jan 13, 2025):

For /api/generate the format is {"model":"Mistral-Nemo-12B-Instruct-2407-Q8_0:latest","prompt":"why is the sky blue"}. The format you are using above is for the /api/chat endpoint.

TY so much! Not sure how I missed that. Its working now!

However, now if that is the syntax, how do I tell it the system prompt and prompts for all the subsequent requests? Can't I have "session behavior"?

<!-- gh-comment-id:2586076794 --> @gl2007 commented on GitHub (Jan 13, 2025): > For `/api/generate` the format is `{"model":"Mistral-Nemo-12B-Instruct-2407-Q8_0:latest","prompt":"why is the sky blue"}`. The format you are using above is for the `/api/chat` endpoint. TY so much! Not sure how I missed that. Its working now! However, now if that is the syntax, how do I tell it the system prompt and prompts for all the subsequent requests? Can't I have "session behavior"?
GiteaMirror commented 2026-05-04 10:23:36 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 13, 2025):

The ollama server is stateless. If want to maintain conversation history, you accumulate the prompts and responses in the client and append them to messages[] when you send the request to the API.

<!-- gh-comment-id:2586078700 --> @rick-github commented on GitHub (Jan 13, 2025): The ollama server is stateless. If want to maintain conversation history, you accumulate the prompts and responses in the client and append them to `messages[]` when you send the request to the API.
GiteaMirror commented 2026-05-04 10:23:38 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Jan 13, 2025):

It's definitely harder to do with /api/generate. You probably want to start off with /api/chat initially.

I'll go ahead and close the issue, but feel free to keep commenting.

<!-- gh-comment-id:2588005776 --> @pdevine commented on GitHub (Jan 13, 2025): It's definitely harder to do with `/api/generate`. You probably want to start off with `/api/chat` initially. I'll go ahead and close the issue, but feel free to keep commenting.
GiteaMirror commented 2026-05-04 10:23:39 -05:00
Author
Owner
Copy Link

@gl2007 commented on GitHub (Jan 13, 2025):

It's definitely harder to do with /api/generate. You probably want to start off with /api/chat initially.

I'll go ahead and close the issue, but feel free to keep commenting.

Agreed. However, does that mean the chat endpoint maintains state? do I have to do anything for it to maintain state and if I pass in a new system prompt, does that start a new "session"?

<!-- gh-comment-id:2588059743 --> @gl2007 commented on GitHub (Jan 13, 2025): > It's definitely harder to do with `/api/generate`. You probably want to start off with `/api/chat` initially. > > I'll go ahead and close the issue, but feel free to keep commenting. Agreed. However, does that mean the chat endpoint maintains state? do I have to do anything for it to maintain state and if I pass in a new system prompt, does that start a new "session"?
GiteaMirror commented 2026-05-04 10:23:40 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 13, 2025):

No, the ollama server is stateless.

<!-- gh-comment-id:2588411284 --> @rick-github commented on GitHub (Jan 13, 2025): No, the ollama server is stateless.
GiteaMirror commented 2026-05-04 10:23:41 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Jan 14, 2025):

Just keep adding each of the messages back to each request (both the user and assistant messages) to have a conversation.

<!-- gh-comment-id:2589325494 --> @pdevine commented on GitHub (Jan 14, 2025): Just keep adding each of the messages back to each request (both the user and assistant messages) to have a conversation.
GiteaMirror commented 2026-05-04 10:23:42 -05:00
Author
Owner
Copy Link

@gl2007 commented on GitHub (Jan 14, 2025):

Ok, sounds ok to me but wondering if making it stateful with an optional session based approach would be worthwhile. Would save some work for every client consuming this to write same code.

<!-- gh-comment-id:2591105860 --> @gl2007 commented on GitHub (Jan 14, 2025): Ok, sounds ok to me but wondering if making it stateful with an optional session based approach would be worthwhile. Would save some work for every client consuming this to write same code.
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 bug
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#67447
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?