github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #11811] API client silently ignores large responses #7838

New Issue
Closed
opened 2026-04-12 20:00:22 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-12 20:00:22 -05:00
Owner
Copy Link

Originally created by @msiebuhr on GitHub (Aug 8, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11811

What is the issue?

When uploading a very large document through the Golang APIs Generate method (https://pkg.go.dev/github.com/ollama/ollama/api#Client.Generate, version 0.11.3), the API returns without error or ever having called fn (GenerateResponseFunc):

In this case a "please turn this HTML table into CSV" and then 2.2 MB of HTML table with plenty of white space.

req := &api.GenerateRequest{
    Model: "<whatever>",
    Prompt: "<something very large>",
    Stream: new(bool),
}

err := client.Generate(ctx, req, func(r api.GenerateResponse) {
    // Never called
    fmt.Print(r.Response)
    return nil
})

if err != nil {
    // No error either
    panic(err)
}

If I manually build the same request and pipe it through curl, it works fine and a reply is returned.

From my reading of the sources at 114c3f2265/api/client.go (L211-L214) and documentation for bufio scanner.Buffer() https://pkg.go.dev/bufio#Scanner.Buffer, it does seem that the Ollama client will ignore answers larger than maxBufferSize which in turn is hard-coded to 512 KB 114c3f2265/api/client.go (L162)

If I use stream: true I get most of the data back but it still swallows the very last message with the context.

I used a script to generate the large-ish requests, here a ~2.2 MB request:

% ./make_request.py | jq . -c | wc --chars                                                                  
2264580

And showing the large context (~1.8 MB) after running through a ollama:

% ./make_request.py | curl -s http://localhost:11434/api/generate --data-binary @- | jq .context -c | wc --chars
1803319

Relevant log output

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.11.3

Originally created by @msiebuhr on GitHub (Aug 8, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11811 ### What is the issue? When uploading a very large document through the Golang APIs `Generate` method (https://pkg.go.dev/github.com/ollama/ollama/api#Client.Generate, version 0.11.3), the API returns without error or ever having called `fn` (GenerateResponseFunc): In this case a "please turn this HTML table into CSV" and then 2.2 MB of HTML table with plenty of white space. ```go req := &api.GenerateRequest{ Model: "<whatever>", Prompt: "<something very large>", Stream: new(bool), } err := client.Generate(ctx, req, func(r api.GenerateResponse) { // Never called fmt.Print(r.Response) return nil }) if err != nil { // No error either panic(err) } ``` If I manually build the same request and pipe it through `curl`, it works fine and a reply is returned. From my reading of the sources at https://github.com/ollama/ollama/blob/114c3f22657750cfb57f70c4a0d6e7389fb7a9fe/api/client.go#L211-L214 and documentation for `bufio scanner.Buffer()` https://pkg.go.dev/bufio#Scanner.Buffer, it does seem that the Ollama client will ignore answers larger than `maxBufferSize` which in turn is hard-coded to 512 KB https://github.com/ollama/ollama/blob/114c3f22657750cfb57f70c4a0d6e7389fb7a9fe/api/client.go#L162 If I use `stream: true` I get most of the data back but it still swallows the very last message with the context. ---- I used a script to generate the large-ish requests, here a ~2.2 MB request: ``` % ./make_request.py | jq . -c | wc --chars 2264580 ``` And showing the large context (~1.8 MB) after running through a ollama: ``` % ./make_request.py | curl -s http://localhost:11434/api/generate --data-binary @- | jq .context -c | wc --chars 1803319 ``` ### Relevant log output ```shell ``` ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.11.3
GiteaMirror added the bug label 2026-04-12 20:00:22 -05:00
GiteaMirror commented 2026-04-12 20:00:23 -05:00
Author
Owner
Copy Link

@msiebuhr commented on GitHub (Aug 8, 2025):

Seem to be introduced in 9e2de1bd2c

<!-- gh-comment-id:3167976331 --> @msiebuhr commented on GitHub (Aug 8, 2025): Seem to be introduced in https://github.com/ollama/ollama/commit/9e2de1bd2c09cfc6a68deb50e7ec5033df6d22ed
GiteaMirror commented 2026-04-12 20:00:23 -05:00
Author
Owner
Copy Link

@msiebuhr commented on GitHub (Jan 16, 2026):

Fixed by 9667c2282f

<!-- gh-comment-id:3758725592 --> @msiebuhr commented on GitHub (Jan 16, 2026): Fixed by 9667c2282f477fb3ba947585c5417ffbc4654a43
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#7838
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?