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 #15817] ollama launch claude --model kimi-k2.6:cloud can not read image #56592

New Issue
Open
opened 2026-04-29 11:04:11 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-04-29 11:04:11 -05:00
Owner
Copy Link

Originally created by @chenhao1umbc on GitHub (Apr 25, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15817

What is the issue?

ollama launch claude --model kimi-k2.6:cloud
the Claude Code can not read image, it said Read tool did not work

Relevant log output

OS

No response

GPU

No response

CPU

No response

Ollama version

No response

Originally created by @chenhao1umbc on GitHub (Apr 25, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15817 ### What is the issue? ollama launch claude --model kimi-k2.6:cloud the Claude Code can not read image, it said Read tool did not work ### Relevant log output ```shell ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version _No response_
GiteaMirror added the bug label 2026-04-29 11:04:11 -05:00
GiteaMirror commented 2026-04-29 11:04:13 -05:00
Author
Owner
Copy Link

@hoyyeva commented on GitHub (Apr 27, 2026):

Hi @chenhao1umbc,

Thank you for the report, I wasn't able to reproduce this on my side. Could you please share your environment details, ollama version, exact steps to reproduce, any logs or error output and whether this happens with all image or only specific ones so we can investigate further.

<!-- gh-comment-id:4328697187 --> @hoyyeva commented on GitHub (Apr 27, 2026): Hi @chenhao1umbc, Thank you for the report, I wasn't able to reproduce this on my side. Could you please share your environment details, ollama version, exact steps to reproduce, any logs or error output and whether this happens with all image or only specific ones so we can investigate further.
GiteaMirror commented 2026-04-29 11:04:14 -05:00
Author
Owner
Copy Link

@chenhao1umbc commented on GitHub (Apr 27, 2026):

Title: Read tool returns empty output for images when using Claude Code via Ollama (ollaude)

Description:
When running Claude Code through an Ollama backend (using ollaude wrapper), the built-in Read tool fails to process image files (PNG/JPG). It returns empty
content instead of describing the image. The same image file reads correctly when using Claude Code's native Anthropic backend.

Environment:

  • Ollama version: 0.21.2
  • Claude Code version: 2.1.119
  • Model: kimi-k2.6:cloud (via Ollama)
  • OS: macOS (Darwin 25.4.0)
  • Container runtime: nerdctl version 2.2.1

Steps to Reproduce:

  1. Launch Claude Code via Ollama wrapper:
    ollama launch claude --model kimi-k2.6:cloud
  2. Inside the Claude Code session, ask it to read/describe an image:
    describe the image /Users/hc/Downloads/Screenshot\ 2026-04-25\ at\ 9.20.06\ PM.png
  3. Observe that the Read tool is invoked on the file, but returns empty output and began to try bash till it failed and reported.

Expected Behavior:
The Read tool should return the image content (or a description/render of it), allowing the model to process it as a multimodal input — just like it does when using the native Anthropic backend.

Actual Behavior:
The Read tool returns empty content. The model then cannot see the image and fails to describe it.

Additional Context:

  • The image file exists and is valid (verified via direct Read on host).
  • This appears to be specific to the Ollama-routed Claude Code session (inside nerdctl container with ANTHROPIC_BASE_URL=http://host.lima.internal:11434).
  • The img_read skill (Python script using OCR/VLM) works as a workaround, but the native Read tool should handle images directly.
<!-- gh-comment-id:4331014842 --> @chenhao1umbc commented on GitHub (Apr 27, 2026): --- Title: Read tool returns empty output for images when using Claude Code via Ollama (ollaude) Description: When running Claude Code through an Ollama backend (using ollaude wrapper), the built-in Read tool fails to process image files (PNG/JPG). It returns empty content instead of describing the image. The same image file reads correctly when using Claude Code's native Anthropic backend. Environment: - Ollama version: 0.21.2 - Claude Code version: 2.1.119 - Model: kimi-k2.6:cloud (via Ollama) - OS: macOS (Darwin 25.4.0) - Container runtime: nerdctl version 2.2.1 Steps to Reproduce: 1. Launch Claude Code via Ollama wrapper: ollama launch claude --model kimi-k2.6:cloud 2. Inside the Claude Code session, ask it to read/describe an image: describe the image /Users/hc/Downloads/Screenshot\ 2026-04-25\ at\ 9.20.06\ PM.png 3. Observe that the Read tool is invoked on the file, but returns empty output and began to try bash till it failed and reported. Expected Behavior: The Read tool should return the image content (or a description/render of it), allowing the model to process it as a multimodal input — just like it does when using the native Anthropic backend. Actual Behavior: The Read tool returns empty content. The model then cannot see the image and fails to describe it. Additional Context: - The image file exists and is valid (verified via direct Read on host). - This appears to be specific to the Ollama-routed Claude Code session (inside nerdctl container with ANTHROPIC_BASE_URL=http://host.lima.internal:11434). - The img_read skill (Python script using OCR/VLM) works as a workaround, but the native Read tool should handle images directly.
GiteaMirror commented 2026-04-29 11:04:15 -05:00
Author
Owner
Copy Link

@balaena01 commented on GitHub (Apr 28, 2026):

same

<!-- gh-comment-id:4333166437 --> @balaena01 commented on GitHub (Apr 28, 2026): same
GiteaMirror commented 2026-04-29 11:04:16 -05:00
Author
Owner
Copy Link

@ParthSareen commented on GitHub (Apr 28, 2026):

Can you confirm that copy-pasting an image works. I think we might be missing some stuff on the tool definition.

<!-- gh-comment-id:4339656678 --> @ParthSareen commented on GitHub (Apr 28, 2026): Can you confirm that copy-pasting an image works. I think we might be missing some stuff on the tool definition.
GiteaMirror commented 2026-04-29 11:04:17 -05:00
Author
Owner
Copy Link

@chenhao1umbc commented on GitHub (Apr 29, 2026):

if I copy and paste, it works perfectly. Check this screenshot

Image
<!-- gh-comment-id:4340290152 --> @chenhao1umbc commented on GitHub (Apr 29, 2026): if I copy and paste, it works perfectly. Check this screenshot <img width="1288" height="528" alt="Image" src="https://github.com/user-attachments/assets/8a731424-a2b3-4411-b52a-0d50854ffab9" />
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#56592
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?