github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 16:40:08 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #9670] Context Modification for Stop Extended Thinking Process #68366

New Issue
Open
opened 2026-05-04 13:35:32 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-04 13:35:32 -05:00
Owner
Copy Link

Originally created by @123gggwnnggg on GitHub (Mar 12, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9670

I’m writing to propose a feature enhancement that could improve how the model handles extended thinking.
Some reasoning model can spend way to long on thinking, so the idea is to implement a context modification mechanism where a marker is automatically inserted after the last period ('.') when the model thinking for 1/2 of the max response.
Example: (limit at 10 words.)

<think> I am thinking. Keep thinking. Still thinking thinking thinking thinking

modify to

<think> I am thinking. Keep thinking.</think>
Originally created by @123gggwnnggg on GitHub (Mar 12, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9670 I’m writing to propose a feature enhancement that could improve how the model handles extended thinking. Some reasoning model can spend way to long on thinking, so the idea is to implement a context modification mechanism where a </think> marker is automatically inserted after the last period ('.') when the model thinking for 1/2 of the max response. Example: (limit at 10 words.) ``` <think> I am thinking. Keep thinking. Still thinking thinking thinking thinking ``` modify to ``` <think> I am thinking. Keep thinking.</think> ```
GiteaMirror added the feature request label 2026-05-04 13:35:32 -05:00
GiteaMirror commented 2026-05-04 13:35:36 -05:00
Author
Owner
Copy Link

@JasonHonKL commented on GitHub (Mar 22, 2025):

I think this may not be suitable because it would be the application layer user (i.e developer) to handle these response properly. For instance you could take a look with how Langchain handle this case with Deepseek. It's not Ollama responsibility which is an applicably focusing on hosting LLM locally.

<!-- gh-comment-id:2745257471 --> @JasonHonKL commented on GitHub (Mar 22, 2025): I think this may not be suitable because it would be the application layer user (i.e developer) to handle these response properly. For instance you could take a look with how Langchain handle this case with Deepseek. It's not Ollama responsibility which is an applicably focusing on hosting LLM locally.
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 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#68366
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?