github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 21:25:46 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #1720] Have response return minimized JSON (not formatted) to save token space #63012

New Issue
Closed
opened 2026-05-03 11:15:47 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-05-03 11:15:47 -05:00
Owner
Copy Link

Originally created by @erf on GitHub (Dec 26, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1720

When i get JSON as a response it seem to be formatted with newlines and spaces. If i want to include the response message in follow-up request it will take up extra token space. Maybe better to have the client format the JSON as needed?

EDIT: I'm not really sure if i need to include the previous messages or if i should just set the context ?

Originally created by @erf on GitHub (Dec 26, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1720 When i get JSON as a response it seem to be formatted with newlines and spaces. If i want to include the response message in follow-up request it will take up extra token space. Maybe better to have the client format the JSON as needed? EDIT: I'm not really sure if i need to include the previous messages or if i should just set the `context` ?
GiteaMirror added the feature request label 2026-05-03 11:15:47 -05:00
GiteaMirror commented 2026-05-03 11:15:49 -05:00
Author
Owner
Copy Link

@BruceMacD commented on GitHub (Jan 2, 2024):

Hi @erf, I'd suggest using multi-shot prompting along with the /chat api to achieve this goal. Start the chat with some minified JSON, and then the LLM will be more likely to respond with minified JSON. It also means it will be easier to reformat the responses to save tokens.

Here is an example:

curl http://localhost:11434/api/chat -d '{
    "model": "llama2",
    "format": "json",
    "stream": false,
    "messages": [
        {
            "role": "user",
            "content": "1234 example ave"
        },
        {
            "role": "assistant",
            "content": "{\"number\":1234,\"street\":\"example ave\"}"
        },
        {
            "role": "user",
            "content": "4567 oak st"
        },
        {
            "role": "assistant",
            "content": "{\"number\":4567,\"street\":\"oak st\"}"
        },
        {
            "role": "user",
            "content": "89 another ave"
        }
    ]
}'

and the response I got:

{
  "model": "llama2",
  "created_at": "2024-01-02T10:30:50.036408Z",
  "message": {
    "role": "assistant",
    "content": "{\"number\":89,\"street\":\"another ave\" }"
  },
  "done": true,
  "total_duration": 605612250,
  "load_duration": 2409375,
  "prompt_eval_count": 102,
  "prompt_eval_duration": 382979000,
  "eval_count": 12,
  "eval_duration": 213132000
}

Hope that helps.

<!-- gh-comment-id:1873851420 --> @BruceMacD commented on GitHub (Jan 2, 2024): Hi @erf, I'd suggest using multi-shot prompting along with the `/chat` api to achieve this goal. Start the chat with some minified JSON, and then the LLM will be more likely to respond with minified JSON. It also means it will be easier to reformat the responses to save tokens. Here is an example: ``` curl http://localhost:11434/api/chat -d '{ "model": "llama2", "format": "json", "stream": false, "messages": [ { "role": "user", "content": "1234 example ave" }, { "role": "assistant", "content": "{\"number\":1234,\"street\":\"example ave\"}" }, { "role": "user", "content": "4567 oak st" }, { "role": "assistant", "content": "{\"number\":4567,\"street\":\"oak st\"}" }, { "role": "user", "content": "89 another ave" } ] }' ``` and the response I got: ``` { "model": "llama2", "created_at": "2024-01-02T10:30:50.036408Z", "message": { "role": "assistant", "content": "{\"number\":89,\"street\":\"another ave\" }" }, "done": true, "total_duration": 605612250, "load_duration": 2409375, "prompt_eval_count": 102, "prompt_eval_duration": 382979000, "eval_count": 12, "eval_duration": 213132000 } ``` Hope that helps.
GiteaMirror commented 2026-05-03 11:15:50 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Mar 13, 2024):

Going to go ahead and close the issue, but feel free to keep commenting.

<!-- gh-comment-id:1996100819 --> @pdevine commented on GitHub (Mar 13, 2024): Going to go ahead and close the issue, but feel free to keep commenting.
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#63012
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?