[PR #15750] [MERGED] launch: keep launch recommended models in a fixed canonical order #41163

Closed
opened 2026-04-23 01:53:07 -05:00 by GiteaMirror · 0 comments
Owner

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/15750
Author: @hoyyeva
Created: 4/22/2026
Status: Merged
Merged: 4/23/2026
Merged by: @hoyyeva

Base: mainHead: hoyyeva/recommended-list


📝 Commits (3)

  • 0becfd3 launch: keep launch recommended models in a fixed canonical order
  • ed80741 add test
  • e0fce6e address recommended list review feedback

📊 Changes

3 files changed (+48 additions, -28 deletions)

View changed files

📝 cmd/launch/integrations_test.go (+36 -12)
📝 cmd/launch/launch_test.go (+4 -2)
📝 cmd/launch/models.go (+8 -14)

📄 Description

Summary

This change keeps the ollama launch recommended model list pinned to a fixed canonical order, regardless of which recommended models are already saved or pre-checked for an integration.

The recommended block is now always:

[]string{"kimi-k2.6:cloud", "qwen3.5:cloud", "glm-5.1:cloud", "minimax-m2.7:cloud", "gemma4", "qwen3.5"}

Checked models still float to the top of the More section, and the current/default model behavior within checked non-recommended items is preserved.

After kimi-k2.6:cloud replaced kimi-k2.5:cloud in the recommended list, previously saved selections could reshuffle the visible recommended block. In practice this made the launch UI show different recommended ordering depending on old saved config, which was confusing and made the recommendations look unstable.

What changed

  • Update buildModelList sorting so recommended models always sort ahead of non-recommended models.
  • Keep recommended models ordered strictly by recommendedModels, ignoring checked/prechecked status.
  • Preserve checked/default prioritization only for non-recommended models in the More section.
  • Update launch tests to cover:
    • fixed recommended ordering
    • checked recommended models not reshuffling the recommended block
    • checked non-recommended models still floating to the top of More
    • launch selector behaviour using the new fixed ordering

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/ollama/ollama/pull/15750 **Author:** [@hoyyeva](https://github.com/hoyyeva) **Created:** 4/22/2026 **Status:** ✅ Merged **Merged:** 4/23/2026 **Merged by:** [@hoyyeva](https://github.com/hoyyeva) **Base:** `main` ← **Head:** `hoyyeva/recommended-list` --- ### 📝 Commits (3) - [`0becfd3`](https://github.com/ollama/ollama/commit/0becfd389ff85da22d352370f510cf0560d34361) launch: keep launch recommended models in a fixed canonical order - [`ed80741`](https://github.com/ollama/ollama/commit/ed80741bdbd3d3c333960eba546663748bd50134) add test - [`e0fce6e`](https://github.com/ollama/ollama/commit/e0fce6e1c70bed8b792c281ac9fc5b209ada36c5) address recommended list review feedback ### 📊 Changes **3 files changed** (+48 additions, -28 deletions) <details> <summary>View changed files</summary> 📝 `cmd/launch/integrations_test.go` (+36 -12) 📝 `cmd/launch/launch_test.go` (+4 -2) 📝 `cmd/launch/models.go` (+8 -14) </details> ### 📄 Description ### Summary This change keeps the ollama launch recommended model list pinned to a fixed canonical order, regardless of which recommended models are already saved or pre-checked for an integration. The recommended block is now always: ``` []string{"kimi-k2.6:cloud", "qwen3.5:cloud", "glm-5.1:cloud", "minimax-m2.7:cloud", "gemma4", "qwen3.5"} ``` Checked models still float to the top of the More section, and the current/default model behavior within checked non-recommended items is preserved. After `kimi-k2.6:cloud` replaced `kimi-k2.5:cloud` in the recommended list, previously saved selections could reshuffle the visible recommended block. In practice this made the launch UI show different recommended ordering depending on old saved config, which was confusing and made the recommendations look unstable. ### What changed - Update buildModelList sorting so recommended models always sort ahead of non-recommended models. - Keep recommended models ordered strictly by recommendedModels, ignoring checked/prechecked status. - Preserve checked/default prioritization only for non-recommended models in the More section. - Update launch tests to cover: - fixed recommended ordering - checked recommended models not reshuffling the recommended block - checked non-recommended models still floating to the top of More - launch selector behaviour using the new fixed ordering --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
GiteaMirror added the pull-request label 2026-04-23 01:53:07 -05:00
GiteaMirror changed title from [PR #15750] launch: keep launch recommended models in a fixed canonical order to [PR #15750] [MERGED] launch: keep launch recommended models in a fixed canonical order 2026-04-25 01:54:58 -05:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#41163