[BUG] Force update the ollama models?

[randomizedcoder]

Problem (one or two sentences)

Friendly feedback

It's unclear how to get roocode to update the models it thinks are available via ollama

I'm running ollama locally, and I have gpt-oss:20b available, but roocode can't find it. Weird.

The docs don't mention anything
https://docs.roocode.com/providers/ollama?utm_source=extension&utm_medium=ide&utm_campaign=provider_docs

My setup is like this ( except the video is a bit old, and I'm tracking NixOS unstable ):
https://youtu.be/mUrVC7oo2_g?si=XjcBwf2O2JOmUhhJ

roocode version is
Version: 3.41.1 (d49bd9a)

Context (who is affected and when)

People using ollama

Reproduction steps

[das@l:~/nixos/desktop/l]$ neofetch
          ▗▄▄▄       ▗▄▄▄▄    ▄▄▄▖            das@l 
          ▜███▙       ▜███▙  ▟███▛            ----- 
           ▜███▙       ▜███▙▟███▛             OS: NixOS 26.05.20260113.1412caf (Yarara) x86_64 
            ▜███▙       ▜██████▛              Host: LENOVO 1046 
     ▟█████████████████▙ ▜████▛     ▟▙        Kernel: 6.18.5 
    ▟███████████████████▙ ▜███▙    ▟██▙       Uptime: 7 hours, 47 mins 
           ▄▄▄▄▖           ▜███▙  ▟███▛       Packages: 1216 (nix-system), 1764 (nix-user) 
          ▟███▛             ▜██▛ ▟███▛        Shell: bash 5.3.3 
         ▟███▛               ▜▛ ▟███▛         Resolution: 3840x2160, 5120x1440, 3840x2160 
▟███████████▛                  ▟██████████▙   DE: GNOME 49.2 (Wayland) 
▜██████████▛                  ▟███████████▛   Theme: Nordic [GTK2/3] 
      ▟███▛ ▟▙               ▟███▛            Icons: Papirus-Dark [GTK2/3] 
     ▟███▛ ▟██▙             ▟███▛             Terminal: kgx 
    ▟███▛  ▜███▙           ▝▀▀▀▀              CPU: AMD Ryzen Threadripper PRO 3945WX s (24) @ 4.427GHz 
    ▜██▛    ▜███▙ ▜██████████████████▛        GPU: AMD ATI Radeon Pro VII/Radeon Instinct MI50 32GB 
     ▜▛     ▟████▙ ▜████████████████▛         GPU: AMD ATI Radeon Pro W7500 
           ▟██████▙       ▜███▙               Memory: 19854MiB / 128661MiB 
          ▟███▛▜███▙       ▜███▙
         ▟███▛  ▜███▙       ▜███▙                                     
         ▝▀▀▀    ▀▀▀▀▘       ▀▀▀▘                                     


[das@l:~/nixos/desktop/l]$ 

Full nixos config here: https://github.com/randomizedcoder/nixos/tree/main/desktop/l

Expected result

ability to refresh the models for ollama

Actual result

not refreshed

Variations tried (optional)

No response

App Version

3.41.1

API Provider (optional)

Ollama

Model Used (optional)

No response

Roo Code Task Links (optional)

No response

Relevant logs or errors (optional)

[roomote]

I've analyzed this issue and here's my proposed implementation plan:

Problem

The Ollama provider settings currently fetch models automatically when the settings page loads, but there's no way to manually refresh the model list. This means if you add a new model to Ollama after opening the settings, you have to close and reopen the settings to see it.

Solution

Add a "Refresh Models" button to the Ollama provider settings UI, similar to what already exists for LiteLLM.

Implementation Details

1. Modify webview-ui/src/components/settings/providers/Ollama.tsx:

   • Add a refreshStatus state to track loading/success/error states
   • Add a handleRefreshModels callback that sends requestOllamaModels message
   • Add a "Refresh Models" button using the existing Button component
   • Display loading/success/error feedback messages

2. No backend changes needed - The requestOllamaModels handler already:

   • Flushes the model cache to ensure fresh data
   • Fetches models from the Ollama API
   • Posts the models back to the webview

3. No new translation keys needed - We can reuse existing settings:providers.refreshModels.* keys

This approach follows the same pattern used by other providers like LiteLLM and provides a consistent user experience.

Please mention @roomote in a comment saying that you approve this plan, and I'll proceed with the implementation.

[randomizedcoder]

Dear roomote, this is a pretty cool suggestion, so thanks! I tried restarting the entire vscode, and that didn't help. So I don't know if the assumption that "The Ollama provider settings currently fetch models automatically when the settings page loads" is true

[dreness]

I was fighting with this today, and sniffing the traffic while doing it so I could see the API calls to my ollama service. I noticed that making a change to something like the api key would usually trigger a refresh. But I also had trouble getting it to display the model list, even though I could see it getting the listing via traffic inspection. After thinking for a minute, I turned back to Roo and there was the model list. So I suspect there is some kind of multi-request flow that has to complete before the list appears. In the traffic I can see it getting the list and then asking for details about each model in the list. There's probably also a time component to the behavior but I haven't nailed it down yet.

Maybe attempts to refresh it by poking stuff are effectively restarting the multi-request flow which serves to delay the appearance of the list in the UI.

[randomizedcoder]

@dreness I'm glad it's not just me!

I doubt this is going to be hard to fix, but I'm on NixOS, so I need the ability to build this

I created
#10811

If you would be so kind to verify it works for you, that would be awesome
https://youtu.be/izhaXYkqT_4?si=watRIaqxkVnkbC8z

Then I'll be able to review and fix the bug