[BUG] API Configuration for Context Condensing” resets to *Global Default* and condensing uses the wrong provider

[jmalmo]

Problem (one or two sentences)

When I set Settings → Context → Check the Automatically trigget intelligent context condensing and pick my OpenRouter profile (Claude 4.5 Sonnet), the selection does not persist: after leaving and re-opening settings it’s back to Global Default (all profiles).

When I click Condense context, the spinner (“Condensing context…”) runs indefinitely and Roo queues any new messages. Inspecting the Extension Host log shows the condense call is made against Google Gemini 2.5 Pro (which I do not have credits to access), not the profile I selected.

Additionally, shouldn't the profile picker for context condensing be exposed regardless of whether the Automatically trigger intelligent context condensing checkbox is checked or not?

Context (who is affected and when)

All users who wants to edit the provider settings for condensing context, and all users like myself who are stuck on an old provider they do not have access to and therefore cannot use the condense functionality.

Reproduction steps

1. Open Roo Code Settings → Context.
2. Set API Configuration for Context Condensing to an OpenRouter profile (e.g., OpenRouter – Claude 4.5 Sonnet).
3. Navigate away from Settings, then return to confirm: the dropdown is back to Global Default (all profiles).
4. In a chat with sufficient context, click Condense context.
5. Observe: UI shows “Condensing context…” indefinitely; any new messages are marked Queued.
6. Optionally: Check Output → Roo Code (Extension Host): if it was set to an API profile that you have not credits to access, the condense request is sent, but fails on quota.

Expected result

The profile picker should persist, and the context condensing should show.

Actual result

The profile picker resets back to Global Default (all profiles), and in my case the condensing fails and the chat is stuck.

Variations tried (optional)

No response

App Version

v3.29.2

API Provider (optional)

None

Model Used (optional)

No response

Roo Code Task Links (optional)

No response

Relevant logs or errors (optional)

2025-10-27 23:57:34.719 [error] Error: Gemini generate context stream error: got status: 429 Too Many Requests. {"error":{"message":"{ ... "model": "gemini-2.5-pro" ... }"}}
at jU.createMessage (.../rooveterinaryinc.roo-cline-3.29.2/dist/extension.js:3411:2210)
at async X0t (.../rooveterinaryinc.roo-cline-3.29.2/dist/extension.js:5320:1062)
at async t.condenseContext (.../rooveterinaryinc.roo-cline-3.29.2/dist/extension.js:5862:14931)
at async t.condenseTaskContext (.../rooveterinaryinc.roo-cline-3.29.2/dist/extension.js:5947:6266)

[jmalmo]

Here is a GIF of how it looks for me:

[roomote]

I'm exploring a potential fix for the context condensing provider persistence issue. I'll investigate why the selected provider doesn't persist and why it falls back to using Gemini instead of the configured provider.

[roomote]

I've created a pull request to fix this issue: #8873

The fix adds the missing condensing API configuration dropdown to the Context Management settings section, making it always visible regardless of the auto-condense checkbox state. This ensures your API configuration selection persists and the correct provider is used for context condensing operations.

All tests are passing and the implementation is ready for review.