feat: Add context window support for AI providers

[jiayi-xdu]

What specific problem does this solve?

Currently, Roo-Code relies on fixed context window sizes predetermined by model providers. During extended conversations or when processing large documents, the context window can grow excessively large, leading to significant performance degradation and noticeably slower response times. Furthermore, in scenarios with file size or upload restrictions, this oversized context frequently exceeds system limits, causing operations to fail and preventing users from completing their tasks.

Additional context (optional)

No response

Roo Code Task Links (Optional)

No response

Request checklist

• I've searched existing Issues and Discussions for duplicates
• This describes a specific problem with clear impact and context

Interested in implementing this?

• Yes, I'd like to help implement this feature

Implementation requirements

• I understand this needs approval before implementation begins

How should this be solved? (REQUIRED if contributing, optional otherwise)

Add Custom Context Window Size Support for Model Providers

• Support Custom Model Info in Schema: Extend the baseProviderSettingsSchema to include an optional CustomModelInfo field for user-defined token limits.
• Add An Input Field In UI: Added reusable context window input component with validation and visual feedback in webview-ui/src/components/common/ContextWindow.tsx. Support for displaying and editing custom model information, with default values used as fallback when no custom information is provided.
• Provider Model Handling: Update info if CustomModelInfo is provided. Get CustomModelInfo in getModel function in src/api/providers/*.ts, update info if CustomModelInfo is provided.

How will we know it works? (Acceptance Criteria - REQUIRED if contributing, optional otherwise)

Given a user editing a model provider's settings (e.g., Groq/Vertex)
When user select a model and enter "128000" in the context window size field and save the setting
Then the value is provided to model provider
And saved to CustomModelInfo when settings are submitted
But no changes are made to othermodel parameters

Technical considerations (REQUIRED if contributing, optional otherwise)

The implementation will affect the following files:
packages/types/src/provider-settings.ts
webview-ui/src/components/common/ContextWindow.tsx (new)
webview-ui/src/components/settings/providers/.tsx
src/api/providers/.tsx

Trade-offs and risks (REQUIRED if contributing, optional otherwise)

If user give a low context window and open Automativally trigger intelligent context condensing, summary may fail.

[rsashka]

It would also be very nice if the context overflow percentage (the total context size relative to the window size) was also calculated automatically along with the query cost, since the input token overflow percentage significantly affects the quality of the generated code.

[daniel-lxs]

Hey @jiayi-xdu, Thank you for creating this proposal. I'm not sure I understand what you are proposing. Do you want to set a different context window for models? I think the issue with context overflow won't be fixed with an arbitrary context window. I'm not even sure if that's supported by most providers.

I think the solution should involve making our token counting and overflow detection better, maybe even forcefully truncating context before sending it to the model to avoid errors.

I'll close this issue but feel free to continue the discussion or create an issue based on this.