<!-- CURSOR_SUMMARY -->
> [!NOTE]
> <sup>[Cursor Bugbot](https://cursor.com/dashboard?tab=bugbot) is
generating a summary for commit
fb0aff0bf5e21e1b57bc8744ed9643b863b888d4. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Adds support for Anthropic Claude Opus 4.5 and standardizes Sonnet
naming. Removes legacy 3.x Anthropic models and updates tests to use the
new names.
- **New Features**
- Added "Claude Opus 4.5" with 32k max output and 200k context.
- **Migration**
- Names standardized to "Claude Sonnet 4" and "Claude Sonnet 4.5".
Update any model selections and tests.
- Removed Anthropic 3.x models: Claude 3.7 Sonnet, Claude 3.5 Sonnet,
Claude 3.5 Haiku, and their Bedrock variants.
<sup>Written for commit fb0aff0bf5e21e1b57bc8744ed9643b863b888d4.
Summary will update automatically on new commits.</sup>
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated description by cubic. -->
## Summary by cubic
Set “balanced” as the default smart context mode. Users now get balanced
when Smart Files Context is enabled and no mode is set; “conservative”
must be explicitly selected.
- **Refactors**
- Default fallback to balanced in UI and engine (proSmartContextOption
undefined -> "balanced").
- ProModeSelector saves "conservative" explicitly; selector reads
undefined as balanced.
- Updated schema and types to allow "balanced" | "conservative".
- Engine payload now includes smart_context_mode with "balanced" by
default; e2e tests and snapshots updated.
- **Migration**
- No action needed. Existing users without an explicit mode will use
balanced by default; selecting conservative persists.
<!-- End of auto-generated description by cubic. -->
1. Show less common AI providers (secondary) in submenu
2. Show $ signs for rough cost guide
3. Show "Pro" for supported AI providers with Pro is enabled
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Improves the Model Picker UX by grouping less-used providers under an
“Other AI providers” submenu and adding clear cost and Pro indicators.
This makes picking models faster and more informative.
- **New Features**
- Grouped secondary providers under “Other AI providers” using a new
provider.secondary flag (Azure marked secondary).
- Added rough cost hints: models can set dollarSigns and the UI shows a
“$” badge accordingly.
- Shows a “Pro” badge on supported cloud providers when Pro is enabled;
added a “Custom” badge for custom providers.
- Extended types: LanguageModelProvider.secondary and
LanguageModel.dollarSigns; populated values across OpenAI, Anthropic,
Google, and OpenRouter.
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated description by cubic. -->
## Summary by cubic
Adds Smart Context v3 with selectable modes (Off, Conservative,
Balanced) and surfaces token savings in chat. Also improves token
estimation by counting per-file tokens when Smart Context is enabled.
- **New Features**
- Smart Context selector in Pro settings with three options.
Conservative is the default when enabled without an explicit choice.
- New setting: proSmartContextOption ("balanced"); undefined implies
Conservative.
- Engine now receives enable_smart_files_context and smart_context_mode.
- Chat shows a DyadTokenSavings card when the message contains
token-savings?original-tokens=...&smart-context-tokens=..., with percent
saved and a tooltip for exact tokens.
- Token estimation uses extracted file contents for accuracy when Pro +
Smart Context is on; otherwise falls back to formatted codebase output.
<!-- End of auto-generated description by cubic. -->
This uses Gemini's native [thinking
summaries](https://cloud.google.com/vertex-ai/generative-ai/docs/thinking#thought-summaries)
which were recently added to the API.
Why? The grafted thinking would sometimes cause weird issues where the
model, especially Gemini 2.5 Flash, got confused and put dyad tags like
`<dyad-write>` inside the `<think>` tags.
This also improves the UX because you can see the native thoughts rather
than having the Gemini response load for a while without any feedback.
I tried adding Anthropic extended thinking, however it requires temp to
be set at 1, which isn't ideal for Dyad's use case where we need precise
syntax following.
