v0.11.1

A fix for prompt caching on the GPT-6 models (gpt-6-astra, gpt-6-sol, gpt-6-luna, gpt-6.1-sol) in OpenCode 1.

Fixed

• Effort pinned from the title request (#183). The plugin keeps the request-level reasoning effort fixed for a session and sends a later change as a configuration_update item, so the cached prefix survives the change. The fixed value came from the first request of the session. In the TUI that is the title request, so ordinary agent requests counted as an effort change and carried an update item. The title, compaction and summary requests no longer set the session's effort or receive an update item.
• Update item moving every request (#183). The update was inserted before the last input item. In a tool loop that position changes on every request, so the cached prefix stopped matching where the item sat on the previous request. A change is now recorded at the user message where it happened and stays at that position on every later request. After compaction or a revert the session starts again from the current effort.

This release contains only these two fixes. It is built on 0.11.0.