feat: rebuild the Cursor adapter for OpenCode V2

[CasualDeveloper]

Hey, @makeittech - was excited to try out Astra and I think it did a great job with our plugin!

Issue for this PR

Refs #31. Please keep it open through publication and release checks.

This PR supersedes #34 and includes its release-flow documentation and constants-test explanation. Please review and merge this PR directly.

Type of change

• New feature
• Refactor / code improvement

What does this PR do?

I rebuilt the Cursor adapter for OpenCode V2 and removed the old proxy and compatibility code. It keeps private models and exact variants while OpenCode manages history, tools, permissions, and compaction.

Each new Cursor Run is built from OpenCode's messages and tool results, including images and reasoning data. This lets the host recover after a lost connection or restart without a second conversation store.

It also fixes usage reporting that could trigger compaction too early. Cursor Run totals are saved separately from per-call usage. Where per-call counts are unavailable, OpenCode's displayed token and cost totals remain incomplete.

I updated the adapter after OpenCode removed context.catalog. It now uses context.provider.transform() and reload(), refreshes models on credential.switched, and preserves failed-tool outcomes in structured compaction requests. Dependencies and the test host are pinned to released @opencode/* 2.0.6. Older hosts with the catalog API are no longer supported.

Grok opaque-reasoning checkpoints no longer fail a completed turn when a root has no unique assistant anchor. Cursor usable-model IDs group into families: none is the non-thinking SKU, and thinking SKUs become low/medium/high/xhigh/max.

The OpenCode id stays the stable catalog name, such as grok-4.6. Each variant still sends Cursor's exact wire id, such as cursor-grok-4.6-xhigh. All non-duplicate models from AvailableModels are published so newly released and parameterized models (such as Grok 4.7 and Grok 4.7 500K) appear in OpenCode's picker. If OpenCode has loaded a models.dev Cursor row, I copy that row's limits and price. I do not copy lab caps, and I do not replace the live catalog with a static list.

I also fixed model grouping against Cursor's real parameter spellings: any effort parameter (including reasoning_effort and minimal) groups into variants on one model, and Claude's thinking flag stays on that same model (none is non-thinking, while effort variants keep thinking enabled) instead of creating separate models. Connect error message details are also preserved in the transport.

The restrictive x-cursor-agent-allowed-tools header is omitted when tools are enabled so Cursor does not strip its internal meta-tools (GET_MCP_TOOLS) or fail at bootstrap, and only sent as "" when tools are explicitly disabled. The runner supports Cursor's native workspace execution requests (shellStreamArgs, shellArgs, grepArgs, readArgs, writeArgs, lsArgs) by mapping them to OpenCode host tools and formatting their native results. Calls to grepArgs with an empty or whitespace pattern automatically route to OpenCode's glob tool with files_with_matches results, and tool result JSON envelopes are unwrapped for native Cursor actions.

How did you verify your code works?

The current head is b99cfe8. I amended this onto the previous head because #35 had no review activity. Unit tests and typecheck pass locally.

Live end-to-end multi-turn tool loops were verified on Mac and WSL for both Composer 2.5 and Grok 4.7 (reading files, searching glob/grep, executing shell commands, and returning final answers).

The earlier CI run for 30f75a3 failed in test:v2-host: OpenCode 2.0.6 rejected a permission reply as Missing key ["decision"].

Instruction priority, opaque-reasoning replay, and continued work after automatic compaction still need live verification.

The test report records the remaining limits.

Checklist

• I have tested my changes locally.
• I have not included unrelated changes in this PR.