DevAgentRadar → OpenCode

You are helping me adopt a real coding-assistant change. Work only from the facts below. Do not invent features.

Context

Assistant: OpenCode Proposal: Upgrade OpenCode to v1.18.30 and route GPT-6 through Astra Summary: OpenCode v1.18.30 adds the Astra system prompt for GPT-6, GitLab GPT/Claude reasoning-effort variants, preserved Bedrock DeepSeek ARN model IDs, and Azure/OpenAI SDK compatibility fixes. Upgrade, then retarget GPT-6, GitLab, and Bedrock DeepSeek so those behaviors apply. Primary source: https://github.com/anomalyco/opencode/releases/tag/v1.18.30

Why it matters

This release changes which prompts, model IDs, and provider SDKs OpenCode actually uses. GPT-6 now has a dedicated Astra system prompt, so staying on an older build means GPT-6 sessions will not get it. Supported GitLab GPT and Claude models gained reasoning-effort variants (#47306), so you can spend more effort on refactors and less on small edits without switching providers. Bedrock DeepSeek IDs, including ARN-based IDs, are preserved so they resolve (#34441); if you shortened or rewrote those IDs to work around resolution failures, restore the original IDs. Azure and OpenAI provider SDKs were updated for compatibility, and console OpenAI usage/tier-threshold handling was normalized (#47342). Bump to v1.18.30 first, then retarget GPT-6, GitLab variants, and Bedrock DeepSeek in the same pass so daily agent sessions pick up the new prompt, variants, and ID handling.

Suggested actions

1. Upgrade OpenCode to v1.18.30 and restart it so GPT-6 loads the Astra system prompt, GitLab GPT/Claude expose reasoning-effort variants, Bedrock DeepSeek keeps ARN-based model IDs, and the Azure/OpenAI provider SDKs include the compatibility fixes.

2. Route all GPT-6 coding sessions through OpenCode v1.18.30 so the new Astra system prompt is applied; do not keep GPT-6 on a pre-v1.18.30 OpenCode binary.

3. For supported GitLab GPT and Claude models in OpenCode, switch to the reasoning-effort variant IDs added in v1.18.30 (#47306): higher effort for refactors and reviews, lower effort for quick edits.

4. In OpenCode Bedrock config, set DeepSeek to the full original model IDs including ARN-based IDs. v1.18.30 preserves those IDs so they resolve—remove any shortened or rewritten DeepSeek IDs from workarounds.

5. After the upgrade, use the updated Azure and OpenAI provider SDKs in OpenCode v1.18.30 with your existing Azure OpenAI and OpenAI credentials so the compatibility fixes and console OpenAI usage/tier-threshold normalization are in effect.

After you finish

Do not report this as applied to DevAgentRadar. You cannot write the visitor's loop.

Tell the human: open https://devagentradar.com/proposals/opencode-v1-18-30-upgrade-opencode-to-v1-18-30-and-route-gpt-6-through-a and mark Applied, Skipped, or Failed. Proposal id: de3b051a-5c40-4233-88e9-8b5681d5caf7

Your job

1. Restate the change in one sentence.
2. Propose a minimal plan for my repo (or a throwaway pilot).
3. Implement only what I approve; prefer small diffs and tests.
4. Call out risks (permissions, breaking APIs, cost).

Start by confirming you understood the proposal.