Skip to content

opencode-go: declare Anthropic SDK for Qwen models - #6856

Closed
Macho0x wants to merge 1 commit into
anomalyco:devfrom
Macho0x:opencode-go-qwen-protocol
Closed

opencode-go: declare Anthropic SDK for Qwen models#6856
Macho0x wants to merge 1 commit into
anomalyco:devfrom
Macho0x:opencode-go-qwen-protocol

Conversation

@Macho0x

@Macho0x Macho0x commented Sep 11, 2026

Copy link
Copy Markdown

Problem

The OpenCode Go provider TOMLs for the Qwen models omit the per-model [provider] npm override, so consumers resolving the API protocol per model (the OpenCode CLI, Zed's new registry fetch in zed-industries/zed#64070) fall back to the provider-level default @ai-sdk/openai-compatible — but the gateway actually serves these models through the Anthropic Messages API.

Evidence already in the catalog:

  • qwen3.8-max.toml's own comments cite the Anthropic messages endpoint (https://help.aliyun.com/en/model-studio/anthropic-api-messages) and describe thinking.type toggle / budget_tokens / output_config.effort request fields.
  • All five entries use budget_tokens reasoning options, an Anthropic-API mechanism.
  • The sibling Go entry qwen3.8-flash.toml already declares [provider] npm = "@ai-sdk/anthropic".
  • The Zen entries qwen3.5-plus.toml and qwen3.6-plus.toml already declare [provider] npm = "@ai-sdk/anthropic".
  • Zed's bundled OpenCode provider likewise maps Qwen to the Anthropic protocol.

Change

Adds [provider] npm = "@ai-sdk/anthropic" to the OpenCode Go entries for qwen3.5-plus, qwen3.6-plus, qwen3.7-plus, qwen3.7-max, and qwen3.8-max, so the registry alone is authoritative about the protocol instead of requiring consumers to special-case the Qwen family.

bun validate passes.

Release Notes:

  • n/a (catalog data fix)

The OpenCode Go gateway serves the Qwen models through the Anthropic
Messages API: the entries already use budget_tokens reasoning, and
qwen3.8-max's own comments cite the Anthropic messages endpoint.
qwen3.8-flash already declares this via [provider] npm, and the Zen
qwen3.5-plus/qwen3.6-plus entries do as well.

Declare [provider] npm = "@ai-sdk/anthropic" on the remaining Go Qwen
entries (qwen3.5-plus, qwen3.6-plus, qwen3.7-plus, qwen3.7-max,
qwen3.8-max) so consumers resolving the protocol per model (e.g. the
OpenCode CLI and Zed's registry fetch) stop falling back to the
provider-level OpenAI-compatible default for these models.
@github-actions

Copy link
Copy Markdown
Contributor

No actionable findings.

@github-actions github-actions Bot added the reviewer: ready Automated review found no actionable items label Sep 11, 2026
@rekram1-node

Copy link
Copy Markdown
Collaborator
Screenshot 2026-09-11 at 10 35 48 AM

These models are served on completions endpoints

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

reviewer: ready Automated review found no actionable items

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants