feat(provider): add QwenCloud International

[navegantesjr]

Summary

Adds the QwenCloud International provider (anomalyco/opencode#43067).

• provider.toml: OpenAI-compatible provider at the DashScope International endpoint https://dashscope-intl.aliyuncs.com/compatible-mode/v1, auth via QWENCLOUD_API_KEY (falls back to DASHSCOPE_API_KEY).
• Models: qwen3.8-max, qwen3.7-plus, qwen3.7-flash (thin wrappers reusing the existing alibaba base models) and qwen3.6-omni-plus (authored from scratch).
• bun validate passes; the compiled catalog resolves qwencloud to @ai-sdk/openai-compatible with the international base URL.

Test plan

• bun install && bun validate (exit 0)
• Reviewer: confirm the 4 models surface under QwenCloud in the opencode provider picker once this catalog is deployed.

🤖 Generated with opencode

[github-actions]

Action items

• [critical] [violation] providers/qwencloud/ - Check: New provider must include a compliant logo.svg. Why: The PR adds provider qwencloud with provider.toml and models but no providers/qwencloud/logo.svg. Missing logos are a merge blocker. Action: Add providers/qwencloud/logo.svg using currentColor, no fixed width/height or hardcoded colors, preferably a square viewBox.
• [high] [violation] providers/qwencloud/models/qwen3.6-omni-plus.toml:1 - Check: Non-lab hosts must use base_model; nameable lab models belong under models/. Why: qwen3.6-omni-plus is a full inline Alibaba/Qwen definition on a separate provider ID, with no base_model and no models/alibaba/… entry. That violates override-only hosting except unique-to-host cases. Action: Add a complete models/alibaba/qwen3.6-omni-plus.toml (required lab fields) and rewrite the provider file as base_model plus only host deltas (cost, reasoning_options, real overrides).
• [high] [violation] providers/qwencloud/models/qwen3.8-max.toml:4 - Check: Every toggle needs a leading top-of-file wire-path comment. Why: reasoning_options includes { type = "toggle" } with no header documenting the request field (lab peers use # Toggle: enable_thinking true|false plus effort/budget wire notes). Action: Add a leading comment block with the exact DashScope/QwenCloud chat wire paths for toggle, effort, and budget before the first key.
• [high] [violation] providers/qwencloud/models/qwen3.7-plus.toml:3 - Check: Every toggle needs a leading top-of-file wire-path comment. Why: File starts with base_model then [[reasoning_options]] type = "toggle" with no wire comment. Action: Add a leading # Toggle: … / # Budget: … comment matching this host’s OpenAI-compatible surface.
• [high] [violation] providers/qwencloud/models/qwen3.7-flash.toml:3 - Check: Every toggle needs a leading top-of-file wire-path comment. Why: Same missing toggle wire documentation as the other hybrid Qwen entries. Action: Add a leading wire-path comment for enable_thinking / thinking_budget (or the actual fields this endpoint accepts).
• [high] [violation] providers/qwencloud/models/qwen3.6-omni-plus.toml:12 - Check: Every toggle needs a leading top-of-file wire-path comment. Why: Toggle is authored with no top-of-file wire path; the only comment is a mid-file pricing note. Action: Move sources/pricing notes above the first key and document the toggle (and budget, if kept) wire syntax there.
• [medium] [possible mistake] providers/qwencloud/models/qwen3.6-omni-plus.toml:14 - Check: budget_tokens only when this host exposes a real reasoning-budget field. Why: First-party omni peers such as providers/alibaba/models/qwen3-omni-flash.toml use toggle only, not budget_tokens. Adding budget here without host/lab evidence looks invented. Action: Drop budget_tokens unless QwenCloud/DashScope docs (or a same-surface peer) confirm thinking_budget for this model; if confirmed, cite it in the leading comment.
• [low] [possible mistake] providers/qwencloud/models/qwen3.7-flash.toml:7 - Check: Material cost data should cite first-party pricing. Why: Comments say international flash pricing is “approximate” / unconfirmed, and the PR body has no mapped pricing sources. Action: Replace approximate costs with documented DashScope/QwenCloud International USD rates and cite what each URL supports, or mark status/omit public cost until rates are confirmed.
• [low] [possible mistake] providers/qwencloud/models/qwen3.6-omni-plus.toml:17 - Check: Material cost/limit claims need reviewable evidence. Why: File labels omni pricing/limits as best-effort and unconfirmed, so catalog consumers may get wrong USD/MTok and window values. Action: Cite first-party international pricing and limit docs for input/output/audio rates and context/output caps, or withhold the model until those are known.