Loading
feat(models): add GPT-6 Sol and GPT-6 Luna model mappings
What does this MR do?
Adds agentic chat mappings for the two GPT-6 models that the upstream AI Gateway now serves but this package doesn't map yet: GPT-6 Sol and GPT-6 Luna.
Source values (ai-assist models.yml)
| Name | gitlab_identifier |
params.model |
Context | Max output | Temperature |
|---|---|---|---|---|---|
| GPT-6 Sol | gpt_6_sol |
gpt-6-sol |
1,050,000 | 128,000 | not set |
| GPT-6 Luna | gpt_6_luna |
gpt-6-luna |
1,050,000 | 128,000 | not set |
Changes
src/model-mappings.ts:duo-chat-gpt-6-solmaps togpt-6-sol.duo-chat-gpt-6-lunamaps togpt-6-luna.- Both use
openaiApiType: 'responses', like GPT-6 Astra and GPT-5.6. The existing comment is widened to cover all GPT-6 models. - As with other OpenAI models, there is no
duo-workflow-*entry.
tests/model-mappings.test.ts:- New mapping tests for both models.
- Coverage in the provider, OpenAI-model, API-type and Responses-API checks.
- Counts updated: total 36 → 38, OpenAI 14 → 16,
getValidModelsForProvider('openai')14 → 16.
README.md: two rows added to the "Available Model Mappings" table, after GPT-6 Astra.dist/: rebuilt.
Not added on purpose: the upstream alias entries gpt-5 (gpt_5_alias) and gpt-5.1 (gpt_5_1_alias), and the generic gpt placeholder. None of these have been mapped before; the dated models they point to are already covered.
How to validate
npm run type-checkis clean.npx vitest run tests/model-mappings.test.ts: 63 passed.npm run lintis clean.npm test: 700 passed.
Not part of this MR
The models.dev registration (the model picker entries) and the opencode dependency bump come separately. The opencode bump waits for this release to be published.