chenxiao5580-cmd
d85acd6083
* feat(provider): add Modelis OpenAI-compatible LLM gateway. One key across Claude, Gemini, DeepSeek and Qwen coding models. Disclosure: I maintain Modelis. * fix(modelis): declare reasoning_effort options, verified against the live endpoint rekram1-node was right to push back on reasoning_options = []. That was 'unverified', not 'verified absent'. Tested every listed model against https://modelishub.com/v1 : all nine accept reasoning_effort with all six values (minimal/low/medium/high/xhigh/max), and usage.completion_tokens_details.reasoning_tokens moves with the setting. An invalid value is rejected with the enum echoed back. Declared the option on all nine, with the exact API syntax as a header comment.
6 lines
150 B
TOML
6 lines
150 B
TOML
name = "Modelis"
|
|
npm = "@ai-sdk/openai-compatible"
|
|
env = ["MODELIS_API_KEY"]
|
|
api = "https://modelishub.com/v1"
|
|
doc = "https://modelishub.com/pricing"
|