Working config (OpenAI vendor, origin host)
big-AGI already speaks OpenAI-compat through the OpenAI vendor. A Chat Completions host whose public catalog is GET https://api.pzero.studio/v1/models (no key) fits that slot.
Models → add an OpenAI service (not the DeepSeek / Groq / Azure vendor rows). Advanced → API Host:
- API Host:
https://api.pzero.studio (origin only)
- API Key: a Bearer key for that host
- Model: a live catalog id, example
deepseek-v4-flash
The UI strips a trailing /v1, then the server appends OPENAI_API_PATHS (/v1/models, /v1/chat/completions). Pasting /v1 is therefore harmless. Pasting /v1/chat/completions is not: strip only removes a trailing /v1, so the request becomes /v1/chat/completions/v1/chat/completions and 404s.
Do not enable image / audio / moderation paths on this host. Chat only.
Not asking for a first-party dialect or autocomplete row. The OpenAI vendor plus a custom host is enough.
Working config (OpenAI vendor, origin host)
big-AGI already speaks OpenAI-compat through the OpenAI vendor. A Chat Completions host whose public catalog is
GET https://api.pzero.studio/v1/models(no key) fits that slot.Models → add an OpenAI service (not the DeepSeek / Groq / Azure vendor rows). Advanced → API Host:
https://api.pzero.studio(origin only)deepseek-v4-flashThe UI strips a trailing
/v1, then the server appendsOPENAI_API_PATHS(/v1/models,/v1/chat/completions). Pasting/v1is therefore harmless. Pasting/v1/chat/completionsis not: strip only removes a trailing/v1, so the request becomes/v1/chat/completions/v1/chat/completionsand 404s.Do not enable image / audio / moderation paths on this host. Chat only.
Not asking for a first-party dialect or autocomplete row. The OpenAI vendor plus a custom host is enough.