POST https://api.naga.ac/v1/responses takes the OpenAI Responses format and serves every chat model in the catalog, whichever company made it. Point the OpenAI SDK at https://api.naga.ac/v1 and call client.responses.create as you would against OpenAI.
Read the response
output_text is a helper of the SDK. In raw JSON the text is in the output array, in the message item’s output_text part. output can also hold reasoning, function_call, web_search_call and image_generation_call items, so find the item by type rather than by position.
Conversation state
NagaAI stores no responses. It ignoresprevious_response_id, store, conversation, background and prompt, and there is no GET /v1/responses/{id}. Send the full conversation in input on every call. See Conversation state.
NagaAI additions
routingin the body and thex-naga-routing-strategyheader, see Routing.models, the fallback models to try whenmodelcan’t answer, see Fallback models.max_uses,allowed_domainsandblocked_domainson theweb_searchtool, see Web search.cache_controlin OpenRouter’s shape, at the top level and on aninput_textpart, next to OpenAI’sprompt_cache_breakpointandprompt_cache_options. It caches the prompt at Claude providers too, see Prompt caching.- NagaAI accepts chat-style tools,
{"type": "function", "function": {...}}, next to the flat Responses shape.
Ignored fields
NagaAI accepts and dropsinclude, metadata, user, top_logprobs, max_tool_calls and stream_options. It also drops server-side tools other than web search. service_tier: "flex" keeps the request at the model’s base price, see Limit the price. More differences are in Differences from OpenAI and Anthropic.