Chat Completions

OpenAI-compatible chat completions, routed through OpenRouter. The request body is OpenRouter’s (a superset of OpenAI’s) and is forwarded as-is — see OpenRouter’s API reference for every field. Set stream: true for Server-Sent Events.

string

required

Model id with provider prefix, e.g. anthropic/claude-sonnet-4.6, openai/gpt-5.2.

array

OpenAI-style chat messages: [{ role, content }]. Either messages or prompt is required.

string

Plain-text prompt (alternative to messages).

string[]

Optional fallback chain — OpenRouter tries these in order if earlier models are unavailable.

object

OpenRouter provider routing preferences (order, only, ignore, sort, allow_fallbacks, data_collection, …).

boolean

default:"false"

Stream the response as SSE. The final chunk carries the usage (incl. cost).

number

Sampling temperature (0–2). top_p, max_tokens, stop, seed, tools, tool_choice, response_format, etc. are also forwarded.

curl -X POST https://api.usenaive.ai/v1/llm/chat/completions \
  -H "Authorization: Bearer nv_sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-4.6",
    "messages": [{ "role": "user", "content": "Summarize REST in one line." }],
    "models": ["anthropic/claude-sonnet-4.6", "openai/gpt-5.2"],
    "provider": { "sort": "throughput" }
  }'

{
  "id": "gen-xxxxxxxx",
  "object": "chat.completion",
  "created": 1749600000,
  "model": "anthropic/claude-sonnet-4.6",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": { "role": "assistant", "content": "REST is a stateless, resource-oriented HTTP API style." }
    }
  ],
  "usage": { "prompt_tokens": 14, "completion_tokens": 16, "total_tokens": 30, "cost": 0.00012 },
  "credits_used": 0.003,
  "credits_remaining": 999
}

Detecting truncated output. When a completion hits the token cap, the output is silently cut off and choices[0].finish_reason is "length". In that case the response also carries a top-level "truncated": true flag — check it (or finish_reason) before treating the output as complete, and re-request with a higher max_tokens if needed. The flag is absent (not false) when the generation finished cleanly. The naive llm chat CLI additionally prints a warning to stderr.

Cost: per-token, derived from OpenRouter’s returned usage.cost × markup (see Credits). Charged after completion (after the final SSE chunk for streams).

Prefer to keep your existing OpenAI/OpenRouter SDK? Point its baseURL at https://api.usenaive.ai/v1/proxy/openrouter and authenticate with your Naive api key — a transparent, drop-in passthrough. See Use Naive instead of OpenRouter.

curl -X POST https://api.usenaive.ai/v1/llm/chat/completions \ -H "Authorization: Bearer nv_sk_live_..." \ -H "Content-Type: application/json" \ -d '{ "model": "anthropic/claude-sonnet-4.6", "messages": [{ "role": "user", "content": "Summarize REST in one line." }], "models": ["anthropic/claude-sonnet-4.6", "openai/gpt-5.2"], "provider": { "sort": "throughput" } }'

{ "id": "gen-xxxxxxxx", "object": "chat.completion", "created": 1749600000, "model": "anthropic/claude-sonnet-4.6", "choices": [ { "index": 0, "finish_reason": "stop", "message": { "role": "assistant", "content": "REST is a stateless, resource-oriented HTTP API style." } } ], "usage": { "prompt_tokens": 14, "completion_tokens": 16, "total_tokens": 30, "cost": 0.00012 }, "credits_used": 0.003, "credits_remaining": 999 }