Create a message
Anthropic-compatible Messages with Raon-run search and client-executed tools; named tool choice, vendor files, containers and other hosted tools are refused.
AuthorizationBearer <token>Raon API key. A Bearer credential takes precedence when both authentication headers are sent.
anthropic-version?stringAnthropic compatibility header commonly sent by SDKs. Raon accepts it but does not validate it or forward it to the serving backend.
application/json- body
Anthropic-compatible Messages request. Organization safety text is prepended to system. web_fetch_, code_execution_, mcp_toolset and other hosted tools are refused with unsupported_tool, as are nonempty mcp_servers arrays. Uploaded file sources, container_upload blocks and nonempty container are refused with invalid_request.
model*stringPublic model ID from the model catalog.
1 <= lengthmessages*array<>max_tokens*integerRequired positive output token cap.
1 <= valuesystem?string|array<>stream?booleanReturn Messages SSE events.
falsetemperature?|Sampling temperature. Raon models always use temperature=1.0, regardless of supplied values, null or omission. Other providers use their own parameter behavior.
0 <= value <= 1top_p?|Nucleus sampling probability. Raon models always use top_p=0.95, regardless of supplied values, null or omission. Other providers use their own parameter behavior.
0 <= value <= 1min_p?|Minimum relative token probability. Raon models always use min_p=0, regardless of supplied values, null or omission. Other providers use their own parameter behavior.
presence_penalty?|Topic repetition penalty. Raon models always use presence_penalty=1.5, regardless of supplied values, null or omission. Other providers use their own parameter behavior.
frequency_penalty?|Token frequency penalty. Raon models always use frequency_penalty=0, regardless of supplied values, null or omission. Other providers use their own parameter behavior.
repetition_penalty?|Token repetition penalty. Raon models always use repetition_penalty=1.0, regardless of supplied values, null or omission. Other providers use their own parameter behavior.
top_k?|Top-k sampling cutoff. Raon models always use top_k=20, regardless of supplied values, null or omission. Other providers use their own parameter behavior.
0 <= valuestop_sequences?array<string>tools?array<||>tool_choice?Named selection (type tool) is refused with 400 because the backend does not reliably return tool_use stop reasons for it.
thinking?||Accepted but ignored on Raon, including type and budget_tokens. Reasoning is always enabled.
output_config?Output configuration forwarded to the serving backend. JSON Schema format constrains the answer, while reasoning remains separate. Supported schema constraints depend on the backend's structured decoding implementation.
chat_template_kwargs?Serving template options. Raon always enables thinking on all three inference APIs.
metadata?Generated output as JSON, or server-sent events when stream is true. Additional backend fields may be returned.
- response
Raon web search emits server_tool_use and web_search_tool_result blocks and reports the usage of its last internal turn, which already holds the whole conversation and every search result; usage.server_tool_use.web_search_requests counts the searches.
id*stringtype*stringrole*stringmodel*stringModel ID returned by the serving backend, preserved by the gateway. It can differ from the public ID in the request. Use GET /v1/models IDs for subsequent requests.
1 <= lengthcontent*array<|||||>stop_reason*|nullstop_sequence*string|nullusage*curl -X POST 'https://raon-api.krafton.ai/v1/messages' \ -H 'Content-Type: application/json' \ -d '{ "model": "raon-1-mini", "max_tokens": 1024, "messages": [ { "role": "user", "content": "Explain why the sky is blue." } ]}'{ "id": "string", "type": "message", "role": "assistant", "model": "raon", "content": [ { "type": "text", "text": "string", "cache_control": { "type": "ephemeral", "ttl": "5m" }, "citations": [ {} ] } ], "stop_reason": "end_turn", "stop_sequence": "string", "usage": { "input_tokens": 0, "output_tokens": 0, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0, "server_tool_use": { "web_search_requests": 0 } }}Create a response POST
Stateless OpenAI-compatible Responses with store forced off and Raon-run web search; vendor storage and other hosted tools are refused.
List models GET
OpenAI-compatible model list scoped to your organization, extended with display name, modality, context window and decimal pricing; upstream provider details are hidden.