错误码	错误说明	返回结果	错误信息列表
200	success	OK	-
400	Bad Request	{ "type": "Bad Request", "error": error message，详见错误信息列表 }	"Force P release link failed." "DMI inference requests can be sent only to the prefill node." "DMI request must have req-type, req-id, d-target headers." "The length of req-id cannot exceed 1024." "req-type must be prefill but got {reqType}." "d-target must be an IPv4 address." "d-target should not be itself." "is-recompute is an optional parameter. When it is set, it must be true or false." "Non DMI does not support dresult request." "Only the Decode node supports dresult but you send to " + {pdRole}."
404	NotFoundError	{ "type": "NotFoundError", "error": error message，详见错误信息列表 }	"Stop request id {stopReqId} not found." "Max wait time is {CV_WAIT_TIME}, input is invalid or too long." "Wait time should be in range of [1, {CV_WAIT_TIME}], input is not valid." "Model {modelName} not found."
422`	Input Validation Error	{ "type": "Input Validation Error", "error": error message，详见错误信息列表 }	"Failed to parse context to infer param." "Failed to parse request context to json" "Request contains not model or model null." "Request param model must be string." "Request param model must be not empty." "Model not found." "Tools not array." "Tool param invalid, tool type not exist." "Tool param invalid tool type." "Tool param invalid tool string type." "Tool param not function type." "Tool param invalid, tool function not exist." "Tool param invalid function object." "Function object invalid, function name not exist." "Function object invalid name type." "The name of function must be a-z, A-Z, 0-9, underscores and dashs within max length of 64" "tool_choice string not in [none, auto, required]." "tool_choice not string or object." "Messages MUST be a string for recompute." "Request param contains not messages or messages null." "Request param messages not arr or arr is empty." "Request param contains not role or role null." "Request param contains role must be system, assistant, user or tool." "Request param contains not tool_call_id or tool_call_id null while role is tool." "Request param tool_call_id len should not be 0 while role is tool." "The type of tool_call_id is abnormal." "Tool call param contains no id." "The type of the tool. Currently, only function is supported." "Tool call param contains no function that the model called." "Tool call param contains no function name." "Tool call param function name expected to be a string." "Tool call param contains no function arguments." "Tool call param function arguments expected to be a string." "Request param contains not content or content null." "Request param content len should not be 0." "Request param contains no tool_call_id." "Tools is empty" "Messages parsed not is a array" "Messages len not in (0, {MAX_INPUTS_NUM}], but the length of inputs is {utf16.length()}" "Request param contains not messages or messages null" "Request param messages not arr or arr is empty" "Request param contains not role or role null" "Request param contains role must be system, assistant, user or tool" "Request param contains not content or content null" "Request param content len should not be 0" "Inputs arr is empty." "Request param image_url null." "Request param audio_url null." "Request param video_url null." "The number of multimodal url. should be no more than {MAX_MULTIMODAL_URL_NUM}." "The type of inputs is abnormal." "Failed to get token from input: {msg}." "Unknown error." "Invalid input prompt length {prompt.length()}" "Cant find pid memory index {pid}." "Cant find share memory in encode" "Encode cast buffer header failed." "Tokenizer encode wait sub process timeout." "Invalid output token length " + std::to_string(tokenIdSize) "Encode cast buffer to int64 failed." "Encode memset_s failed." "[InferTokenizer::DownloadUrl] download fail: {e.what()}." "[InferTokenizer::DownloadUrl] get unknown error." "Messages token length must be in(0, {MAX_TOKENS_NUM)], but got {reqTokens.size()}" "req should contain 'inputs' and the type should be array." "Inputs count must be 1." "Cannot find name in inputs or its type is not string." "The length of name set in inputs exceeds 256." "Cannot find shape in inputs or its type is not array" "Shape element must be integer type." "Shape must be at most two-dimensional, and the last element's value must be in (0, {MAX_TOKENS_NUM}]." "Can not find datatype in inputs or its type is not string." "Unsupported datatype, got {typeName}." "Cannot find data in inputs or its type is not array." "Data element must be no more than max token id length="MAX_TOKENS_NUM"." "Data size should be {inputShape[0][1]}, got {body["data"].size()}." "Data element must be integer type." "The id can contain only digits, letters, underscores(_), hyphens(-) and no more than {MAX_INPUT_ID_LENGTH} words in length." "Req should contain 'outputs' and the type should be array." "Outputs size should be {inputDataType.size()}, got {outputs.size()}." "Outputs name must not be null." "Input conflict." "Inputs must not be null." "The type of inputs is abnormal." "Inputs must be necessary and data type must be string and length in (0, {MAX_INPUTS_NUM}], but the length of inputs is {utf16.length()}." "Req should contain 'inputs' and the type should be string." "The text_input not found." "text_input arr is empty." "The type of text_input is abnormal for id = {logId}." "text_input must be necessary and data type must be string and length in (0, {MAX_INPUTS_NUM}], but the length of inputs is {utf16.length()}" "text_input token length must be in (0, {MAX_TOKENS_NUM}], but got {reqTokens.size()}." "Contains not prompt or prompt null." "Prompt arr is empty." "The type of prompt is abnormal." "Prompt must be necessary and data type must be string and length in (0, {MAX_INPUTS_NUM}], but the length of inputs is {utf16.length()}." "Param stream not boolean." "repetition_penalty must be in (0.0, 2.0], got {jsonObj[key]}." "Input validation error: `stop` item count more than {MAX_STOP_STRING_NUM}, but got {stopStrings.size()}" "Input validation error: `stop` must be list[string] if list, and item length in [1, {MAX_STOP_STRING_LEN}] with total <= {MAX_TOTAL_STOP}." "Input validation error: length of `stop` must be in [1, {MAX_STOP_STRING_LEN} but got {GetU16Str(stopStrings).length()}." "Param stop must be string or list[string]." "presence_penalty not float number." "presence_penalty not in [-2.0, 2.0]." "frequency_penalty not number." "frequency_penalty not in [-2.0, 2.0]." "Request param include_stop_str_in_output is not support in dmi mode." "max_tokens not number_integer." "max_tokens must be (0, 2147483647], got {jsonObj[key]}." "temperature must be in [0.0,2.0], got {inputNum}." "top_k must be in [0,2147483647], got {jsonObj[key]}." "stop_token_ids not array." "stop_token_ids not list[int]." "top must be float." "top_p must be in (0.0, 1.0], got {jsonObj[key]}." "Check open ai req parameter error" "typical_p must be in (0.0,1.0], got {jsonObj[key]}." "Check self develop req parameter error." "truncate must be in (0,2147483647], got {body[key]}." "The decoder_input_details must not be true when stream is true." "The format of adapter_id is invalid." "Check tgi req parameter error." "firstTokenCost exceeds the max value of size_t type." "The decodeTime length is invalid." "The decodeTime exceeds the max value of size_t type." "Check triton text req parameter error" "repetition_penalty must be strictly positive, got {jsonObj[key]}." "seed must be in (0, 18446744073709551615], got {jsonObj[key]}." "temperature must be in (0.0, {MAX_FLOAT_VALUE}], got {jsonObj[key]}." "top_k must be in [0,2147483647], got {jsonObj[key]}." "top_p must be in (0.0,1.0], got {jsonObj[key]}." "batch_size must be in (0,{MAX_INT32_VALUE}], got {jsonObj[key]}." "max_new_tokens must be in (0, 2147483647], got {jsonObj[key]}." "priority must be in [1,5], got {jsonObj[key]}." "timeout must be strictly positive and small than 1 hour, got {jsonObj[key]}." "Check triton token req parameter error." "seed is not number unsigned." "Input validation error: param stop must be string or list[string]." "model must be string type." "repetition_penalty must be in (0.0,2.0], got {jsonObj[key]}." "Check vllm req parameter error." "Failed to parse context to json body." "Not found {key}." "key must not be null." "key must be {type} type." "The id no more than {MAX_INPUT_ID_LENGTH} words in length." "Stop request id = {stopReqId} is invalid." "Req body converts to json fail. Reset to previous node status." "Error: `inputs` or `prompt` must be necessary and data type must be string. Additionally, the request body must be valid json." "reqType does not match type InferReqType." "stream must be boolean type." "Req body converts to json fail." "Inputs in request body is invalid." "Inputs must be necessary and data type must be string and length in [0, {MAX_INPUTS_NUM}], but the length of inputs is {utf16.length()}." "TikToken process fail." "Failed to check the input text. Can't convert string to UTF-16."
424	Generation Error	{ "type": "Generation Error", "error": error message，详见错误信息列表 }	"Failed to generate infer request." "Failed to enqueue inferRequest: remainBlocks is nullptr." "Failed to enqueue inferRequest: remainPrefillSlots is nullptr." "Failed to enqueue inferRequest: remainPrefill is nullptr."
500	Incomplete Generation	{ "type": "Incomplete Generation", "error": error message，详见错误信息列表 }	"Failed to generate simulate infer request." "Stop request failed, internal failed." "Failed to enqueue inferRequest: backend manager has not been initialized."; "Failed to enqueue inferRequest: engine has not been initialized." "Failed to enqueue inferRequest: Model instance has been finalized or not initialized." "Failed to enqueue inferRequest: llmInferEngine is not initialized!" "Failed forward in for infer engine." "Engine callback timeout." "Failed to get engine response." "Health status changed during health detector." "[P Node] Failed to get tokenIds from inferResponse." "[P Node] Failed to decode tokenIds to respStr." "[P Node] Send D request failed!" "[P Node] Unknown transfer flag." "Role status is invalid." "Environment variable MIES_SERVICE_MONITOR_MODE is not set." "Failed to get health status." "No contact node detected."
503	Service Unavailable	{ "type": "Service Unavailable", "error": error message，详见错误信息列表 }	"The service has been stopped." "Parse req json failed. Reset to previous node status." "The server cannot process the inference request due to an unknown status."

RESTful响应状态码