技能对话接口，API多轮对话的实现

首页版块访问AI主站注册发帖

技能对话接口，API多轮对话的实现

精

半醉半醒半疯Z 发布于2018-07 浏览:54293 回复:27

技能对话接口，API多轮对话的实现

快速回复

最后编辑于2021-12

补充：client_session参数可以不写，省事。

使用API对话接口时，想要实现多轮对话，需要传递bot_session或session_id的信息。

第一轮对话bot_session需要置空，置空的写法是"bot_session": ""。从第二轮开始，多轮对话的每轮都需将上轮返回结果里的bot_session（或session_id）的值完整不动的传给本轮的bot_session。

1.传bot_session的值

返回数据里bot_session的形式如下：

"bot_session": "{\"bot_id\":\"1201\",......\"timestamp\":\" request time \"}],\"session_id\":\" value \"}\n"

可以将bot_session的值（整个双引号里的内容）完整不做任何修改的传入请求参数的bot_session里。也可以去掉最后的 \n 。其他部分不需要做任何修改。

2.传session_id的值

如果传的是session_id，则请求参数形式是："bot_session": "{\"session_id\":\" value \"}"

返回数据里bot_session的形式如下：

"bot_session": "{\"bot_id\":\"1201\",......,\"session_id\":\" value \"}\n"

可以直接将 \"session_id\":\"value\" 取出，使用 {} 包起来，然后再传入入参的bot_session里。

格式： "bot_session": "{\"session_id\":\"value\"}"

另外：DMKIT如果进行多轮对话，需要传入整个bot_session的内容，不能只传session_id。

智能客服

个赞

共27条回复最后由路人都市报回复于2021-12

#28路人都市报回复于2021-12

String params =
"{\"bot_session\":\"{\\\"session_id\\\":\\\" " + session_id + "\\\"}\"," +
"\"log_id\":\"" + 1106648 + "\"" +
",\"request\":" +
"{\"bernard_level\":1" +
",\"client_session\":\"" +
"{\\\"client_results\\\":\\\"\\\", " +
"\\\"candidate_options\\\":[]}\"," +
"\"query\":\"" + content + "\"," +
"\"query_info\":" +
"{\"asr_candidates\":[]," +
"\"source\":\"ASR\"," +
"\"type\":\"TEXT\"}," +
"\"updates\":\"\"," +
"\"user_id\":\"" + 88888 + "\"}," +
"\"bot_id\":\"S54982\"}," +
"\"version\":\"2.0\"}";

报错：ult":{"ref_id":"q9cn6_20211214105107_4257016938"},"error_code":282008,"error_msg":"The request content type is illegal."}

#271154867045回复于2020-07

传入上轮的bot_session后出现类型非法。。。。

不知道该怎么该了

{'error_code': 282008, 'error_msg': 'The request content type is illegal.'}

bot_session = response.json()['result']['bot_session']
print(bot_session)
post_data = "{\"bot_session\":"+bot_session+",\"log_id\":\"7758521\",\"request\":{\"bernard_level\":1,\"client_session\":\"{\\\"client_results\\\":\\\"\\\", \\\"candidate_options\\\":[]}\",\"query\":\""+query+"\",\"query_info\":{\"asr_candidates\":[],\"source\":\"KEYBOARD\",\"type\":\"TEXT\"},\"updates\":\"\",\"user_id\":\"88888\"},\"bot_id\":\"1035476\",\"version\":\"2.0\"}"
response = requests.post(url, data=post_data.encode('utf-8'), headers=headers)
print(response.json())

#26wangwei8638回复于2020-06

实现多轮对话更适合与人的交流

#25用户已被禁言回复于2019-07

干活

#24rose20135188回复于2019-01

谢谢无私的分享。

#23zhaojiang1990回复于2019-01

对#21 mr_hanhai回复

已经解决

能把多轮调用这部分代码分享么，谢谢啦

#22lu9944y回复于2019-01

简单易懂，学会了，谢谢

#21mr_hanhai回复于2018-12

已经解决

#20inderzhu回复于2018-11

对#19 荒墨丶迷失回复

按道理不会出现这个问题的呀，除非sessionID一样了

请求百度ai把上次对话的bot_session也传进去可以达到多轮对话的实现，这个是没问题的。但我遇到一个问题：我用电脑A访问UI界面和机器人对话是没问题，但当我又用另一台电脑B访问UI界面和机器人对话，此时B的对话会出现A的对话回答缓存，这是为什么？我的后台程序做了session用户的区分，当是新用户的时候把bot_session给置空，相当于访问后台程序bot_session第一次传空参，以后bot_session就传上次对话的bot_session，这样做还是不行，他们的对话回答还是会串在一起，大神能指点指点么？告诉我怎么解决！！

#19荒墨丶迷失回复于2018-11

对#18 inderzhu回复

我传sessionid保存上下文的对话，为什么会出现A用户和B用户同时访问出现回复混乱的样子？？求解答

展开

按道理不会出现这个问题的呀，除非sessionID一样了

#18inderzhu回复于2018-11

我传sessionid保存上下文的对话，为什么会出现A用户和B用户同时访问出现回复混乱的样子？？求解答

#17love卵疼回复于2018-11

这是我一次对话百度json返回的bot_session内容，我不确定，在第二轮对话的时候是否需要把这么多再次传回去（如果我语文学的没错的情况下，楼主应该就是这个意思）

{{"bot_id":"12968","bot_views":{"bernard_res":[{"action_list":[{"action_id":"","confidence":0.0,"custom_reply":"","refine_detail":{"clarify_reason":"","interact":"","option_list":[]},"say":"","type":"understood"}],"msg":"ok","qu_res":{"candidates":[{"confidence":100.0,"domain_confidence":0.0,"extra_info":{"group_id":"5","real_threshold":"1","threshold":"0.7"},"from_who":"pow-slu-lev1","intent":"QUERYPOSITION","intent_confidence":100.0,"intent_need_clarify":false,"match_info":"{\"group_id\":\"5\",\"id\":\"739186\",\"match_keywords\":\" kw_querypositionyuqi:在哪里\",\"match_pattern\":\"[D:user_queryposition]\\t[D:kw_querypositionyuqi]\",\"ori_pattern\":\"[D:user_queryposition]\\t[D:kw_querypositionyuqi]\",\"ori_slots\":{\"confidence\":100.0,\"domain_confidence\":0.0,\"extra_info\":{},\"from_who\":\"smart_qu\",\"intent\":\"QUERYPOSITION\",\"intent_confidence\":100.0,\"intent_need_clarify\":false,\"match_info\":\"[D:user_queryposition] \\t[D:kw_querypositionyuqi] kw_querypositionyuqi:在哪里\",\"slots\":[{\"begin\":0,\"confidence\":100.0,\"father_idx\":-1,\"length\":4,\"name\":\"user_queryposition\",\"need_clarify\":false,\"normalized_word\":\"\",\"original_word\":\"电梯\",\"word_type\":\"\"},{\"begin\":4,\"confidence\":100.0,\"father_idx\":-1,\"length\":6,\"name\":\"kw_querypositionyuqi\",\"need_clarify\":false,\"normalized_word\":\"\",\"original_word\":\"在哪里\",\"word_type\":\"\"}]},\"real_threshold\":1.0,\"threshold\":0.6999999880790710}","slots":[{"begin":0,"confidence":100.0,"father_idx":-1,"length":4,"name":"user_queryposition","need_clarify":false,"normalized_word":"电梯","original_word":"电梯","word_type":""}]}],"lexical_analysis":[{"basic_word":["电梯"],"etypes":["[D:user_queryposition]"],"term":"电梯","type":"user_queryposition","weight":0.6080},{"basic_word":["在"],"etypes":[],"term":"在","type":"28","weight":0.130},{"basic_word":["哪里"],"etypes":["[D:COREF_LOCATION]","sys_coref_location","sys_coref_location"],"term":"哪里","type":"sys_coref_location","weight":0.260}],"qu_res_chosen":"{\"confidence\":100.0,\"domain_confidence\":0.0,\"extra_info\":{\"group_id\":\"5\",\"real_threshold\":\"1\",\"threshold\":\"0.7\"},\"from_who\":\"pow-slu-lev1\",\"intent\":\"QUERYPOSITION\",\"intent_confidence\":100.0,\"intent_need_clarify\":false,\"match_info\":\"{\\\"group_id\\\":\\\"5\\\",\\\"id\\\":\\\"739186\\\",\\\"match_keywords\\\":\\\" kw_querypositionyuqi:在哪里\\\",\\\"match_pattern\\\":\\\"[D:user_queryposition]\\\\t[D:kw_querypositionyuqi]\\\",\\\"ori_pattern\\\":\\\"[D:user_queryposition]\\\\t[D:kw_querypositionyuqi]\\\",\\\"ori_slots\\\":{\\\"confidence\\\":100.0,\\\"domain_confidence\\\":0.0,\\\"extra_info\\\":{},\\\"from_who\\\":\\\"smart_qu\\\",\\\"intent\\\":\\\"QUERYPOSITION\\\",\\\"intent_confidence\\\":100.0,\\\"intent_need_clarify\\\":false,\\\"match_info\\\":\\\"[D:user_queryposition] \\\\t[D:kw_querypositionyuqi] kw_querypositionyuqi:在哪里\\\",\\\"slots\\\":[{\\\"begin\\\":0,\\\"confidence\\\":100.0,\\\"father_idx\\\":-1,\\\"length\\\":4,\\\"name\\\":\\\"user_queryposition\\\",\\\"need_clarify\\\":false,\\\"normalized_word\\\":\\\"\\\",\\\"original_word\\\":\\\"电梯\\\",\\\"word_type\\\":\\\"\\\"},{\\\"begin\\\":4,\\\"confidence\\\":100.0,\\\"father_idx\\\":-1,\\\"length\\\":6,\\\"name\\\":\\\"kw_querypositionyuqi\\\",\\\"need_clarify\\\":false,\\\"normalized_word\\\":\\\"\\\",\\\"original_word\\\":\\\"在哪里\\\",\\\"word_type\\\":\\\"\\\"}]},\\\"real_threshold\\\":1.0,\\\"threshold\\\":0.6999999880790710}\",\"slots\":[{\"begin\":0,\"confidence\":100.0,\"father_idx\":-1,\"length\":4,\"name\":\"user_queryposition\",\"need_clarify\":false,\"normalized_word\":\"电梯\",\"original_word\":\"电梯\",\"word_type\":\"\"}]}\n","raw_query":"电梯在哪里","sentiment_analysis":{"label":"1","pval":0.9960},"status":0,"timestamp":0},"schema":{"domain_confidence":0.0,"intent":"QUERYPOSITION","intent_confidence":100.0,"slots":[{"begin":0,"confidence":100.0,"length":4,"merge_method":"updated","name":"user_queryposition","normalized_word":"电梯","original_word":"电梯","session_offset":0,"sub_slots":[],"word_type":""}]},"status":0}],"bernard_status":[{"index":0,"step":"AFTER_DM_TRIGGER"}],"intervention":{"interv_qu_res":"","interv_query":"","qu_res_interved":"","qu_res_original":"","query_original":"","type":"","user_id":""},"user_slots":{"user_queryposition":{"state":2,"static_slot":{"default_state":0,"default_tag_name":"DEFAULT","extensible":true,"name":"user_queryposition","print":["区域位置"],"type":0,"update_type":"single_turn","weight":1.0},"tag_map":{"电梯":{"begin":0,"confidence":100.0,"length":4,"merge_method":"updated","normalized_name":"电梯","original_name":"电梯","state":2,"turn":0,"word_type":""}}}}},"dialog_state":{"contexts":{},"intents":[{"index":0,"name":"QUERYPOSITION"}],"user_slots":{"user_queryposition":{"attrs":{"default_state":0,"default_tag_name":"DEFAULT","extensible":true,"name":"user_queryposition","print":["区域位置"],"type":0,"update_type":"single_turn","weight":1.0},"state":2,"values":{"电梯":{"begin":0,"confidence":100.0,"length":4,"merge_method":"updated","normalized_name":"电梯","original_name":"电梯","session_offest":0,"state":2,"word_type":""}}}}},"interactions":[{"interaction_id":"interaction-1541226572696-746135415-8013-497","request":{"bernard_level":1,"client_session":"{\"client_results\":\"\", \"candidate_options\":[]}","hyper_params":{"slu_tags":[]},"query":"电梯在哪里","query_info":{"asr_candidates":[],"source":"KEYBOARD","type":"TEXT"},"updates":"","user_id":"99635"},"response":{"action_list":[{"action_id":"queryposition_satisfy","confidence":100.0,"custom_reply":"","refine_detail":{"clarify_reason":"","interact":"","option_list":[]},"say":"结果请查看地图","type":"satisfy"}],"msg":"ok","qu_res":{"candidates":[{"confidence":100.0,"domain_confidence":0.0,"extra_info":{"group_id":"5","real_threshold":"1","threshold":"0.7"},"from_who":"pow-slu-lev1","intent":"QUERYPOSITION","intent_confidence":100.0,"intent_need_clarify":false,"match_info":"{\"group_id\":\"5\",\"id\":\"739186\",\"match_keywords\":\" kw_querypositionyuqi:在哪里\",\"match_pattern\":\"[D:user_queryposition]\\t[D:kw_querypositionyuqi]\",\"ori_pattern\":\"[D:user_queryposition]\\t[D:kw_querypositionyuqi]\",\"ori_slots\":{\"confidence\":100.0,\"domain_confidence\":0.0,\"extra_info\":{},\"from_who\":\"smart_qu\",\"intent\":\"QUERYPOSITION\",\"intent_confidence\":100.0,\"intent_need_clarify\":false,\"match_info\":\"[D:user_queryposition] \\t[D:kw_querypositionyuqi] kw_querypositionyuqi:在哪里\",\"slots\":[{\"begin\":0,\"confidence\":100.0,\"father_idx\":-1,\"length\":4,\"name\":\"user_queryposition\",\"need_clarify\":false,\"normalized_word\":\"\",\"original_word\":\"电梯\",\"word_type\":\"\"},{\"begin\":4,\"confidence\":100.0,\"father_idx\":-1,\"length\":6,\"name\":\"kw_querypositionyuqi\",\"need_clarify\":false,\"normalized_word\":\"\",\"original_word\":\"在哪里\",\"word_type\":\"\"}]},\"real_threshold\":1.0,\"threshold\":0.6999999880790710}","slots":[{"begin":0,"confidence":100.0,"father_idx":-1,"length":2,"name":"user_queryposition","need_clarify":false,"normalized_word":"电梯","original_word":"电梯","word_type":""}]}],"lexical_analysis":[{"basic_word":["电梯"],"etypes":["[D:user_queryposition]"],"term":"电梯","type":"user_queryposition","weight":0.6080},{"basic_word":["在"],"etypes":[],"term":"在","type":"28","weight":0.130},{"basic_word":["哪里"],"etypes":["[D:COREF_LOCATION]","sys_coref_location","sys_coref_location"],"term":"哪里","type":"sys_coref_location","weight":0.260}],"qu_res_chosen":"{\"confidence\":100.0,\"domain_confidence\":0.0,\"extra_info\":{\"group_id\":\"5\",\"real_threshold\":\"1\",\"threshold\":\"0.7\"},\"from_who\":\"pow-slu-lev1\",\"intent\":\"QUERYPOSITION\",\"intent_confidence\":100.0,\"intent_need_clarify\":false,\"match_info\":\"{\\\"group_id\\\":\\\"5\\\",\\\"id\\\":\\\"739186\\\",\\\"match_keywords\\\":\\\" kw_querypositionyuqi:在哪里\\\",\\\"match_pattern\\\":\\\"[D:user_queryposition]\\\\t[D:kw_querypositionyuqi]\\\",\\\"ori_pattern\\\":\\\"[D:user_queryposition]\\\\t[D:kw_querypositionyuqi]\\\",\\\"ori_slots\\\":{\\\"confidence\\\":100.0,\\\"domain_confidence\\\":0.0,\\\"extra_info\\\":{},\\\"from_who\\\":\\\"smart_qu\\\",\\\"intent\\\":\\\"QUERYPOSITION\\\",\\\"intent_confidence\\\":100.0,\\\"intent_need_clarify\\\":false,\\\"match_info\\\":\\\"[D:user_queryposition] \\\\t[D:kw_querypositionyuqi] kw_querypositionyuqi:在哪里\\\",\\\"slots\\\":[{\\\"begin\\\":0,\\\"confidence\\\":100.0,\\\"father_idx\\\":-1,\\\"length\\\":4,\\\"name\\\":\\\"user_queryposition\\\",\\\"need_clarify\\\":false,\\\"normalized_word\\\":\\\"\\\",\\\"original_word\\\":\\\"电梯\\\",\\\"word_type\\\":\\\"\\\"},{\\\"begin\\\":4,\\\"confidence\\\":100.0,\\\"father_idx\\\":-1,\\\"length\\\":6,\\\"name\\\":\\\"kw_querypositionyuqi\\\",\\\"need_clarify\\\":false,\\\"normalized_word\\\":\\\"\\\",\\\"original_word\\\":\\\"在哪里\\\",\\\"word_type\\\":\\\"\\\"}]},\\\"real_threshold\\\":1.0,\\\"threshold\\\":0.6999999880790710}\",\"slots\":[{\"begin\":0,\"confidence\":100.0,\"father_idx\":-1,\"length\":4,\"name\":\"user_queryposition\",\"need_clarify\":false,\"normalized_word\":\"电梯\",\"original_word\":\"电梯\",\"word_type\":\"\"}]}\n","raw_query":"电梯在哪里","sentiment_analysis":{"label":"1","pval":0.9960},"status":0,"timestamp":0},"schema":{"domain_confidence":0.0,"intent":"QUERYPOSITION","intent_confidence":100.0,"slots":[{"begin":0,"confidence":100.0,"length":2,"merge_method":"updated","name":"user_queryposition","normalized_word":"电梯","original_word":"电梯","session_offset":0,"sub_slots":[],"word_type":""}]},"status":0},"timestamp":"2018-11-03 14:29:32.696"}],"session_id":"session-1541226572612-746135415-8013-490"}}

把这么多的玩意再次传回去。。确定吗？

#16smarthomeLYZ回复于2018-10

拼接参数...:

String params =
"{\"bot_session\":\"{\\\"session_id\\\":\\\" " + session_id + "\\\"}\"," +
"\"log_id\":\"" + logId + "\"" +
",\"request\":" +
"{\"bernard_level\":1" +
",\"client_session\":\"" +
"{\\\"client_results\\\":\\\"\\\", " +
"\\\"candidate_options\\\":[]}\"," +
"\"query\":\"" + content + "\"," +
"\"query_info\":" +
"{\"asr_candidates\":[]," +
"\"source\":\"ASR\"," +
"\"type\":\"TEXT\"}," +
"\"updates\":\"\"," +
"\"user_id\":\"" + user_id + "\"}," +
"\"bot_id\":" + Constants.BAIDU_UNIT_BOT_ID + "," +
"\"version\":\"2.0\"}";

#15glary_augus回复于2018-10

String params = "{\"bot_session\":"+bot_session+",\"log_id\":\"7758521\",\"request\":{\"bernard_level\":1,\"client_session\":\"{\\\"client_results\\\":\\\"\\\", \\\"candidate_options\\\":[]}\",\"query\":\""+ask+"\",\"query_info\":{\"asr_candidates\":[],\"source\":\"KEYBOARD\",\"type\":\"TEXT\"},\"updates\":\"\",\"user_id\":\"88888\"},\"bot_id\":\"你自己的bot_id\",\"version\":\"2.0\"}";

这么拼接就行了。然后第一次传的bot_session="\"\"";

之后的bot_session就把从result中取出的bot_session直接传进去就行了。

#14redlavi回复于2018-10

对#12 sun_lei0422回复

首先 bot_session 和 client_session 都是 json格式的字符串类型，所以需要对这两个json值的引号进行转移，可以参考 9 楼的同学的写法

展开

“转移”应该是错别字，应该是“转义”。引号写成”\"“

#13tiancorn回复于2018-10

对#9 星空下hh回复

问一下，传入session_id但是两轮对话还是没有关联，是怎么回事啊 { "bot_session": "{\"session_id\":\" session-1536030295392-1220742608-8013-218 \"}", "log_id": "199606", "request": { "bernard_level": 0, "client_session": "{\"client_results\":\"\", \"candidate_options\":[]}", "query": "流水账单下载", "query_info": { "asr_candidates": [ ], "source": "KEYBOARD", "type": "TEXT" }, "updates": "", "user_id": "1234" }, "bot_id": "10729", "version": "2.0" }

展开

这前后空格是你自己加上去的？还是怎么弄进去的？

#12sun_lei0422回复于2018-10

对#10 love卵疼回复

好烦一直是请求数据不对 "{ "bot_session":{ "session_id":"session-1539308171714-2261805925-8013-24"},"log_id":"7758521","request":{ "bernard_level":2,"client_session":"{"client_results":"", "candidate_options":[]}","query":"hyc在不在","query_info":{ "asr_candidates":[],"source":"KEYBOARD","type":"TEXT"},"updates":"","user_id":"99635"},"bot_id":12752,"version":"2.0"}" 都不知道这个拼接哪里有错。

展开

首先 bot_session 和 client_session 都是 json格式的字符串类型，所以需要对这两个json值的引号进行转移，可以参考 9 楼的同学的写法

#11sun_lei0422回复于2018-10

对#9 星空下hh回复

展开

你的上轮返回的session_id 值的前后就包含空格吗？？"bot_session": "{\"session_id\":\" session-1536030295392-1220742608-8013-218 \"}"

#10love卵疼回复于2018-10

好烦一直是请求数据不对

"{ "bot_session":{ "session_id":"session-1539308171714-2261805925-8013-24"},"log_id":"7758521","request":{ "bernard_level":2,"client_session":"{"client_results":"", "candidate_options":[]}","query":"hyc在不在","query_info":{ "asr_candidates":[],"source":"KEYBOARD","type":"TEXT"},"updates":"","user_id":"99635"},"bot_id":12752,"version":"2.0"}"

都不知道这个拼接哪里有错。

#9星空下hh回复于2018-09

问一下，传入session_id但是两轮对话还是没有关联，是怎么回事啊

{
"bot_session": "{\"session_id\":\" session-1536030295392-1220742608-8013-218 \"}",
"log_id": "199606",
"request": {
"bernard_level": 0,
"client_session": "{\"client_results\":\"\", \"candidate_options\":[]}",
"query": "流水账单下载",
"query_info": {
"asr_candidates": [ ],
"source": "KEYBOARD",
"type": "TEXT"
},
"updates": "",
"user_id": "1234"
},
"bot_id": "10729",
"version": "2.0"
}

快速回复

小编推荐

【7.18升级】自动预标注上线、词典值管理升级

TroubleMaker源 7回复

【智能对话深度实战营】首批训练师认证名单公示

魏亚非669 38回复

语音语义一体化全新升级，一次数据交互全搞定

用户已被禁言 18回复

TOP

操作指南

常见问答

平台公告

经验交流

技术专区

文字识别

人脸识别

语音技术

PaddlePaddle

EasyDL

BML

EasyData

AI Studio

UNIT

人体分析

图像搜索

图像识别

内容审核

自然语言处理

机器人视觉

视频技术

增强现实

知识图谱

智能创作

智能呼叫中心

文心

EdgeBoard

DuerOS

EasyEdge

度目硬件

百度AI市场

Doris

AI赛事

百度之星大赛

AI Studio人工智能竞赛

语言与智能技术竞赛

千言数据集

集思广益

共享工具

头脑风暴

成果展示

智能客服