前言
最近项目上有个数字人问数需求,即用户通过自然语言向大模型提问相关数据问题,大模型根据用户提问生成 SQL,查询数据库,根据结果总结回答用户。根据网上相关教程和提示词,在Dify 简单糊一个工作流看看效果。
效果演示


构建知识库
将数据库表结构存入知识库;
知识库的分段和索引模型等设置看着选吧,保证能正确分段和召回即可。
我使用的是 PostgreSQL,表结构如下:(建议表结构有完整的注释)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
| ### 事件统计结果表: CREATE TABLE dwd.dwd_event_statistics ( id SERIAL PRIMARY KEY, event_num INTEGER, deal_event_num INTEGER, full_date DATE, date_of_year INTEGER, date_of_month INTEGER, date_of_day INTEGER, info_src VARCHAR(100), level1_type_name VARCHAR(100), level2_type_name VARCHAR(100), level3_type_name VARCHAR(100), area_name VARCHAR(100), street_name VARCHAR(100), village_name VARCHAR(100), etl_time TIMESTAMP DEFAULT CURRENT_TIMESTAMP );
COMMENT ON TABLE dwd.dwd_event_statistics IS '事件统计结果表'; COMMENT ON COLUMN dwd.dwd_event_statistics.id IS '主键'; COMMENT ON COLUMN dwd.dwd_event_statistics.event_num IS '事件总数'; COMMENT ON COLUMN dwd.dwd_event_statistics.deal_event_num IS '已办结事件总数'; COMMENT ON COLUMN dwd.dwd_event_statistics.full_date IS '事件上报日期'; COMMENT ON COLUMN dwd.dwd_event_statistics.date_of_year IS '事件上报年份'; COMMENT ON COLUMN dwd.dwd_event_statistics.date_of_month IS '事件上报月份'; COMMENT ON COLUMN dwd.dwd_event_statistics.date_of_day IS '事件上报日'; COMMENT ON COLUMN dwd.dwd_event_statistics.info_src IS '事件来源'; COMMENT ON COLUMN dwd.dwd_event_statistics.level1_type_name IS '一级事件类别'; COMMENT ON COLUMN dwd.dwd_event_statistics.level2_type_name IS '二级事件类别'; COMMENT ON COLUMN dwd.dwd_event_statistics.level3_type_name IS '三级事件类别'; COMMENT ON COLUMN dwd.dwd_event_statistics.area_name IS '区名称'; COMMENT ON COLUMN dwd.dwd_event_statistics.street_name IS '街镇名称'; COMMENT ON COLUMN dwd.dwd_event_statistics.village_name IS '社区名称'; COMMENT ON COLUMN dwd.dwd_event_statistics.etl_time IS '数据入库时间';
|
构建ChatFlow应用
具体工作流如下:

工作流介绍
1、主要流程:
开始->知识库检索表结构->大模型生成 SQL->提取大模型返回的 SQL->调用接口执行 SQL 获取结果->大模型总结结果->输出结果
2、大模型生成 SQL 节点,提示词如下:
SYSTEM:
1
| #角色:你是一位精通SQL语言的数据库专家,精通PostgreSQL,擅长理解用户需求并编写正确的SQL语句
|
USER:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
| #任务:你的任务是理解用户的输入{{#sys.query#}}和上下文内容{{#context#}},编写符合用户需求以及可正常执行的SQL语句。 #关键步骤: 1、对用户输入的内容进行识别和判断,如果内容涉及政治、时事、社会问题以及违背道德和法律法规的情形,一律输出:”您提出的问题超出我应当回答的范围,请询问与业务相关的问题,否则我无法作出回答“,如果内容不涉及上述内容,但是你根据已有信息无法按要求回答,请输出“暂无法根据您的问题查询信息,请尝试换种方式提问“ 2、根据用户输入的内容和上下文信息,形成内容分类,根据内容分类按照以下规则从知识库“数据库表结构”中检索数据表结构信息: -内容分类与事件相关,则检索“事件统计结果表” -内容分类与隐患点相关,则检索“隐患点信息表” 注意:务必严格按照上述分类获得对应的检索关键词,不得生成新的检索关键词。如果你认为用户的提问无法匹配到合适的分类,请输出提示:为确保查询获得准确信息,请再把你的需求描述细致一些 3、根据用户输入的内容和上下文信息,形成一个符合用户意图的完整问题,以此作为输入在知识库“sql示例”中检索SQL语句参考示例 4、基于对上下文和对用户提问的理解,按照检索到的数据表结构信息,以及SQL参考示例,编写SQL查询语句。注意,若内容分类与参考示例中的分类不符时,则忽略这个示例。另外,不是所有情况下都有示例参考,没有示例时请按照自己的理解和掌握的知识编写SQL语句 5、去除SQL语句中多余的注释、换行符等无用信息,Markdown语法,输出一个纯净的、可直接执行的SQL语句 #编写SQL时的注意事项: 1. 务必根据上下文提供的数据表结构描述来编写SQL语句,确保仅使用数据表结构描述中提到的表名和字段名,并参考对字段的解释 2. 确保SQL兼容PostgreSQL 3. 只用简体中文 4. 只输出一个完整SQL语句,无注释,无前缀或后缀,不需要格式化,返回输出的时候不要使用代码块,仅输出纯字符串的 SQL 语句,确保返回的 SQL 语句可直接执行并获得预期的结果 5. 对于字符串和长文本类型的字段,除非用户有特别说明,否则都用LIKE操作,而不是等于操作,例如:WHERE 产品型号 LIKE N'%关键词%',而不是WHERE 产品型号='关键词' 6. 除法处理:参考以下模板以避免错误: CASE WHEN [除数] = 0 THEN 0 ELSE CAST([被除数] AS FLOAT) / [除数] END AS [结果列名] 7.如果编写 SQL 语句是需要确定当前时间,请记住当前时间为{{#1741074867787.text#}}
|
3、大模型总结结果节点,提示词如下:
SYSTEM:
1
| #角色:你是一个专业的问题分析师和图表绘制专家,能根据用户的提问及查询结果,回答用户的问题,并可以绘制相关的图表。
|
USER:
1 2 3 4 5 6 7
| #任务:根据用户的提问:{{#sys.query#}},已经生成了可以执行的 SQL:{{#1741074989846.result#}},SQL 执行结果为:{{#1741075031589.body#}},请判断用户的提问和SQL结果是否适合生成图表展示 1、如果不适合绘制图表,请直接根据 SQL 结果回答用户的问题。输出格式为 json字符串:"{"canDraw":0,"answer":"你的回答内容","data":null,"type":null}"。 2、如果适合绘制图表,请先根据 SQL 结果回答用户的问题,再生成图表,暂时仅支持饼图、线性图表和柱状图,请先根据问题和数据选择合适的图表,图表和问题结果请按以下要求输出: - 如果适合绘制饼图,请根据 SQL 结果,输出格式为 json字符串:"{"canDraw":1,"answer":"你的回答内容","data":"用于生成饼图的数据,每个数字之间用 ; 分隔",",type":"饼图的分类,每个分类之间用 ; 分隔"}" - 如果适合绘制线性图表,请根据 SQL 结果,输出格式为 json字符串:"{"canDraw":2,"answer":"你的回答内容","data":"用于生成线性图表的数据,每个数字之间用 ; 分隔",",type":"线性图表的 x 轴,每个文本之间用 ; 分隔"}" - 如果适合绘制柱状图,请根据 SQL 结果,输出格式为 json字符串:"{"canDraw":3,"answer":"你的回答内容","data":"用于生成柱状图的数据,每个数字之间用 ; 分隔",",type":"柱状图的 x 轴,每个文本之间用 ; 分隔"}" #输出要求: 请严格按照上述的 JSON 格式输出,不要输出除 JSON 外的多余文字。
|
4、执行SQL接口,在服务器上写个flask接口,代码如下:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93
| from flask import Flask, request, jsonify import psycopg2 import os from dotenv import load_dotenv
load_dotenv()
app = Flask(__name__)
DB_HOST = os.getenv("DB_HOST", "localhost") DB_NAME = os.getenv("DB_NAME", "postgres") DB_USER = os.getenv("DB_USER", "postgres") DB_PASSWORD = os.getenv("DB_PASSWORD", "difyai123456") DB_PORT = os.getenv("DB_PORT", "5432")
def get_db_connection(): """创建并返回数据库连接""" conn = psycopg2.connect( host=DB_HOST, database=DB_NAME, user=DB_USER, password=DB_PASSWORD, port=DB_PORT ) return conn
@app.route('/execute_sql', methods=['POST']) def execute_sql(): """执行SQL语句的API接口""" try: data = request.get_json() if not data or 'sql' not in data: return jsonify({"error": "Missing SQL query in request"}), 400 sql_query = data['sql'] conn = get_db_connection() cursor = conn.cursor() cursor.execute(sql_query) if cursor.description: columns = [desc[0] for desc in cursor.description] rows = cursor.fetchall() results = [] for row in rows: result = {} for i, column in enumerate(columns): result[column] = row[i] results.append(result) response = { "status": "success", "count": len(results), "results": results } else: conn.commit() affected_rows = cursor.rowcount response = { "status": "success", "message": f"Query executed successfully. Affected rows: {affected_rows}" } cursor.close() conn.close() return jsonify(response) except Exception as e: if 'conn' in locals() and conn: conn.rollback() conn.close() return jsonify({"error": str(e)}), 500
if __name__ == '__main__': app.run(debug=True, host='0.0.0.0', port=5000)
|
5、注意要点:
大模型生成 SQL 节点,若使用推理模型,其返回的结果会是思考过程+SQL结果,所以需要通过代码正则匹配提取最后的SQL结果。
由于大模型无法知道当前时间,如果提问涉及到与当前时间相关,需要提前获取当前时间,填入提示词中。
6、DSL文件(Version 0.15.3)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000
| app: description: '' icon: 🤖 icon_background: '#FFEAD5' mode: advanced-chat name: 数字人聊天流 use_icon_as_answer_icon: false kind: app version: 0.1.5 workflow: conversation_variables: [] environment_variables: [] features: file_upload: allowed_file_extensions: - .JPG - .JPEG - .PNG - .GIF - .WEBP - .SVG allowed_file_types: - image allowed_file_upload_methods: - local_file - remote_url enabled: false fileUploadConfig: audio_file_size_limit: 50 batch_count_limit: 5 file_size_limit: 15 image_file_size_limit: 10 video_file_size_limit: 100 workflow_file_upload_limit: 10 image: enabled: false number_limits: 3 transfer_methods: - local_file - remote_url number_limits: 3 opening_statement: '' retriever_resource: enabled: true sensitive_word_avoidance: enabled: false speech_to_text: enabled: false suggested_questions: [] suggested_questions_after_answer: enabled: false text_to_speech: enabled: false language: '' voice: '' graph: edges: - data: isInIteration: false sourceType: knowledge-retrieval targetType: llm id: 1741074832553-source-llm-target selected: false source: '1741074832553' sourceHandle: source target: llm targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: tool targetType: knowledge-retrieval id: 1741074867787-source-1741074832553-target selected: false source: '1741074867787' sourceHandle: source target: '1741074832553' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: llm targetType: code id: llm-source-1741074989846-target source: llm sourceHandle: source target: '1741074989846' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: http-request targetType: llm id: 1741075031589-source-1741075092617-target source: '1741075031589' sourceHandle: source target: '1741075092617' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: start targetType: question-classifier id: 1741074811356-source-1741075993453-target source: '1741074811356' sourceHandle: source target: '1741075993453' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: question-classifier targetType: tool id: 1741075993453-1-1741074867787-target source: '1741075993453' sourceHandle: '1' target: '1741074867787' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: question-classifier targetType: answer id: 1741075993453-2-1741076049272-target source: '1741075993453' sourceHandle: '2' target: '1741076049272' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: question-classifier targetType: answer id: 1741075993453-1741076958110-1741076983758-target source: '1741075993453' sourceHandle: '1741076958110' target: '1741076983758' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: code targetType: code id: 1741074989846-source-1741077295508-target source: '1741074989846' sourceHandle: source target: '1741077295508' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: code targetType: if-else id: 1741077295508-source-1741077672910-target source: '1741077295508' sourceHandle: source target: '1741077672910' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: if-else targetType: http-request id: 1741077672910-true-1741075031589-target source: '1741077672910' sourceHandle: 'true' target: '1741075031589' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: if-else targetType: answer id: 1741077672910-false-1741077713161-target source: '1741077672910' sourceHandle: 'false' target: '1741077713161' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: llm targetType: code id: 1741075092617-source-1741080767918-target source: '1741075092617' sourceHandle: source target: '1741080767918' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: code targetType: if-else id: 1741080767918-source-1741081080222-target source: '1741080767918' sourceHandle: source target: '1741081080222' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: if-else targetType: answer id: 1741081080222-true-answer-target source: '1741081080222' sourceHandle: 'true' target: answer targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: tool targetType: answer id: 1741082537467-source-1741081418797-target source: '1741082537467' sourceHandle: source target: '1741081418797' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: if-else targetType: tool id: 1741081080222-0042cbd7-7395-457a-aa6c-86a74a9892f0-1741082537467-target source: '1741081080222' sourceHandle: 0042cbd7-7395-457a-aa6c-86a74a9892f0 target: '1741082537467' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: if-else targetType: tool id: 1741081080222-f6e336df-d32e-4cc3-84f2-18b4f625cb79-1741156091023-target source: '1741081080222' sourceHandle: f6e336df-d32e-4cc3-84f2-18b4f625cb79 target: '1741156091023' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: if-else targetType: tool id: 1741081080222-61cbc9c9-8893-4de6-879b-f891e4302ccf-1741156101562-target source: '1741081080222' sourceHandle: 61cbc9c9-8893-4de6-879b-f891e4302ccf target: '1741156101562' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: tool targetType: answer id: 1741156091023-source-1741081418797-target source: '1741156091023' sourceHandle: source target: '1741081418797' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: tool targetType: answer id: 1741156101562-source-1741081418797-target source: '1741156101562' sourceHandle: source target: '1741081418797' targetHandle: target type: custom zIndex: 0 - data: isInIteration: false sourceType: if-else targetType: answer id: 1741081080222-false-answer-target source: '1741081080222' sourceHandle: 'false' target: answer targetHandle: target type: custom zIndex: 0 nodes: - data: desc: '' selected: false title: 开始 type: start variables: [] height: 54 id: '1741074811356' position: x: 30 y: 288 positionAbsolute: x: 30 y: 288 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: context: enabled: true variable_selector: - '1741074832553' - result desc: '' memory: query_prompt_template: '{{#sys.query#}}' role_prefix: assistant: '' user: '' window: enabled: false size: 10 model: completion_params: temperature: 0.7 mode: chat name: deepseek-ai/DeepSeek-V3 provider: siliconflow prompt_template: - id: f14db6d7-353e-4bdd-bf28-17e365f9bb45 role: system text: '#角色:你是一位精通SQL语言的数据库专家,精通Postgres SQL,擅长理解用户需求并编写正确的SQL语句' - id: 2c6b188f-5624-4f72-a90b-68a90620ca4f role: user text: '#任务:你的任务是理解用户的输入{{#sys.query#}}和上下文内容{{#context#}},编写符合用户需求以及可正常执行的SQL语句。
#关键步骤:
1、对用户输入的内容进行识别和判断,如果内容涉及政治、时事、社会问题以及违背道德和法律法规的情形,一律输出:”您提出的问题超出我应当回答的范围,请询问与业务相关的问题,否则我无法作出回答“,如果内容不涉及上述内容,但是你根据已有信息无法按要求回答,请输出“暂无法根据您的问题查询信息,请尝试换种方式提问“
2、根据用户输入的内容和上下文信息,形成内容分类,根据内容分类按照以下规则从知识库“数据库表结构”中检索数据表结构信息:
-内容分类与事件相关,则检索“事件统计结果表”
-内容分类与隐患点相关,则检索“隐患点信息表”
注意:务必严格按照上述分类获得对应的检索关键词,不得生成新的检索关键词。如果你认为用户的提问无法匹配到合适的分类,请输出提示:为确保查询获得准确信息,请再把你的需求描述细致一些
3、根据用户输入的内容和上下文信息,形成一个符合用户意图的完整问题,以此作为输入在知识库“sql示例”中检索SQL语句参考示例
4、基于对上下文和对用户提问的理解,按照检索到的数据表结构信息,以及SQL参考示例,编写SQL查询语句。注意,若内容分类与参考示例中的分类不符时,则忽略这个示例。另外,不是所有情况下都有示例参考,没有示例时请按照自己的理解和掌握的知识编写SQL语句
5、去除SQL语句中多余的注释、换行符等无用信息,Markdown语法,输出一个纯净的、可直接执行的SQL语句
#编写SQL时的注意事项:
1. 务必根据上下文提供的数据表结构描述来编写SQL语句,确保仅使用数据表结构描述中提到的表名和字段名,并参考对字段的解释
2. 确保SQL兼容Postgres SQL
3. 只用简体中文
4. 只输出一个完整SQL语句,无注释,无前缀或后缀,不需要格式化,返回输出的时候不要使用代码块,仅输出纯字符串的 SQL 语句,确保返回的 SQL 语句可直接执行并获得预期的结果
5. 对于字符串和长文本类型的字段,除非用户有特别说明,否则都用LIKE操作,而不是等于操作,例如:WHERE 产品型号 LIKE N''%关键词%'',而不是WHERE 产品型号=''关键词''
6. 除法处理:参考以下模板以避免错误:
CASE WHEN [除数] = 0 THEN 0
ELSE CAST([被除数] AS FLOAT) / [除数]
END AS [结果列名]
7.如果编写 SQL 语句是需要确定当前时间,请记住当前时间为{{#1741074867787.text#}}' selected: false title: 生成 SQL type: llm variables: [] vision: enabled: false height: 98 id: llm position: x: 1246 y: 288 positionAbsolute: x: 1246 y: 288 selected: true sourcePosition: right targetPosition: left type: custom width: 244 - data: answer: '{{#1741080767918.answer#}}' desc: '' selected: false title: 直接回复 type: answer variables: [] height: 103 id: answer position: x: 3952 y: 353.7142857142857 positionAbsolute: x: 3952 y: 353.7142857142857 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: dataset_ids: - bb33c628-92b1-4368-b0d4-77dd7b6a35cd desc: '' multiple_retrieval_config: reranking_enable: true reranking_mode: reranking_model reranking_model: model: netease-youdao/bce-reranker-base_v1 provider: siliconflow top_k: 4 query_variable_selector: - '1741074811356' - sys.query retrieval_mode: multiple selected: false title: 知识检索 type: knowledge-retrieval height: 92 id: '1741074832553' position: x: 942 y: 288 positionAbsolute: x: 942 y: 288 sourcePosition: right targetPosition: left type: custom width: 244 - data: desc: '' provider_id: time provider_name: time provider_type: builtin selected: false title: 获取当前时间 tool_configurations: format: '%Y-%m-%d %H:%M:%S' timezone: UTC tool_label: 获取当前时间 tool_name: current_time tool_parameters: {} type: tool height: 116 id: '1741074867787' position: x: 638 y: 288 positionAbsolute: x: 638 y: 288 sourcePosition: right targetPosition: left type: custom width: 244 - data: code: "import re\ndef main(arg1: str) -> dict:\n # 提取 text 字段\n text\ \ = arg1\n # 使用正则表达式提取 </details> 标签后的内容\n pattern = r'</details>\\\ n\\n(.*)'\n match = re.search(pattern, text)\n if match:\n \ \ # 提取 </details> 标签后的内容\n sql_statement = match.group(1).strip()\n\ \ return {\"result\":sql_statement}\n else:\n return {\"\ result\": text}" code_language: python3 desc: '' outputs: result: children: null type: string selected: false title: 提取SQL type: code variables: - value_selector: - llm - text variable: arg1 height: 54 id: '1741074989846' position: x: 1550 y: 288 positionAbsolute: x: 1550 y: 288 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: authorization: config: null type: no-auth body: data: - id: key-value-566 key: '' type: text value: '{"sql":"{{#1741074989846.result#}}"}' type: json desc: '' headers: '' method: post params: '' retry_config: max_retries: 3 retry_enabled: true retry_interval: 100 selected: false timeout: max_connect_timeout: 0 max_read_timeout: 0 max_write_timeout: 0 title: 执行 SQL type: http-request url: http://180.00.00.00:5000/execute_sql variables: [] height: 135 id: '1741075031589' position: x: 2335.428571428571 y: 288 positionAbsolute: x: 2335.428571428571 y: 288 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: context: enabled: false variable_selector: [] desc: '' model: completion_params: temperature: 0.7 mode: chat name: Qwen/Qwen2.5-32B-Instruct provider: siliconflow prompt_template: - id: 10d9b6b1-a3ee-425f-8b8e-04d37f99e9ec role: system text: '#角色:你是一个专业的问题分析师和图表绘制专家,能根据用户的提问及查询结果,回答用户的问题,并可以绘制相关的图表。' - id: d5077a8b-944e-4529-82ad-01a3bcc36a5b role: user text: '#任务:根据用户的提问:{{#sys.query#}},已经生成了可以执行的 SQL:{{#1741074989846.result#}},SQL 执行结果为:{{#1741075031589.body#}},请判断用户的提问和SQL结果是否适合生成图表展示
1、如果不适合绘制图表,请直接根据 SQL 结果回答用户的问题。输出格式为 json字符串:"{"canDraw":0,"answer":"你的回答内容","data":null,"type":null}"。
2、如果适合绘制图表,请先根据 SQL 结果回答用户的问题,再生成图表,暂时仅支持饼图、线性图表和柱状图,请先根据问题和数据选择合适的图表,图表和问题结果请按以下要求输出:
- 如何适合绘制饼图,请根据 SQL 结果,输出格式为 json字符串:"{"canDraw":1,"answer":"你的回答内容","data":"用于生成饼图的数据,每个数字之间用 ; 分隔",",type":"饼图的分类,每个分类之间用 ; 分隔"}"
- 如何适合绘制线性图表,请根据 SQL 结果,输出格式为 json字符串:"{"canDraw":2,"answer":"你的回答内容","data":"用于生成线性图表的数据,每个数字之间用 ; 分隔",",type":"线性图表的 x 轴,每个文本之间用 ; 分隔"}"
- 如何适合绘制柱状图,请根据 SQL 结果,输出格式为 json字符串:"{"canDraw":3,"answer":"你的回答内容","data":"用于生成柱状图的数据,每个数字之间用 ; 分隔",",type":"柱状图的 x 轴,每个文本之间用 ; 分隔"}"
#输出要求:
请严格按照上述的 JSON 格式输出,不要输出除 JSON 外的多余文字。' selected: false title: 解析SQL 执行结果 type: llm variables: [] vision: enabled: false height: 98 id: '1741075092617' position: x: 2632.2857142857138 y: 288 positionAbsolute: x: 2632.2857142857138 y: 288 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: classes: - id: '1' name: 数据查询问题 - id: '2' name: 操作指令 - id: '1741076958110' name: 其他 desc: '' instruction: '' instructions: '' model: completion_params: temperature: 0.7 mode: chat name: Qwen/Qwen2.5-32B-Instruct provider: siliconflow query_variable_selector: - '1741074811356' - sys.query selected: false title: 问题分类器 topics: [] type: question-classifier vision: enabled: false height: 212 id: '1741075993453' position: x: 334 y: 288 positionAbsolute: x: 334 y: 288 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: answer: 指令操作开发中。 desc: '' selected: false title: 直接回复 2 type: answer variables: [] height: 100 id: '1741076049272' position: x: 638 y: 444 positionAbsolute: x: 638 y: 444 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: answer: 暂无法理解您的问题或需求。 desc: '' selected: false title: 直接回复 3 type: answer variables: [] height: 100 id: '1741076983758' position: x: 638 y: 583 positionAbsolute: x: 638 y: 583 sourcePosition: right targetPosition: left type: custom width: 244 - data: code: "\ndef main(arg1: str) -> dict:\n if arg1.lower().startswith('select'):\n\ \ return {\n \"is_select_sql\": 1,\n \"result\"\ : arg1\n }\n else :\n return {\n \"is_select_sql\"\ : 0,\n \"result\": arg1\n }\n\n \n" code_language: python3 desc: '' outputs: is_select_sql: children: null type: number result: children: null type: string selected: false title: 判断是否为 SQL type: code variables: - value_selector: - '1741074989846' - result variable: arg1 height: 54 id: '1741077295508' position: x: 1872.5714285714284 y: 288 positionAbsolute: x: 1872.5714285714284 y: 288 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: cases: - case_id: 'true' conditions: - comparison_operator: '=' id: 34f07303-7757-4f85-bf69-8e260999a284 value: '1' varType: number variable_selector: - '1741077295508' - is_select_sql id: 'true' logical_operator: and desc: '' selected: false title: 条件分支 type: if-else height: 126 id: '1741077672910' position: x: 1986.8571428571427 y: 430.57142857142856 positionAbsolute: x: 1986.8571428571427 y: 430.57142857142856 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: answer: '{{#1741077295508.result#}}' desc: '' selected: false title: 直接回复 4 type: answer variables: [] height: 103 id: '1741077713161' position: x: 2335.428571428571 y: 462 positionAbsolute: x: 2335.428571428571 y: 462 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: code: "import json\ndef main(arg1: str) -> dict:\n # 解析 JSON 数据\n data\ \ = json.loads(arg1)\n return data" code_language: python3 desc: '' outputs: answer: children: null type: string canDraw: children: null type: number data: children: null type: string type: children: null type: string selected: false title: 解析问题结果参数提取 type: code variables: - value_selector: - '1741075092617' - text variable: arg1 height: 54 id: '1741080767918' position: x: 2956.285714285714 y: 393 positionAbsolute: x: 2956.285714285714 y: 393 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: cases: - case_id: 'true' conditions: - comparison_operator: '=' id: 3c9f6dea-3f0c-4037-a964-42427b206679 value: '0' varType: number variable_selector: - '1741080767918' - canDraw id: 'true' logical_operator: and - case_id: 0042cbd7-7395-457a-aa6c-86a74a9892f0 conditions: - comparison_operator: '=' id: cc2b3f2e-f428-48eb-90bf-bc0466e1e9bf value: '1' varType: number variable_selector: - '1741080767918' - canDraw id: 0042cbd7-7395-457a-aa6c-86a74a9892f0 logical_operator: and - case_id: f6e336df-d32e-4cc3-84f2-18b4f625cb79 conditions: - comparison_operator: '=' id: 69a6583e-35c5-4eea-a352-8a3d52b9d5cc value: '2' varType: number variable_selector: - '1741080767918' - canDraw id: f6e336df-d32e-4cc3-84f2-18b4f625cb79 logical_operator: and - case_id: 61cbc9c9-8893-4de6-879b-f891e4302ccf conditions: - comparison_operator: '=' id: 06b02ba1-482e-449e-8a48-18505e1d57d3 value: '3' varType: number variable_selector: - '1741080767918' - canDraw id: 61cbc9c9-8893-4de6-879b-f891e4302ccf logical_operator: and desc: '' selected: false title: 判断是否生成图表 type: if-else height: 270 id: '1741081080222' position: x: 3261.714285714286 y: 393 positionAbsolute: x: 3261.714285714286 y: 393 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: answer: '{{#1741080767918.answer#}}{{#1741082537467.files#}}{{#1741156091023.files#}}{{#1741156101562.files#}}' desc: '' selected: false title: 直接回复 5 type: answer variables: [] height: 158 id: '1741081418797' position: x: 4252 y: 513.7142857142858 positionAbsolute: x: 4252 y: 513.7142857142858 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: desc: '' provider_id: chart provider_name: chart provider_type: builtin selected: false title: 饼图 tool_configurations: {} tool_label: 饼图 tool_name: pie_chart tool_parameters: categories: type: mixed value: '{{#1741080767918.type#}}' data: type: mixed value: '{{#1741080767918.data#}}' type: tool height: 54 id: '1741082537467' position: x: 3952 y: 513.7142857142858 positionAbsolute: x: 3952 y: 513.7142857142858 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: desc: '' provider_id: chart provider_name: chart provider_type: builtin selected: false title: 线性图表 tool_configurations: {} tool_label: 线性图表 tool_name: line_chart tool_parameters: data: type: mixed value: '{{#1741080767918.data#}}' x_axis: type: mixed value: '{{#1741080767918.type#}}' type: tool height: 54 id: '1741156091023' position: x: 3952 y: 606.7142857142858 positionAbsolute: x: 3952 y: 606.7142857142858 selected: false sourcePosition: right targetPosition: left type: custom width: 244 - data: desc: '' provider_id: chart provider_name: chart provider_type: builtin selected: false title: 柱状图 tool_configurations: {} tool_label: 柱状图 tool_name: bar_chart tool_parameters: data: type: mixed value: '{{#1741080767918.data#}}' x_axis: type: mixed value: '{{#1741080767918.type#}}' type: tool height: 54 id: '1741156101562' position: x: 3952 y: 699.7142857142858 positionAbsolute: x: 3952 y: 699.7142857142858 selected: false sourcePosition: right targetPosition: left type: custom width: 244 viewport: x: -1044.298931079242 y: -100.79975541473073 zoom: 0.7020879457826513
|
总结
可以通过Agent方式实现?
可以,修改下提示词,让大模型既能生成SQL,也能分析SQL结果,也能生成图表,添加相关工具,让Agent自动编排,效果如下:



生成SQL的准确率怎么样?
这种基于知识库和提示词的大模型text2Sql生成的SQL并不能保证100%正确,不同类型和参数的大模型生成的 SQL 效果不一,当提问比较复杂或者涉及到多表关联查询时,其生成的SQL正确率也会比较低。
怎么提高Text2SQL正确率?
待研究🧐
相关的开源软件?
1、Supersonic:https://github.com/tencentmusic/supersonic
2、DB-GPT:https://github.com/eosphoros-ai/DB-GPT
等等,还有很多。