制作数据可视化助手创建workflow完成用户excel数据柱状图可视化。基本思路是用户输入文档-然后用文档提取器提取文字--然后用本地模型去清洗数据--然后利用代码执行模块进行输出显示。期间遇到的问题1是本地模型处理速度比在线的大模型慢很多微调模型参数模型使用这三个参数从几十秒 缩短至几秒。同时注意提示词提示词应该也还可以优化2.代码执行的python程序不会写实际用豆包生成实测可行但需要加约束豆包生成的显示柱状图的程序import csv import json from collections import defaultdict def main(csv_string): try: raw_lines [line.strip() for line in csv_string.split(\n) if line.strip()] if not raw_lines: return {output: echarts\n{\error\:\没有CSV数据\}\n} reader csv.reader(raw_lines) headers next(reader) x_axis_name headers[0] if len(headers)1 else data_dict defaultdict(lambda: defaultdict(float)) x_categories [] for row in reader: if len(row) 3: continue x_val row[0].strip() s_val row[1].strip() if x_val not in x_categories: x_categories.append(x_val) try: num float(row[2]) except Exception: num 0.0 data_dict[x_val][s_val] num all_series_names sorted({s for x in data_dict for s in data_dict[x]}) series_list [] for s_name in all_series_names: series_list.append({ name: s_name, type: bar, data: [data_dict[x].get(s_name, 0) for x in x_categories] }) option { tooltip: {trigger: axis}, legend: { data: all_series_names, type: scroll }, grid: { left: 3%, right: 4%, bottom: 20%, containLabel: True }, xAxis: { type: category, data: x_categories, name: x_axis_name, boundaryGap: True, axisLabel: { rotate: 30 } }, yAxis: { type: value, name: headers[2] if len(headers)3 else }, series: series_list } render_text echarts\n json.dumps(option, ensure_asciiFalse, indent2) \n return {output: render_text} except Exception as e: return {output: fecharts\n{{\error\:\程序异常{str(e)}\}}\n}豆包生成的显示饼图的程序import csv import json from collections import defaultdict def main(csv_string): try: raw_lines [line.strip() for line in csv_string.split(\n) if line.strip()] if not raw_lines: return {output: echarts\n{\error\:\没有CSV数据\}\n} reader csv.reader(raw_lines) headers next(reader) sum_dict defaultdict(float) for row in reader: if len(row) 3: continue product row[1].strip() try: sales float(row[2]) except: sales 0 sum_dict[product] sales data_list [{name:k,value:v} for k,v in sum_dict.items()] option { tooltip: {trigger:item}, legend:{orient:vertical,left:left}, series:[ { name:销售数量, type:pie, radius:60%, data:data_list } ] } render_text echarts\n json.dumps(option, ensure_asciiFalse, indent2) \n return {output: render_text} except Exception as e: return {output: fecharts\n{{\error\:\程序异常{str(e)}\}}\n}换用本地的千问 4B模型运行速度可以接受。
制作数据可视化助手创建workflow完成用户excel数据柱状图可视化。基本思路是用户输入文档-然后用文档提取器提取文字--然后用本地模型去清洗数据--然后利用代码执行模块进行输出显示。期间遇到的问题1是本地模型处理速度比在线的大模型慢很多微调模型参数模型使用这三个参数从几十秒 缩短至几秒。同时注意提示词提示词应该也还可以优化2.代码执行的python程序不会写实际用豆包生成实测可行但需要加约束豆包生成的显示柱状图的程序import csv import json from collections import defaultdict def main(csv_string): try: raw_lines [line.strip() for line in csv_string.split(\n) if line.strip()] if not raw_lines: return {output: echarts\n{\error\:\没有CSV数据\}\n} reader csv.reader(raw_lines) headers next(reader) x_axis_name headers[0] if len(headers)1 else data_dict defaultdict(lambda: defaultdict(float)) x_categories [] for row in reader: if len(row) 3: continue x_val row[0].strip() s_val row[1].strip() if x_val not in x_categories: x_categories.append(x_val) try: num float(row[2]) except Exception: num 0.0 data_dict[x_val][s_val] num all_series_names sorted({s for x in data_dict for s in data_dict[x]}) series_list [] for s_name in all_series_names: series_list.append({ name: s_name, type: bar, data: [data_dict[x].get(s_name, 0) for x in x_categories] }) option { tooltip: {trigger: axis}, legend: { data: all_series_names, type: scroll }, grid: { left: 3%, right: 4%, bottom: 20%, containLabel: True }, xAxis: { type: category, data: x_categories, name: x_axis_name, boundaryGap: True, axisLabel: { rotate: 30 } }, yAxis: { type: value, name: headers[2] if len(headers)3 else }, series: series_list } render_text echarts\n json.dumps(option, ensure_asciiFalse, indent2) \n return {output: render_text} except Exception as e: return {output: fecharts\n{{\error\:\程序异常{str(e)}\}}\n}豆包生成的显示饼图的程序import csv import json from collections import defaultdict def main(csv_string): try: raw_lines [line.strip() for line in csv_string.split(\n) if line.strip()] if not raw_lines: return {output: echarts\n{\error\:\没有CSV数据\}\n} reader csv.reader(raw_lines) headers next(reader) sum_dict defaultdict(float) for row in reader: if len(row) 3: continue product row[1].strip() try: sales float(row[2]) except: sales 0 sum_dict[product] sales data_list [{name:k,value:v} for k,v in sum_dict.items()] option { tooltip: {trigger:item}, legend:{orient:vertical,left:left}, series:[ { name:销售数量, type:pie, radius:60%, data:data_list } ] } render_text echarts\n json.dumps(option, ensure_asciiFalse, indent2) \n return {output: render_text} except Exception as e: return {output: fecharts\n{{\error\:\程序异常{str(e)}\}}\n}换用本地的千问 4B模型运行速度可以接受。
