microsoft
Graph creation using code interpreter using this solution accelerator
Hi,
I am trying to ask the agent in this solution accelerator using code interpreter tool. I have modified the agent definition as follow to enable code interpreter for the agent ("coding_tools" : true)
######## { "input_key": "", "type": "", "name": "VisualizationAgent", "deployment_name": "gpt-4.1-mini", "icon": "", "system_message": "You are a visualization agent with access to a Python sandbox (“python” tool). When asked to visualize or analyze data, you must: 1- Run Python to load data into pandas and plot with matplotlib. 2- Do not print or return source code. 3- Return only the rendered chart plus a short 1–2 sentence caption. 4- Label axes, add a clear title, avoid subplots unless asked, never use seaborn, and don’t specify colors unless the user requests. 5- If input looks like code or pseudo-code, treat it as instructions to execute in Python (don’t echo it). If you are about to output code, stop and instead run it in Python and return the image.", "description": "An agent that visualize analyzed data in graph form based on actual data and prediction.", "use_rag": false, "use_mcp": false, "use_bing": false, "use_reasoning": false, "index_name": "", "index_foundry_name": "", "coding_tools": true } ############
However, when I asked the agent to generate graph, for example using following prompt, the agent just return the python code to generate the graph into the chat window and not showing the graph itself. Is the code interpreter not being used correctly here?
User Prompt: Visualize sales trends by region from the following CSV data: Region,Month,Sales East,Jan,120 East,Feb,150 West,Jan,80 West,Feb,110
Output from agent looks like this:
import pandas as pd import matplotlib.pyplot as plt
Create# dataframe Create from dataframe the from CSV the data CSV datadata =data { = { " Region "":Region ["":East ["",East "",East "",East "",West "",West "",West ""], West "], " Month "":Month ["":Jan ["",Jan "",Feb "",Feb "",Jan "",Jan "",Feb ""], Feb "], " Sales "":Sales [":120 [,120 ,150 ,150 ,80 ,80 ,110 ] 110} ] df} =df pd =.Data pdFrame.Data(dataFrame)
(data#)
Pivot# data Pivot for data visualization for visualizationpivot _dfpivot =_df df =.p dfivot.p(indexivot="(indexMonth="",Month columns",=" columnsRegion="",Region values",=" valuesSales="")
Sales#")
Plot# sales Plot trends sales by trends region by regionpivot _dfpivot.plot_df(kind.plot='(kindline='',line marker',=' markero='') oplt') .titleplt('.titleSales(' TrendsSales by Trends Region by ( RegionJan (-FebJan)-Feb') )plt') .xlabelplt('.xlabelMonth('') Monthplt') .ylabelplt('.ylabelSales('') Salesplt') .gridplt(True.grid) (Trueplt) .legendplt(title.legend='(titleRegion='') Regionplt') .tightplt_layout.tight() _layoutplt() .showplt().show()
Hello @clement-lucas , Thanks for raising this issue. We will look into this and let you know on this.
Hey @clement-lucas - Thank you for pointing that out. This functionality isn’t available in the current version. We’ll connect with the engineering team regarding this and keep you updated.
