Hugging Face's logo

Spaces:
Shrouk04
/
data_analysis_agent2
Sleeping

App Files Community
data_analysis_agent2
File size: 2,942 Bytes
b814c5a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4c2e99b
b814c5a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import ollama


SYSTEM_PROMPT = """

You are an expert Python Data Analyst.



You are given:

- A pandas DataFrame named df

- Dataset metadata

- Conversation history

- A user question



Generate ONLY executable Python code.



Rules:

1. Use only pandas and numpy.

2. Store final answer in variable named result.

3. Do not print().

4. Do not explain.

5. Do not use markdown.

6. Return code only.



Examples:



Question:

How many rows are in the dataset?



Code:

result = len(df)



Question:

How many columns are there?



Code:

result = len(df.columns)



Question:

List all columns



Code:

result = list(df.columns)



Question:

What is the average sales?



Code:

result = df["Sales"].mean()



Question:

What is the maximum sales?



Code:

result = df["Sales"].max()



Question:

What is the minimum sales?



Code:

result = df["Sales"].min()



Question:

Show first 5 rows



Code:

result = df.head()



Question:

What percentage of rows have missing values?



Code:

result = (df.isnull().any(axis=1).mean()) * 100



Question:

What percentage of missing values does each column have?



Code:

result = (df.isnull().sum() / len(df)) * 100



Question:

Which category has the highest average sales?



Code:

result = (

    df.groupby("Category")["Sales"]

        .mean()

        .sort_values(ascending=False)

        .head(1)

)



Question:

What factors affect price the most?



Code:

numeric_df = df.select_dtypes(include='number')



result = (

    numeric_df.corr()['price']

    .sort_values(ascending=False)

)



"""


def generate_code(

    question,

    metadata,

    memory=None

):
    
    #generate code from 
    

    memory_text = ""

    if memory:
        


        memory_text = ""

        for item in memory[-10:]:

            if item["role"] == "user":
                memory_text += f"User: {item['content']}\n"

            elif item["role"] == "assistant":
                memory_text += f"Assistant: {item['content']}\n"



    prompt = f"""

Dataset Metadata:



Columns:

{metadata['columns']}



Numeric Columns:

{metadata['numeric_columns']}



Categorical Columns:

{metadata['categorical_columns']}



Conversation History:

{memory_text}



Current Question:

{question}



Generate ONLY Python code.



Requirements:

- Use dataframe named df

- Save final output into variable result

- Return code only

"""

    response = ollama.chat(
        model="qwen2.5:3b",
        messages=[
            {
                "role": "system",
                "content": SYSTEM_PROMPT
            },
            {
                "role": "user",
                "content": prompt
            }
        ]
    )

    code = response["message"]["content"]

    # remove markdown if Qwen generates it
    code = code.replace("```python", "")
    code = code.replace("```", "")
    code = code.strip()

    return code