File size: 4,519 Bytes
141cc25
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# Code Routes Documentation

This document describes the API endpoints available for code execution, editing, fixing, and cleaning operations in the Auto-Analyst backend.

## Base URL

All code-related endpoints are prefixed with `/code`.

## Endpoints

### Execute Code
Executes Python code against the current session's dataframe.

**Endpoint:** `POST /code/execute`

**Request Body:**
```json
{
    "code": "string",              // Python code to execute
    "session_id": "string",        // Optional session ID
    "message_id": 123              // Optional message ID for tracking
}
```

**Response:**
```json
{
    "output": "string",            // Execution output
    "plotly_outputs": [            // Optional array of plotly outputs
        "string"
    ]
}
```

**Error Responses:**
- `400 Bad Request`: No dataset loaded or no code provided
- `500 Internal Server Error`: Execution error

### Edit Code
Uses AI to edit code based on user instructions.

**Endpoint:** `POST /code/edit`

**Request Body:**
```json
{
    "original_code": "string",     // Code to be edited
    "user_prompt": "string"        // Instructions for editing
}
```

**Response:**
```json
{
    "edited_code": "string"        // The edited code
}
```

**Error Responses:**
- `400 Bad Request`: Missing original code or editing instructions
- `500 Internal Server Error`: Editing error

### Fix Code
Uses AI to fix code with errors, employing a block-by-block approach with DSPy refinement.

**Endpoint:** `POST /code/fix`

**Request Body:**
```json
{
    "code": "string",              // Code containing errors
    "error": "string"              // Error message to fix
}
```

**Response:**
```json
{
    "fixed_code": "string"         // The fixed code
}
```

**Error Responses:**
- `400 Bad Request`: Missing code or error message
- `500 Internal Server Error`: Fixing error

### Clean Code
Cleans and formats code by organizing imports and ensuring proper code block formatting.

**Endpoint:** `POST /code/clean-code`

**Request Body:**
```json
{
    "code": "string"               // Code to clean
}
```

**Response:**
```json
{
    "cleaned_code": "string"       // The cleaned code
}
```

**Error Responses:**
- `400 Bad Request`: No code provided
- `500 Internal Server Error`: Cleaning error

### Get Latest Code
Retrieves the latest code from a specific message.

**Endpoint:** `POST /code/get-latest-code`

**Request Body:**
```json
{
    "message_id": 123              // Message ID to retrieve code from
}
```

**Response:**
```json
{
    "code": "string"               // The retrieved code
}
```

**Error Responses:**
- `400 Bad Request`: Missing message ID
- `404 Not Found`: Message not found
- `500 Internal Server Error`: Retrieval error

## Code Processing Features

### Import Organization
The code processing system automatically:
- Moves all import statements to the top of the file
- Deduplicates imports
- Sorts imports alphabetically

### Code Block Management
The system supports code blocks marked with special comments:
- Start marker: `# agent_name code start`
- End marker: `# agent_name code end`

### Error Handling with DSPy Refinement
When fixing code, the system uses DSPy's refinement mechanism:
- Identifies specific code blocks with errors
- Processes error messages to extract relevant information
- Uses a scoring function to validate fixes
- Employs iterative refinement with up to 3 attempts
- Fixes each block individually while maintaining the overall structure
- Preserves code block markers and relationships

### Dataset Context
When editing or fixing code, the system provides context about the current dataset including:
- Number of rows and columns
- Column names and data types
- Null value counts
- Sample values for each column

### Code Execution Safety
The execution system includes safety measures:
- Removes blocking calls like `plt.show()`
- Handles `__main__` block extraction
- Cleans up print statements with unwanted newlines
- Executes code in isolated namespaces

## Session Management
All endpoints require a valid session ID, which is used to:
- Access the current dataset
- Maintain state between requests
- Track code execution history
- Store execution results for analysis

## Error Handling
The system provides detailed error messages while maintaining security by:
- Logging errors for debugging
- Returning user-friendly error messages
- Preserving original code in case of processing failures
- Using code scoring to validate fixes before returning results