File size: 2,898 Bytes
05e3517
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import json
import openai
from config import PROJECT_ID, DATASET_ID
from utils.bigquery_utils import get_bigquery_schema_info

def table_selection_agent(state):
    """Identifies relevant tables for the natural language query based on schema."""
    natural_language_query = state["sql_query"]
    client = state["client"]
    
    if client is None:
        return {"relevant_tables": [], "error": "Failed to connect to BigQuery."}
    
    schema_info = get_bigquery_schema_info(client, PROJECT_ID, DATASET_ID)
    
    # Format the schema for the prompt
    schema_text = ""
    for table_name, columns in schema_info.items():
        schema_text += f"- **{DATASET_ID}.{table_name}** ({', '.join(columns)})\n"
    
    prompt = f"""
    Based on the following natural language query and BigQuery schema, identify the tables that would be needed to answer the query.
    
    **Query:** "{natural_language_query}"
    
    **BigQuery Schema:**
    {schema_text}
    
    Analyze the query and determine which tables contain the necessary information.
    
    IMPORTANT: Return ONLY a raw JSON array of table names without any markdown formatting, code blocks, or explanations.
    
    Example of correct response format:
    ["{DATASET_ID}.users", "{DATASET_ID}.orders"]
    
    Example of INCORRECT response format:
    ```json
    ["{DATASET_ID}.users", "{DATASET_ID}.orders"]
    ```
    
    DO NOT use code blocks, backticks, or any other formatting. Return ONLY the raw JSON array.
    """
    
    try:
        response = openai.chat.completions.create(
            model="gpt-4o",
            messages=[{"role": "user", "content": prompt}],
            temperature=0.0
        )
        
        # Get the content from the response
        content = response.choices[0].message.content.strip()
        
        # Remove markdown code block formatting if present
        if content.startswith("```"):
            # Extract content between the code block markers
            parts = content.split("```")
            if len(parts) >= 3:  # There should be at least 3 parts if there are code blocks
                content = parts[1]
                # If there's a language identifier (like json), remove it
                if content.startswith("json"):
                    content = content.replace("json", "", 1).strip()
        
        # Parse the JSON
        relevant_tables = json.loads(content)
        print(f"Parsed relevant tables: {relevant_tables}")
        return {"relevant_tables": relevant_tables}
    except json.JSONDecodeError as e:
        print(f"JSON Decode Error: {e}")
        print(f"Response content: {response.choices[0].message.content}")
        return {"relevant_tables": [], "error": "Invalid JSON response from OpenAI"}
    except Exception as e:
        print(f"Unexpected error: {e}")
        return {"relevant_tables": [], "error": f"Error: {str(e)}"}