探秘 LLM Agents：ReAct 框架藏着哪些惊喜？

import sqlite3def create_sample_database():    """    Create a simple SQLite database with a 'companies' table and a 'customers' table.    Each company has an id, title, and description.    Each customer has an id, name, revenue, and a foreign key company_id referencing the companies table.    This function resets the database for a clean slate.    """    conn = sqlite3.connect("sample.db")    cur = conn.cursor()        # Enable foreign key support in SQLite.    cur.execute("PRAGMA foreign_keys = ON;")        # Create the companies table.    cur.execute("""        CREATE TABLE companies (            id INTEGER PRIMARY KEY,            title TEXT NOT NULL,            description TEXT        )    """)        # Create the customers table with a foreign key to companies.    cur.execute("""        CREATE TABLE customers (            id INTEGER PRIMARY KEY,            name TEXT NOT NULL,            revenue REAL NOT NULL,            company_id INTEGER,            FOREIGN KEY (company_id) REFERENCES companies(id) ON DELETE SET NULL        )    """)        # Insert sample data into companies.    companies_data = [        (1, 'Acme Corp', 'Leading provider of innovative solutions'),        (2, 'Beta Inc', 'Specializes in data analytics and insights'),        (3, 'Gamma LLC', 'Pioneer in cloud computing services')    ]    cur.executemany("INSERT INTO companies (id, title, description) VALUES (?, ?, ?)", companies_data)        # Insert sample data into customers.    customers_data = [        (1, 'Alice', 1500.0, 1),        (2, 'Bob', 2400.0, 2),        (3, 'Charlie', 1800.0, 1),        (4, 'Diana', 3200.0, 3),        (5, 'Ethan', 2100.0, 2)    ]    cur.executemany("INSERT INTO customers (id, name, revenue, company_id) VALUES (?, ?, ?, ?)", customers_data)        conn.commit()    return conn

system_prompt = f"""You are a helpful assistant that translates natural language queries into SQL queries for the following SQLite database schema.Table "companies":  - id: INTEGER PRIMARY KEY  - title: TEXT NOT NULL  - description: TEXTTable "customers":  - id: INTEGER PRIMARY KEY  - name: TEXT NOT NULL  - revenue: REAL NOT NULL  - company_id: INTEGER (Foreign key referencing companies(id))Instructions:  1. When you receive a text query that is relevant to this schema, generate the corresponding SQL query. Return only the SQL query with no additional commentary or characters like ``` or quotes e.g. SELECT * FROM COMPANIES.  2. If you receive a text query that includes an error message, generate a corrected SQL query that fixes the error, again returning only the SQL query.  3. If the query is irrelevant, do not output any extra text."""

import timedef test():    nl_request = input("Enter natural language query: ").strip()    if not nl_request:        print("No natural language query provided.")        return    conn = sqlite3.connect("sample.db")    max_iterations = 5    iteration = 0    feedback = None    final_result = None    while iteration < max_iterations:        iteration += 1        print(f"--- Iteration {iteration} ---")        sql_query = generate_sql_query(nl_request, feedback)        if not sql_query:            print("No corresponding SQL query for this query.")            break        result, error = execute_query(conn, sql_query)                if error:            print(f"SQL execution error: {error}")            feedback = f"The query resulted in the following error: {error}. Please adjust the SQL query accordingly."            time.sleep(1)        else:            if not result:                print("No results found.")            else:                print("Query executed successfully. Results:")                for row in result:                    print(row)            final_result = result            break    if final_result is None:        print("Failed to generate a valid SQL query after several iterations.")    conn.close()