fastAPI 심화

- Chart.js
- pdf, csv 파일 업로드 후 데이터 정제하여 llm으로 처리 후 결과 도출
- sqlite로 데이터 저장
- ORM - SQLAlchemy

project/CREDIT_APP/backend/services/card_service.py

@@ -0,0 +1,140 @@
+from fastapi import UploadFile
+from langchain_core.runnables import RunnablePassthrough
+
+from backend.services.db_service import get_connection
+from backend.services.db_service import get_table_columns
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.output_parsers import StrOutputParser
+from backend.ai.llm import watson_llm
+import csv
+
+async def upload_csv(file :UploadFile):
+    """
+    csv 파일을 테이블에 저장
+    """
+
+    conn =get_connection()
+    cursor = conn.cursor()
+
+    # file.read() : 비동기 함수임
+    contents = await file.read()
+    csv_text = contents.decode('utf-8')
+    reader = csv.DictReader(csv_text.splitlines())
+    count = 0
+
+    for row in reader:
+        cursor.execute("""
+                        INSERT INTO transactions (date, category, merchant, amount) 
+                        VALUES (?, ?,?, ?)"""
+                        ,
+                        (row['date'], row['category'], row['merchant'], row['amount'])
+                    )
+        count += 1
+
+    conn.commit()
+    conn.close()
+    return {"message": f"{count} 건 저장 완료"}
+
+def card_history():
+    """
+    db에서 카드 정보 조회
+    """
+    conn =get_connection()
+    cursor = conn.cursor()
+
+    try :
+        cursor.execute("""
+                       SELECT * FROM transactions ORDER BY date DESC
+                       """)
+        rows = cursor.fetchall()
+        query_result = [dict(row) for row in rows]
+
+        conn.close()
+    except Exception as e:
+        query_result = [f"SQL 실행 오류 : {e}"]
+
+    return query_result
+
+def get_dashboard():
+    """
+    db에서 대시보드 정보 조회
+    """
+    conn = get_connection()
+    cursor = conn.cursor()
+
+    # 카테고리 별 사용금액
+    cursor.execute(
+        """
+        SELECT category, SUM(amount)
+        FROM transactions
+        GROUP BY category
+        ORDER BY SUM(amount) DESC
+        """
+    )
+
+    category_rows = cursor.fetchall()
+
+    # 월별 사용금액
+    cursor.execute(
+        """
+        SELECT strftime('%Y-%m', date), SUM(amount)
+        FROM transactions
+        GROUP BY strftime('%Y-%m', date)
+        ORDER BY strftime('%Y-%m', date)
+        """
+    )
+    month_rows = cursor.fetchall()
+    conn.close()
+
+    return  {
+        "category": [{"category":row[0], "amount":row[1]} for row in category_rows],
+        "monthly": [{"month":row[0], "amount":row[1]} for row in month_rows],
+    }
+
+def sql_generate_llm(question):
+    """자연어 -> SQL"""
+    sql_prompt = ChatPromptTemplate.from_template("""
+    당신은 SQLite 전문가입니다.
+    
+    테이블명 : transactions
+    
+    컬럼 : {columns}
+    
+    질문 : {question}
+    
+    SQL 만 출력하세요
+    """)
+
+    columns = get_table_columns("transactions")
+    sql_chain = sql_prompt | watson_llm | StrOutputParser()
+
+    sql = sql_chain.invoke({"columns": columns, "question": question})
+    print(f"sql: {sql}")
+
+    sql = sql.replace("```sql","").replace("```", "").strip()
+
+    # sql 문 실제 실행
+    conn = get_connection()
+    cursor = conn.cursor()
+    cursor.execute(sql)
+    rows = cursor.fetchall()
+    query_result = [dict(row) for row in rows]
+    conn.close()
+
+    return query_result
+
+def card_analysis(question):
+    query_result = sql_generate_llm(question)
+
+    analysis_prompt = ChatPromptTemplate.from_template("""
+    사용자 질문 : {question}
+
+    SQL 결과
+    {result}
+
+    결과를 설명하고 소비 습관을 분석하고 절약 팁을 제시해 주세요.
+    """)
+    analysis_chain = analysis_prompt | watson_llm | StrOutputParser()
+    answer = analysis_chain.invoke({"question": question, "result": query_result})
+
+    return {"message": answer}