save tmp (need to be fixed)

examples/use-rdb-resource/README.md

@@ -22,3 +22,5 @@
   - `poetry run python make_example_table_data.py`
     - if this command doesn't work, then run `poetry env use 3.12`
       - change python version to resolve dependensies version
+  - `poetry run python main.py`
+    - run main file to answer question by DANA using DbResource, and see the result in the terminal.

examples/use-rdb-resource/main.py

@@ -24,6 +24,7 @@ def solve(question, query) -> str:
     )
 
     query = generate_sql_from_prompt(QUESTION)
+    print(query)
     answer = solve(QUESTION, query)
 
     print('--------------------------------')

examples/use-rdb-resource/pyproject.toml

@@ -15,6 +15,7 @@ faker = "^30.1.0"
 sqlalchemy = "^2.0.35"
 openai = "^1.51.2"
 openssa = "^0.24.10.10"
+plotly = "^5.24.1"
 
 
 [build-system]

openssa/core/resource/db.py

@@ -14,6 +14,7 @@
 import os
 from sqlalchemy import create_engine, text
 from sqlalchemy.orm import sessionmaker
+from llama_index.core import SummaryIndex, Document
 
 from .abstract import AbstractResource
 from ._global import global_register
@@ -70,11 +71,15 @@ def fetch_data(self) -> list[tuple[Any]]:
     def answer(self, question: str, n_words: int = 1000) -> str:
         """Answer question from database-stored Informational Resource."""
         data = self.fetch_data()
-        # Here you can implement a more sophisticated way to generate answers from the data
-        # For simplicity, we will just return the fetched data as a string
-        return str(data)
-
-    def __del__(self):
-        """Ensure the database connection is closed when the object is deleted."""
-        if hasattr(self, 'db'):
-            self.db.Session.close_all()
+        print(data)
+        # データベースから取得したデータをドキュメントに変換
+        documents = [Document(text=str(row[0]), metadata={'id': row[1]}) for row in data]
+        # print(documents)
+        index = SummaryIndex.from_documents(documents)
+        # print(index)
+        # set Logging to DEBUG for more detailed outputs
+        query_engine = index.as_query_engine()
+        # print(query_engine)
+        response = query_engine.query(question)
+        # print(response)
+        return response

tests/core/resource/test_db_resource.py

@@ -3,16 +3,23 @@
 
 load_dotenv()
 
+#TODO: Fix hallucination
+# Given Data: [(1, 'Laptop', 100000), (2, 'Smartphone', 60000), (3, 'Headphones', 8000), (4, 'Keyboard', 3000), (5, 'Mouse', 2000), (6, 'Monitor', 25000), (7, 'Tablet', 50000), (8, 'Smartwatch', 20000), (9, 'Camera', 45000), (10, 'Speaker', 15000)]
+# Answer: The item that is the most expensive from the given data is the Camera.
+# The Answer Should be: Laptop(100000).
+# How to fix it? Make query by vanna or change the process in llama_index?
 
 def test_db_resource():
-    test_url = "http://paulgraham.com/worked.html"
-    test_question = "What did the author do growing up?"
+    test_query = "SELECT * FROM items"
+    test_question = "Which item is the most expensive from given data?"
 
-    webpage1 = WebPageResource(test_url)
-    # print(f"unique name = {webpage1.name}")
-    # print(f"unique name = {webpage1.unique_name}")
-    # print(f"answer = {webpage1.answer(test_question)}")
-    # print(f"summary = {webpage1.get_summary()}")
-    _ = webpage1.answer(test_question)
-    _ = webpage1.get_summary()
+    rdb1 = DbResource(query=test_query)
+    # print(f"unique name = {rdb1.name}")
+    # print(f"unique name = {rdb1.unique_name}")
+    # print(f"answer = {rdb1.answer(test_question)}")
+    # print(f"summary = {rdb1.get_summary()}")
+    _ = rdb1.answer(test_question)
     # assert isinstance(answer, str)
+    print(_)
+
+test_db_resource()