panda query

app.py

@@ -10,6 +10,7 @@ For more information on `huggingface_hub` Inference API support, please check th
 
 import pandas as pd
 import pandasql
+from collections import defaultdict
 
 # Create a sample DataFrame
 data = [
@@ -19,8 +20,11 @@ data = [
     {"Name": "Alice", "Age": 24, "Gender": "female", "Votes": 120},
 ]
 
+# types = defaultdict(str,'Votes'=int,'Votes Percentage'=float)
+# {'Votes':int,'Votes Percentage':float}
+#
 # Create a pandas dataframe from the list of dictionaries
-df = pd.DataFrame(data)
+df = pd.read_csv('election_results.csv',dtype={'Votes':'int64'})
 
 
 def respond(
@@ -32,7 +36,7 @@ def respond(
     messages = [
         {
             "role": "system",
-            "content": "Your task is to convert the input query into a sql statement to be used against a panda dataframe.\n\nGiven the below columns, \n\nColumn1: Age\nColumn2: Name\nColumn3: Gender\nColumn4: Votes\nColumn5: Location\nColumn6: Party\n\n and Table name as df \n and the user input text, \n\nconvert it into a proper sql statement.\n\nIn the where condition make sure you do a case insensitive comparison for text columns, and where possible use like, instead of 'equal' condition.  Also when you compare with text always use a lowercase, for example use 'female', not 'Female'. \n\nOutput format:\nIn the response give only the SQL statement starts with 'SELECT', do not add any note or any other explanations"
+            "content": "You are an election result analysis bot. Your task is to convert the user's natural language query into a SQL SELECT statement suitable for a pandas DataFrame named df.\n\nAvailable Columns:\n\nCandidate (string)\nParty (string)\nVotes (integer)\nVotes Percentage (decimal)\nConstituency (string)\nState (string)\n\nCase-Insensitive Text Comparisons:\n\nUse UPPER function to convert text columns (Candidate, Party, Constituency, State) to uppercase for case-insensitive comparisons.\nEmploy LIKE operator with wildcards (%) for pattern matching when appropriate.\n\nUser Input:\n\nThe system will provide the user's query in natural language.\n\nOutput Format:\n\nGenerate the SQL SELECT statement only, starting with SELECT. Do not add any other extra instruction text, your output will be directly executed, so send only the SQL Statement\n\nExample:\n\nUser Input: Find all candidates from Coimbatore who received more than 50% of the votes.\n\nYour Output:\nSELECT * FROM df\nWHERE UPPER(State) = 'COIMBATORE' AND Votes Percentage > 50"
         }
     ]
     for val in history:
@@ -46,15 +50,20 @@ def respond(
         model="llama3-70b-8192",
         messages=messages,
         temperature=1,
-        max_tokens=1024,
+        max_tokens=2048,
         top_p=1,
         stream=False,
         stop=None,
     )
 
     sql_command = completion.choices[0].message.content
-    result = pandasql.sqldf(sql_command, globals())
-    yield str(result)
+    print(sql_command)
+
+    if  sql_command.startswith('SELECT'):
+        result = pandasql.sqldf(sql_command, globals())
+        yield result.to_string()
+    else:
+        yield str(sql_command)
 
 """
 For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface