Update app.py

app.py

@@ -10,93 +10,49 @@ tokenizer = None
 generator = None
 csv_name = "disease_database_mini.csv"
 df = pd.read_csv(csv_name)
-openai.api_key = "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"
+openai.api_key = "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"
 
-def csv_prompter(question,csv_name):
+def csv_prompter(question,csv_name):    
+    json_file = open('order.json')
+    json_data = json.load(json_file)
+    json_data = json_data['records']
     
+    print(json_data)
+    fulltext = []
+    #print all nm in json file
+    for i in json_data:
     
-
-    fulltext = "A question is provided below. Given the question, extract " + \
-               "keywords from the text. Focus on extracting the keywords that we can use " + \
-               "to best lookup answers to the question. \n" + \
-               "---------------------\n" + \
-               "{}\n".format(question) + \
-               "---------------------\n" + \
-               "Provide keywords in the following comma-separated format.\nKeywords: "
-
+        # identify if opts in this item
+        opt_list = []
+        if 'opt' in i:
+            for opt1 in i['opt']:
+                for opt2 in opt1['opts']:
+                    opt_list.append(opt2['nm'])
+        if len(opt_list) > 100:
+            print(str(i['pid'])+" "+i['nm']+" Options:"+str(opt_list))
+            each = i['nm']+" Options:"+str(opt_list)
+            fulltext.append(each)
+        else:
+            print(str(i['pid'])+" "+i['nm'])
+            each = i['nm']
+            fulltext.append(each)
+    
+    
+    fulltext = '\n'.join(fulltext)
+    #fulltext = fulltext + "\n 00:00 - Customer: Hey, dear, can I please take out? 00:03 - Waiter: I'm up here. Okay go ahead. 00:06 - Customer: Can I get two sesame chicken dinners? 00:10 - Customer: One with lo mein and one with fried rice? 00:13 - Customer: Can I get, I know spring rolls come with each of those, but can I get just as long 00:17 - Customer: as I've got a total of two spring rolls and two egg rolls? 00:21 - Waiter: Okay. Anything? 00:22 - Customer: And a small bowl of hot and sour soup. 00:25 - Customer: And that's it. 00:26 - Waiter: Alright. Two sesame chickens, small one with fried rice, small one with lo mein, and both of those with spring rolls, two egg rolls and a small hot and sour soup. 00:30 - Customer: It's 205-473-1750. 00:36 - Waiter: What's the fun number please? 00:43 - Waiter: Just got around 20 minutes. Thank you. 00:43 - Customer: Okay, dear, thank you so much."
+    fulltext = fulltext+'\n'
+    #read txt file in lines
+    f = open("talk.txt", "r")
+    for x in f:
+        fulltext = fulltext + x
+    
+    
+    
+    fulltext = fulltext+'Based on the above dialogue and menu, output the dishes ordered by the customer. Note that some dishes have sub-options (e.g. some are set meals, some are a la carte)'
+    print(fulltext)
     messages = [
-        {"role": "system", "content": ""},
-    ]
-    messages.append(
-        {"role": "user", "content": f"{fulltext}"}
-    )
-    rsp = openai.ChatCompletion.create(
-        model="gpt-3.5-turbo",
-        messages=messages
-    )
-    keyword_list = rsp.get("choices")[0]["message"]["content"]
-    keyword_list = keyword_list.replace(",","").split(" ")
-
-    print(keyword_list)
-    divided_text = []
-    csvdata = df.to_dict('records')
-    step_length = 15
-    for csv_item in range(0,len(csvdata),step_length):
-        csv_text = str(csvdata[csv_item:csv_item+step_length]).replace("}, {", "\n\n").replace("\"", "")#.replace("[", "").replace("]", "")
-        divided_text.append(csv_text)
-
-    answer_llm = ""
-
-    score_textlist = [0] * len(divided_text)
-
-    for i, chunk in enumerate(divided_text):
-        for t, keyw in enumerate(keyword_list):
-            if keyw.lower() in chunk.lower():
-                score_textlist[i] = score_textlist[i] + 1
-
-    answer_list = []
-    divided_text = [item for _, item in sorted(zip(score_textlist, divided_text), reverse=True)]
-
-    for i, chunk in enumerate(divided_text):
-
-        if i>4:
-            continue
-
-        fulltext = "{}".format(chunk) + \
-                   "\n---------------------\n" + \
-                   "Based on the Table above and not prior knowledge, " + \
-                   "Select the Table Entries that will help to answer the question: {}\n Output in the format of \" Disease: <>; Symptom: <>; Medical Test: <>; Medications: <>;\". If there is no useful form entries, output: 'No Entry'".format(question)
-
-        print(fulltext)
-        messages = [
             {"role": "system", "content": ""},
         ]
-        messages.append(
-            {"role": "user", "content": f"{fulltext}"}
-        )
-        rsp = openai.ChatCompletion.create(
-            model="gpt-3.5-turbo",
-            messages=messages
-        )
-        answer_llm = rsp.get("choices")[0]["message"]["content"]
-
-        print("\nAnswer: " + answer_llm)
-        print()
-        if not "No Entry" in answer_llm:
-            answer_list.append(answer_llm)
-
-
-
-    fulltext = "The original question is as follows: {}\n".format(question) + \
-               "Based on this Table:\n" + \
-               "------------\n" + \
-               "{}\n".format(str("\n\n".join(answer_list))) + \
-               "------------\n" + \
-               "Answer: "
-    print(fulltext)
-    messages = [
-        {"role": "system", "content": ""},
-    ]
     messages.append(
         {"role": "user", "content": f"{fulltext}"}
     )
@@ -104,10 +60,10 @@ def csv_prompter(question,csv_name):
         model="gpt-3.5-turbo",
         messages=messages
     )
-    answer_llm = rsp.get("choices")[0]["message"]["content"]
-
-    print("\nFinal Answer: " + answer_llm)
+    response = rsp.get("choices")[0]["message"]["content"]
     print()
+    
+    print(response)
 
     return answer_llm