Update app.py

app.py

@@ -9,7 +9,7 @@ import requests
 
 
 # define the text summarizer function
-def text_prompt(request, system_role, page_urls_str, contrasenya, api_base, deployment_id, temp):
+def text_prompt(request, system_role, page_urls_str, api_key, api_base, deployment_id, temp):
     tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
 
     page_urls = [page_url_str for page_url_str in page_urls_str.split("\n") if page_url_str]
@@ -44,17 +44,22 @@ def text_prompt(request, system_role, page_urls_str, contrasenya, api_base, depl
             if len(tokens) > 3100:
                 break
             page_text += sentence + ". "
-
+        tokens.extend(tokenizer.tokenize(request))
+        tokens.extend(tokenizer.tokenize(system_role))
         # Delete the last space
         page_text = page_text.strip()
 
         num_tokens = len(tokens)
 
-        if num_tokens > 10 and len(contrasenya) > 6 and len(deployment_id) > 6:
+        tokens_condition = num_tokens > 10
+        api_key_condition = len(api_key) > 6
+        deployment_id_condition = len(deployment_id) > 6
+        if tokens_condition and api_key_condition and deployment_id_condition:
             openai.api_type = "azure"
-            openai.api_version = "2023-03-15-preview"
+            openai.api_version = "2023-05-15"
             openai.api_base = api_base
-            openai.api_key = contrasenya
+            openai.api_key = api_key
+            max_tokens = 4000 - num_tokens  # TODO: change 4096 to a dictionary with the max tokens for each deploymend_id
             # get the response from openai API
             try:
                 response = openai.ChatCompletion.create(
@@ -63,7 +68,7 @@ def text_prompt(request, system_role, page_urls_str, contrasenya, api_base, depl
                         {"role": "system", "content": system_role},
                         {"role": "user", "content": request + "\n\n" + 'Text:\n\n""""' + page_text + '\n""""'}
                     ],
-                    max_tokens=4000,
+                    max_tokens=max_tokens,
                     temperature=temp,
                     top_p=1.0,
                 )
@@ -73,16 +78,33 @@ def text_prompt(request, system_role, page_urls_str, contrasenya, api_base, depl
 
                 # clean the response text
                 response_text = re.sub(r'\s+', ' ', response_text)
-                response_text = f"#### [{page.title}]({page_url})\n\n{response_text.strip()}"
+                response_text = f"#### [{page.title}]({page_url})\n\n{response_text.strip()}\n"
 
-                page_texts.append(page.text)
-                response_texts.append(response_text)
             except Exception as e:
-                return page.text, f"--- An error occurred while processing the request: {e} ---", num_tokens
-        else:
-            return page.text, "--- Check API-Key or deployment_id or Min number of tokens:", str(num_tokens)
+                response_text = f"#### [{page.title}]({page_url})\n\n"
+                response_text += f"--- An error occurred while processing the request: {e} ---\n"
+            page_texts.append(page.text)
+            response_texts.append(response_text)
 
-    page_texts_str = "\n\n============\n".join([page_text for page_text in page_texts])
+        else:
+            page_text_temp = "ERROR:\n\n"
+            if page.text:
+                page_text_temp += page.text
+
+            response_text_temp = "#### "
+            if page.title:
+                response_text_temp += f"[{page.title}]({page_url})"
+            if not tokens_condition:
+                response_text_temp += "\n\nERROR: Tokens problems! Maybe it can't read the URL. "
+            if not api_key_condition:
+                response_text_temp += "\n\nERROR: API Key problems! Copy and paste the API Key (be careful with copying spaces at the beginning or end of the API Key). "
+            if not deployment_id_condition:
+                response_text_temp += "\n\nERROR: Deployment_id problems! Copy and paste the deployment_id (be careful with copying spaces at the beginning or end of the deployment_id). "
+
+            page_texts.append(page_text_temp)
+            response_texts.append(response_text_temp)
+
+    page_texts_str = "".join([f"====== NEW URL: {URL} ======\n{page_text}\n\n" for page_text, URL in zip(page_texts, page_urls)])
     response_texts_str = "\n\n".join([response_text for response_text in response_texts])
     total_tokens_str = str(total_tokens) + " (${:.2f} USD)".format(total_tokens / 1000 * 0.03)
     return page_texts_str, response_texts_str, total_tokens_str
@@ -95,7 +117,7 @@ iface = gr.Interface(
             gr.Textbox(lines=1, placeholder="Enter your system-role description here...", label="System Role:", type="text"),
             gr.Textbox(lines=10, placeholder="Enter the Articles' URLs here...", label="Articles' URLs to parse (one per line up to 10):", type="text"),
             gr.Textbox(lines=1, placeholder="Enter your API-key here...", label="API-Key:", type="password"),
-            gr.Textbox(lines=1, placeholder="Enter your Azure OpenAI endpoint here...", label="Endpoint:", type="text"),
+            gr.Textbox(lines=1, placeholder="Enter your Azure OpenAI API base here...", label="Enter Azure API base (Endpoint):", type="text"),
             gr.Textbox(lines=1, placeholder="Enter your model name here...", label="Deployment ID:", type="text"),
             gr.Slider(0.0, 1.0, value=0.0, label="Temperature:")
             ],