fix bugs

.gitignore

@@ -0,0 +1,2 @@
+__pycache__
+logs

agentverse/agents/simulation_agent/conversation.py

@@ -66,7 +66,7 @@ class ConversationAgent(BaseAgent):
                 raise
             except Exception as e:
                 logger.error(e)
-                logger.warning("Retrying...")
+                logger.warn("Retrying...")
                 continue
 
         if parsed_response is None:

agentverse/output_parser/output_parser.py

@@ -134,13 +134,14 @@ class NlpClassroom9PlayersParser(OutputParser):
         cleaned_output = re.sub(r"\n+", "\n", cleaned_output)
         cleaned_output = cleaned_output.split("\n")
         if not (
-            len(cleaned_output) == 2
+            len(cleaned_output) >= 2
             and cleaned_output[0].startswith("Action:")
             and cleaned_output[1].startswith("Action Input:")
         ):
             raise OutputParserError(text)
         action = cleaned_output[0][len("Action:") :].strip()
-        action_input = cleaned_output[1][len("Action Input:") :].strip()
+        # action_input = cleaned_output[1][len("Action Input:") :].strip()
+        action_input = "\n".join(cleaned_output[1:]).strip()[len("Action Input:") :].strip()
         if action == "Speak":
             return AgentFinish({"output": action_input}, text)
         elif action == "CallOn":
@@ -216,13 +217,13 @@ class PrisonerDilemmaParser(OutputParser):
         cleaned_output = re.sub(r"\n+", "\n", cleaned_output)
         cleaned_output = cleaned_output.split("\n")
         if not (
-            len(cleaned_output) == 2
+            len(cleaned_output) >= 2
             and cleaned_output[0].startswith("Action:")
             and cleaned_output[1].startswith("Action Input:")
         ):
             raise OutputParserError(text)
         action = cleaned_output[0][len("Action:") :].strip()
-        action_input = cleaned_output[1][len("Action Input:") :].strip()
+        action_input = "\n".join(cleaned_output[1:]).strip()[len("Action Input:") :].strip()
 
         if action == "Speak":
             # make sure the police count the round right

agentverse/tasks/simulation/nlp_classroom_9players/config.yaml

@@ -26,7 +26,7 @@ prompts:
     Here is the conversation history
     ${chat_history}
 
-    Remember to pay attention to the response format instructions, and strictly follow the rules specified above! 
+    Remember to pay attention to the response format instructions, and strictly follow the rules specified above! Only output ONE action and action input in your response!
     You should give your response based on the above history. What will you, ${agent_name}, do next?
 
   student_prompt: &student_prompt |-
@@ -60,7 +60,7 @@ prompts:
     Here is the conversation history
     ${chat_history}
 
-    Remember to pay attention to the response format instructions, and strictly follow the rules specified above! 
+    Remember to pay attention to the response format instructions, and strictly follow the rules specified above! Only output ONE action and action input in your response!
     You should give your response based on the above history. What will you, ${agent_name}, do next?
 
 
@@ -96,7 +96,7 @@ agents:
     prompt_template: *professor_prompt
     llm:
       llm_type: gpt-4
-      model: "gpt-4"
+      model: "gpt-4o"
       temperature: 0.7
       max_tokens: 250
     output_parser:
@@ -112,7 +112,7 @@ agents:
       memory_type: chat_history
     llm:
       llm_type: gpt-4
-      model: "gpt-4"
+      model: "gpt-4o"
       temperature: 0.7
       max_tokens: 100
     output_parser:
@@ -126,7 +126,7 @@ agents:
       memory_type: chat_history
     llm:
       llm_type: gpt-4
-      model: "gpt-4"
+      model: "gpt-4o"
       temperature: 0.7
       max_tokens: 100
     output_parser:
@@ -140,7 +140,7 @@ agents:
       memory_type: chat_history
     llm:
       llm_type: gpt-4
-      model: "gpt-4"
+      model: "gpt-4o"
       temperature: 0.7
       max_tokens: 100
     output_parser:
@@ -154,7 +154,7 @@ agents:
       memory_type: chat_history
     llm:
       llm_type: gpt-4
-      model: "gpt-4"
+      model: "gpt-4o"
       temperature: 0.7
       max_tokens: 100
     output_parser:
@@ -168,7 +168,7 @@ agents:
       memory_type: chat_history
     llm:
       llm_type: gpt-4
-      model: "gpt-4"
+      model: "gpt-4o"
       temperature: 0.7
       max_tokens: 100
     output_parser:
@@ -182,7 +182,7 @@ agents:
       memory_type: chat_history
     llm:
       llm_type: gpt-4
-      model: "gpt-4"
+      model: "gpt-4o"
       temperature: 0.7
       max_tokens: 100
     output_parser:
@@ -196,7 +196,7 @@ agents:
       memory_type: chat_history
     llm:
       llm_type: gpt-4
-      model: "gpt-4"
+      model: "gpt-4o"
       temperature: 0.7
       max_tokens: 100
     output_parser:
@@ -210,7 +210,7 @@ agents:
       memory_type: chat_history
     llm:
       llm_type: gpt-4
-      model: "gpt-4"
+      model: "gpt-4o"
       temperature: 0.7
       max_tokens: 100
     output_parser:

agentverse/tasks/simulation/prisoner_dilemma/config.yaml

@@ -53,7 +53,7 @@ agents:
     prompt_template: *prompt
     llm:
       model: "gpt-4"
-      llm_type: gpt-4
+      llm_type: gpt-4o
       temperature: 1.2
       max_tokens: 200
     output_parser:
@@ -75,7 +75,7 @@ agents:
     prompt_template: *prompt
     llm:
       model: "gpt-4"
-      llm_type: gpt-4
+      llm_type: gpt-4o
       temperature: 1.2
       max_tokens: 100
     output_parser:
@@ -97,7 +97,7 @@ agents:
     prompt_template: *prompt
     llm:
       model: "gpt-4"
-      llm_type: gpt-4
+      llm_type: gpt-4o
       temperature: 1.2
       max_tokens: 100
     output_parser:

app.py

@@ -130,10 +130,10 @@ class GUI:
             gr.Box.update(visible=any(self.solution_status)),
         )
 
-    def delay_reset(self, task_dropdown, api_key_text, organization_text):
+    def delay_reset(self, task_dropdown, api_key_text, organization_text, api_base_text):
         self.autoplay = False
         self.image_now, self.text_now = self.reset(
-            task_dropdown, api_key_text, organization_text
+            task_dropdown, api_key_text, organization_text, api_base_text
         )
         return (
             self.image_now,
@@ -150,9 +150,11 @@ class GUI:
         task_dropdown="simulation/nlp_classroom_9players",
         api_key_text="",
         organization_text="",
+        api_base_text=""
     ):
         openai.api_key = api_key_text
         openai.organization = organization_text
+        openai.api_base = api_base_text if api_base_text else None
         """
         tell backend the new number of students and generate new empty image
         :param stu_num:
@@ -472,6 +474,7 @@ class GUI:
                     )
                     api_key_text = gr.Textbox(label="OPENAI API KEY")
                     organization_text = gr.Textbox(label="Organization")
+                    api_base_text = gr.Textbox(label="OpenAI Base URL", default="", placeholder="if not set, will use openai's default url")
                 with gr.Row():
                     with gr.Column():
                         image_output = gr.Image()
@@ -550,7 +553,7 @@ class GUI:
                 # reset_btn.click(fn=self.reset, inputs=None, outputs=[image_output, text_output], show_progress=False)
                 reset_btn.click(
                     fn=self.delay_reset,
-                    inputs=[task_dropdown, api_key_text, organization_text],
+                    inputs=[task_dropdown, api_key_text, organization_text, api_base_text],
                     outputs=[
                         image_output,
                         text_output,