added beta_models capabilities

OpenAI_GPT.py

@@ -11,16 +11,26 @@
 
 
 #####
-def simpler_gpt_call(apikey, messages, model_engine, max_tokens, temperature, **kwargs):
+def simpler_gpt_call(apikey, messages, model_engine, max_tokens, temperature, beta_model=False, **kwargs):
     client = OpenAI(api_key=apikey)
 
+    # beta models limitation: https://platform.openai.com/docs/guides/reasoning
+    # o1 will may not provide an answer if the max_completion_tokens is lower than 2000
+
+#    with open("msg.json", 'w') as f:
+#        json.dump(messages, f, indent=4)
+
+    if beta_model is True:
+        kwargs['max_completion_tokens'] = max_tokens
+    else:
+        kwargs['max_tokens'] = max_tokens
+        kwargs['temperature'] = temperature
+
     # Generate a response (20231108: Fixed for new API version)
     try:
-        completion = client.chat.completions.create(
+        response = client.chat.completions.create(
             model=model_engine,
             messages = messages,
-            max_tokens=max_tokens,
-            temperature=temperature,
             **kwargs
         )
     # using list from venv/lib/python3.11/site-packages/openai/_exceptions.py
@@ -35,7 +45,9 @@ def simpler_gpt_call(apikey, messages, model_engine, max_tokens, temperature, **
     except openai.OpenAIError as e:
         return(f"OpenAI API request failed: {e}", "")
 
-    return "", completion.choices[0].message.content
+#    with open("response.txt", 'w') as f:
+#        f.write(f"{response}")
+    return "", response.choices[0].message.content
 
 ##########
 class OAI_GPT:
@@ -63,6 +75,7 @@ def __init__(self, apikey, base_save_location, username):
         self.gpt_roles = {}
         self.gpt_roles_help = ""
         self.model_capability = {}
+        self.beta_models = {}
 
 
 #####
@@ -96,6 +109,10 @@ def get_gpt_roles_help(self):
     def get_save_location(self):
         return self.save_location
 
+    def get_beta_models(self):
+        return self.beta_models
+
+
 #####
 # https://platform.openai.com/docs/models/continuous-model-upgrades
     def set_parameters(self, models_list, av_models_list):
@@ -131,6 +148,10 @@ def set_parameters(self, models_list, av_models_list):
                 self.model_capability[key] = models[key]["capability"]
             else:
                 self.model_capability[key] = "None"
+            if 'beta_model' in models[key]:
+                self.beta_models[key] = models[key]['beta_model']
+            else:
+                self.beta_models[key] = False
             if cf.isNotBlank(models[key]["status_details"]):
                 per_model_help += " NOTE: " + models[key]["status_details"]
             self.per_model_help[key] = per_model_help
@@ -310,7 +331,11 @@ def chatgpt_it(self, model_engine, prompt, max_tokens, temperature, clear_chat,
         msg_file = f"{dest_dir}/msg.json"
         with open(msg_file, 'w') as f:
             json.dump(clean_messages, f, indent=4)
-        err, response = simpler_gpt_call(self.apikey, clean_messages, model_engine, max_tokens, temperature, **kwargs)
+
+        beta_model = False
+        if model_engine in self.beta_models:
+            beta_model = True
+        err, response = simpler_gpt_call(self.apikey, clean_messages, model_engine, max_tokens, temperature, beta_model=beta_model, **kwargs)
 
         if cf.isNotBlank(err):
             return err, ""

OpenAI_GPT_WUI.py

@@ -37,6 +37,7 @@ def __init__(self, oai_gpt: OAI_GPT, enable_vision: bool = True, prompt_presets_
         self.gpt_presets = oai_gpt.get_gpt_presets()
         self.gpt_presets_help = oai_gpt.get_gpt_presets_help()
         self.per_model_help = oai_gpt.get_per_model_help()
+        self.beta_models = oai_gpt.get_beta_models()
 
         self.enable_vision = enable_vision
 
@@ -193,6 +194,7 @@ def set_ui(self):
         clear_chat = False
         prompt_preset = None
         msg_extra = None
+        beta_model = False
 
         if 'gpt_last_prompt' in st.session_state:
             if st.session_state['gpt_last_prompt'] != "":
@@ -221,6 +223,8 @@ def set_ui(self):
                     st.info(f"{model}: {self.models_status[model]}")
                 if self.model_capability[model] == "vision":
                     vision_capable = True
+                if model in self.beta_models and self.beta_models[model] is True:
+                    beta_model = True
                 m_token = self.models[model]['max_token']
 
                 # vision mode bypass
@@ -236,12 +240,22 @@ def set_ui(self):
 
                 role = list(self.gpt_roles.keys())[0]
                 if vision_mode is False:
-                    role = st.selectbox("Role", options=self.gpt_roles, index=0, key="input_role", help = "Role of the input text\n\n" + self.gpt_roles_help)
+                    tmp_roles = self.gpt_roles.copy()
+                    if beta_model is True:
+                        tmp_roles.pop("system")
+                    tmp_txt = "" if beta_model is False else "\n\nBeta models do not support the 'system' role"
+                    role = st.selectbox("Role", options=tmp_roles, index=0, key="input_role", help = "Role of the input text\n\n" + self.gpt_roles_help + tmp_txt)
+
+                if beta_model is False:
+                    max_tokens = st.slider('max_tokens', 0, m_token, 1000, 100, "%i", "max_tokens", "The maximum number of tokens to generate in the completion. The token count of your prompt plus max_tokens cannot exceed the model\'s context length.")
+                else:
+                    max_tokens = st.slider('max_completion_tokens', 0, m_token, 4000, 100, "%i", "max_completion_tokens", "For beta models: control the total number of tokens generated by the model, including both reasoning and visible completion tokens. If the token value is too low, the answer might be empty")
 
-                max_tokens = st.slider('max_tokens', 0, m_token, 1000, 100, "%i", "max_tokens", "The maximum number of tokens to generate in the completion. The token count of your prompt plus max_tokens cannot exceed the model\'s context length.")
-                temperature = st.slider('temperature', 0.0, 1.0, 0.5, 0.01, "%0.2f", "temperature", "The temperature of the model. Higher temperature results in more surprising text.")
+                temperature = 0.5
+                if beta_model is False:
+                    temperature = st.slider('temperature', 0.0, 1.0, 0.5, 0.01, "%0.2f", "temperature", "The temperature of the model. Higher temperature results in more surprising text.")
 
-                if vision_mode is False:
+                if vision_mode is False and beta_model is False:
                     presets = st.selectbox("GPT Task", options=list(self.gpt_presets.keys()), index=0, key="presets", help=self.gpt_presets_help)
                 else:
                     presets = list(self.gpt_presets.keys())[0]
@@ -295,10 +309,13 @@ def set_ui(self):
         if 'gpt_last_prompt' not in st.session_state:
             st.session_state['gpt_last_prompt'] = ''
 
-        prompt_value=f"GPT ({model}) Input (role: {role}) [max_tokens: {max_tokens} | temperature: {temperature}"
+        prompt_value=f"GPT ({model}) Input (role: {role}) [max_tokens: {max_tokens} "
+        if beta_model is False:
+            prompt_value += f"| temperature: {temperature}"
         if vision_mode:
             prompt_value += f" | vision details: {vision_details}"
-        prompt_value += f" | preset: {presets}"
+        if beta_model is False:
+            prompt_value += f" | preset: {presets}"
         if prompt_preset is not None:
             prompt_value += f" | prompt preset: {prompt_preset}"
         if 'gpt_clear_chat' in st.session_state or clear_chat is True:
@@ -351,7 +368,9 @@ def set_ui(self):
                 st.warning("You requested an estimated %i tokens, which might exceed the model's context window of %i tokens. We are still proceeding with the request, but an error return is possible." % (requested_token_count, self.models[model]["context_token"]))
 
             if max_tokens > 0:
-                with st.spinner(f"Asking OpenAI ({model} for {max_tokens} tokens with temperature {temperature}. Prompt est. tokens : {prompt_token_count})"):
+                tmp_txt = "" if beta_model is True else f" (temperature: {temperature})"
+                st.toast(f"Requesting OpenAI ({model} for {max_tokens} tokens{tmp_txt})")
+                with st.spinner(f"Asking OpenAI ({model} for {max_tokens} tokens{tmp_txt}. Prompt est. tokens : {prompt_token_count})"):
                     if 'gpt_clear_chat' in st.session_state:
                         clear_chat = True
                         del st.session_state['gpt_clear_chat']

README.md

@@ -98,8 +98,8 @@ The following table shows the [models](https://platform.openai.com/docs/models/)
 | GPT | gpt-4o-2024-08-06 | active | vision | | 0.9.8 |
 | GPT | gpt-4o-mini | active | vision | | 0.9.7 |
 | GPT | gpt-4o-mini-2024-07-18 | active | vision | | 0.9.7 |
-| GPT | o1-preview | active | vision | untested | 0.9.8 |
-| GPT | o1-mini | active | vision | untested | 0.9.8 |
+| GPT | o1-preview | active | | untested | 0.9.8 |
+| GPT | o1-mini | active | | untested | 0.9.8 |
 
 (MM) Note on "untested": we do not yet have access to the model (`... or you do not have access to it`)
 

models.json

@@ -58,8 +58,9 @@
             "context_token": 128000,
             "data": "Up to Oct 2023 (as of 20240912)",
             "status": "active",
-            "status_details": "",
-            "capability": "vision"
+            "status_details": "beta model: limited control of features and capabilities. For details, see https://platform.openai.com/docs/guides/reasoning",
+            "capability": "",
+            "beta_model": true
         },
         "o1-mini":
         {
@@ -68,8 +69,9 @@
             "context_token": 128000,
             "data": "Up to Oct 2023 (as of 20240912)",
             "status": "active",
-            "status_details": "",
-            "capability": "vision"
+            "status_details": "beta model: limited control of features and capabilities. For details, see https://platform.openai.com/docs/guides/reasoning",
+            "capability": "",
+            "beta_model": true
         },
         "gpt-4-turbo":
         {

pyproject.toml

@@ -14,7 +14,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry.dependencies]
 python = "^3.10"
-openai = "==1.51.0"
+openai = "==1.56.1"
 streamlit = "^1.39.0"
 extra-streamlit-components = "^0.1.71"
 streamlit-extras = "^0.4.7"