Fix/chatsequence

langclient/__main__.py

@@ -94,14 +94,14 @@ def interactve_chat(
         or _prompt_for_api_key()
     )
 
-    input_name = user_name()
+    user = user_name()
     model_selected = select_language_model()
 
-    stream_chat_ = use_key(api_key)(
-        partial(stream_chat, model=model_selected, max_tokens=3500)
-    )
+    stream_chat_ = use_key(api_key)(partial(stream_chat, model=model_selected))
 
-    for _ in chat_sequence_process(chat_input(input_name), stream_chat_):
+    for _ in chat_sequence_process(
+        chat_input(user), stream_chat_, model_selected, user_name=user
+    ):
         pass
 
 

langclient/chat_functions.py

@@ -35,7 +35,7 @@ def _sum_accumulated_tokens(tokens):
     return input_cost + output_cost
 
 
-def _token_usage_stats(messages: list[Message], model: LanguageModel) -> str:
+def token_usage_stats(messages: list[Message], model: LanguageModel) -> str:
     input_content = _filter_content_by_role(messages, "user")
     output_content = _filter_content_by_role(messages, "assistant")
 
@@ -54,8 +54,11 @@ def _token_usage_stats(messages: list[Message], model: LanguageModel) -> str:
     usage_stats = f"{_format_token_number(tokens_messages)}/"
     usage_stats += f"{_format_token_number(model.max_token)}"
 
-    stats_string = f" {Fore.MAGENTA}({usage_stats}){Fore.RESET}"
-    stats_string += f" {Fore.GREEN}${cost_stats}{Fore.RESET}"
+    color_token = Fore.LIGHTBLACK_EX
+    color_cost = Fore.LIGHTGREEN_EX
+
+    stats_string = f" {color_token}({usage_stats}){Fore.RESET}"
+    stats_string += f" {color_cost}${cost_stats}{Fore.RESET}"
     return stats_string
 
 
@@ -65,7 +68,6 @@ def stream_chat(
     api_key: str,
     model: LanguageModel,
     temperature=1,
-    max_tokens=14505,
     top_p=1,
     frequency_penalty=0,
     presence_penalty=0,
@@ -75,8 +77,7 @@ def stream_chat(
     generator: Iterable[ChatCompletionChunk] = client.chat.completions.create(
         model=model.name,
         messages=list_of_dict_messages,
-        temperature=temperature,
-        max_tokens=max_tokens,
+        temperature=temperature,  # next parameters have no effect
         top_p=top_p,
         frequency_penalty=frequency_penalty,
         presence_penalty=presence_penalty,
@@ -85,9 +86,13 @@ def stream_chat(
 
     deltas_content = map(lambda chunk: chunk.choices[0].delta.content, generator)
 
-    assistant_head = f"\n{Fore.CYAN}Assistant:{Fore.RESET}"
-    assistant_head += _token_usage_stats(messages, model)
+    color_assistant = Fore.BLUE
+    assistant_name = "Assistant"
+
+    assistant_head = f"{color_assistant}{assistant_name}:{Fore.RESET}"
+    assistant_head += token_usage_stats(messages, model)
 
+    print()
     print(assistant_head)
 
     def _typeguard(content: str | None) -> TypeGuard[str]:

langclient/data/models.json

@@ -6,13 +6,7 @@
         "cost_1kT_output": 0.0015
     },
     {
-        "name": "gpt-4-0125-preview",
-        "max_token": 128000,
-        "cost_1kT_input": 0.01,
-        "cost_1kT_output": 0.03
-    },
-    {
-        "name": "gpt-4-1106-preview",
+        "name": "gpt-4-turbo-2024-04-09",
         "max_token": 128000,
         "cost_1kT_input": 0.01,
         "cost_1kT_output": 0.03

langclient/interactive_chat_handling.py

@@ -4,7 +4,8 @@
 from itertools import accumulate
 from functools import partial
 
-from langclient.models import Message, Role
+from langclient.models import Message, Role, LanguageModel
+from langclient.chat_functions import token_usage_stats
 
 
 def _get_files_from_message(message: str) -> list[str]:
@@ -53,7 +54,6 @@ def chat_input(user_name: str) -> Iterable[Message]:
     """Chat input stream."""
 
     while True:
-        print(Fore.CYAN + f"{user_name}:" + Fore.RESET)
         input_message = input()
         yield _enhance_user_input(input_message)
 
@@ -79,7 +79,11 @@ def _print_intercept(iter: Iterable):
 
 
 def _step_process(
-    previous_messages: list[Message], user_message: Message, chat_function: Callable
+    previous_messages: list[Message],
+    user_message: Message,
+    chat_function: Callable,
+    model: LanguageModel,
+    user_name: str,
 ) -> list[Message]:
     """Get the next step in the conversation."""
     assistant_response_chunks = chat_function([*previous_messages, user_message])
@@ -88,13 +92,32 @@ def _step_process(
         role=Role.ASSISTANT,
         content="".join(_print_intercept(assistant_response_chunks)),
     )
+
+    chat_updated = [*previous_messages, user_message, assistant_response]
+
+    user_name_color = Fore.CYAN
+    user_head = f"{user_name_color}{user_name}:{Fore.RESET}"
+    user_head += token_usage_stats(chat_updated, model)
+
     print()
     print()
-    return [*previous_messages, user_message, assistant_response]
+    print(user_head)
+
+    return chat_updated
 
 
 def chat_sequence_process(
-    user_input: Iterable[Message], chat_function: Callable
+    user_input: Iterable[Message],
+    chat_function: Callable,
+    model: LanguageModel,
+    user_name: str,
 ) -> Iterable[Message]:
-    accumulate_function = partial(_step_process, chat_function=chat_function)
+    accumulate_function = partial(
+        _step_process, chat_function=chat_function, user_name=user_name, model=model
+    )
+
+    user_name_color = Fore.CYAN
+    first_user_head = f"{user_name_color}{user_name}:{Fore.RESET}"
+    print(first_user_head)
+
     return accumulate(user_input, accumulate_function, initial=[])

setup.py

@@ -9,7 +9,7 @@
 
 setup(
     name="langclient",
-    version="0.4.0",
+    version="0.4.1",
     description="Simple OpenAI language repl client",
     author="Juan Molina Riddell",
     author_email="[email protected]",