Refine agent interaction and improve CLI output

This commit includes two main changes:

1. In `aicodebot/agents.py`, the prompt for the AI agent has been updated to emphasize the importance of properly formatting responses in JSON. This change is crucial to ensure that the agent's responses can be correctly parsed and processed.

2. In `aicodebot/cli.py`, the verbosity level passed to the `get_agent` function is now the one provided by the user, instead of always being `True`. Additionally, the agent's response is now displayed in a more user-friendly way, with a spinner indicating that the agent is "thinking", and the response being printed separately from the spinner. This improves the user experience by providing clearer feedback about what the agent is doing.

aicodebot/agents.py

@@ -9,8 +9,10 @@
 
 If asking the human for more clarification would produce better results, you can ask the human for more information.
 
-Before changing any local files, you should ALWAYS check with the human developer first, explaining what you are doing,
-you can give human multiple choices.
+Important: When you respond calling a tool, you should make sure your response is properly json formatted, ie escape
+quotes and newlines.
+
+Before writing any local files, you should ALWAYS check with the human developer first, explaining what you are doing.
 """
 
 

aicodebot/cli.py

@@ -243,10 +243,12 @@ def sidekick(task, verbose):
     model = get_llm_model()
     llm = ChatOpenAI(model=model, temperature=DEFAULT_TEMPERATURE, max_tokens=3500, verbose=verbose)
 
-    agent = get_agent("sidekick", llm, True)
+    agent = get_agent("sidekick", llm, verbose)
 
-    response = agent({"input": task})
-    console.print(response, style=bot_style)
+    with console.status("Thinking", spinner=DEFAULT_SPINNER):
+        response = agent({"input": task})
+        console.print("")
+        console.print(response["output"], style=bot_style)
 
 
 # ---------------------------------------------------------------------------- #