fix: decrease max_tokens to 1024

The previous value of 8192 is too large and causes the following error: "max_tokens is too large: 8192. This model supports at most 4096 completion tokens, whereas you provided 8192."

max_tokens indicates the number of tokens in the *output* and input + max_tokens < context length for model. 1024 should be a safe value to use and sufficiently large for creating even the most advanced CLI commands.

Resolves: #28

src/fish_ai/engine.py

@@ -218,7 +218,7 @@ def get_response(messages):
     else:
         completions = get_openai_client().chat.completions.create(
             model=get_config('model') or 'gpt-4',
-            max_tokens=8192,
+            max_tokens=1024,
             messages=messages,
             stream=False,
             temperature=float(get_config('temperature') or '0.2'),