feat: Update CLI to use Haiku model with Y default and accurate cost info

summarize_yt/cli.py

@@ -18,12 +18,13 @@ from . import __version__
 CLAUDE_COST_PER_1K_INPUT = 0.015  # Cost per 1K tokens for input
 CLAUDE_COST_PER_1K_OUTPUT = 0.075  # Cost per 1K tokens for output
 ESTIMATED_TOKENS_PER_CHAR = 0.25  # Rough estimate of tokens per character
+HAIKU_OUTPUT_TOKENS = 50  # Haiku summaries are very short
 
 
 def estimate_api_cost(text: str) -> float:
     """Estimate the cost of sending text to Claude API."""
     estimated_input_tokens = len(text) * ESTIMATED_TOKENS_PER_CHAR
-    estimated_output_tokens = 1024  # max_tokens setting
+    estimated_output_tokens = HAIKU_OUTPUT_TOKENS
 
     input_cost = (estimated_input_tokens / 1000) * CLAUDE_COST_PER_1K_INPUT
     output_cost = (estimated_output_tokens / 1000) * CLAUDE_COST_PER_1K_OUTPUT
@@ -196,7 +197,7 @@ def get_summary_from_claude(text: str, duration_mins: int, prompt: str = None) -
         target_words = max(500, (duration_mins // 10) * 500)
         
         if prompt is None:
-            prompt = f"Please summarize this transcript in approximately {target_words} words"
+            prompt = "Please summarize this transcript as a haiku. A haiku is a three-line poem with 5 syllables in the first line, 7 syllables in the second line, and 5 syllables in the third line."
             
         message = client.messages.create(
             model="claude-3-sonnet-20240229",
@@ -247,8 +248,8 @@ def main():
 
     if not args.yes:
         try:
-            response = input("\nDo you want to proceed with getting the summary? (y/N): ").strip().lower()
-            if response != 'y':
+            response = input("\nDo you want to proceed with getting the summary? (Y/n): ").strip().lower()
+            if response == 'n':
                 print("Operation cancelled by user.")
                 cleanup_files(vtt_path)
                 sys.exit(0)