zev/yts
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refactor: Simplify CLI, remove optional features, and streamline summary generation

Browse Source
This commit is contained in:
Zev Averbach
2025-01-03 14:54:45 +01:00
committed by Zev Averbach (aider)
parent 1fbe9094c2
commit d3451c19e4
1 changed files with 60 additions and 71 deletions
Download Patch File Download Diff File Expand all files Collapse all files

131
summarize_yt/cli.py
View File

@@ -1,30 +1,47 @@
# Standard library imports remain the same
import argparse
import os
from pathlib import Path
import re
import signal
import sys
from datetime import datetime
from yt_dlp import YoutubeDL
from pathlib import Path
from typing import Optional, Tuple
# Third-party packages
import anthropic
from yt_dlp import YoutubeDL
# Local modules
from . import __version__
# Add cost estimation constants
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
CLAUDE_COST_PER_1K_INPUT = 0.80 / 1000
CLAUDE_COST_PER_1K_OUTPUT = 4 / 1000
ESTIMATED_TOKENS_PER_CHAR = 0.25
def estimate_api_cost(text: str) -> float:
def setup_terminal_control():
"""Set up terminal control at program start."""
try:
# Put process in its own process group and take control of terminal
os.setpgrp()
# Ignore terminal control signals
signal.signal(signal.SIGTTOU, signal.SIG_IGN)
signal.signal(signal.SIGTTIN, signal.SIG_IGN)
signal.signal(signal.SIGTSTP, signal.SIG_IGN)
# Take control of terminal if we're running in one
if sys.stdin.isatty():
import termios
termios.tcsetpgrp(sys.stdin.fileno(), os.getpgrp()) # type: ignore
except Exception:
# If we can't get terminal control, just continue
pass
def estimate_api_cost(text: str, target_words: float) -> float:
"""Estimate the cost of sending text to Claude API."""
estimated_input_tokens = len(text) * ESTIMATED_TOKENS_PER_CHAR
estimated_output_tokens = HAIKU_OUTPUT_TOKENS
estimated_output_tokens = target_words
input_cost = (estimated_input_tokens / 1000) * CLAUDE_COST_PER_1K_INPUT
output_cost = (estimated_output_tokens / 1000) * CLAUDE_COST_PER_1K_OUTPUT
@@ -47,7 +64,7 @@ def sanitize_filename(url: str) -> str:
return f"{timestamp}_{video_id if video_id else 'video'}"
def ensure_yts_dirs() -> Tuple[Path, Path]:
def ensure_yts_dirs() -> tuple[Path, Path]:
"""Create and return paths to transcript and summary directories."""
base_dir = Path.home() / ".yts"
transcript_dir = base_dir / "transcripts"
@@ -57,23 +74,23 @@ def ensure_yts_dirs() -> Tuple[Path, Path]:
return transcript_dir, summary_dir
def download_subtitles(url: str, quiet: bool = False) -> Tuple[bool, Optional[Path], int]:
def download_subtitles(url: str, quiet: bool = False) -> tuple[bool, Path | None, int]:
"""Download subtitles from YouTube using yt-dlp and return success, path, and duration in minutes."""
try:
ydl_opts = {
'skip_download': True,
'writeautomaticsub': True,
'subtitleslangs': ['en'],
'quiet': quiet,
'no_warnings': quiet,
"skip_download": True,
"writeautomaticsub": True,
"subtitleslangs": ["en"],
"quiet": quiet,
"no_warnings": quiet,
}
# First get video duration
info_opts = dict(ydl_opts)
info_opts['extract_flat'] = True
info_opts["extract_flat"] = True
with YoutubeDL(info_opts) as ydl:
info = ydl.extract_info(url, download=False)
duration_mins = int(info.get('duration', 0) / 60)
duration_mins = int(info.get("duration", 0) / 60) # type: ignore
if not quiet:
print(f"Debug: Downloading subtitles for {url}")
@@ -84,10 +101,10 @@ def download_subtitles(url: str, quiet: bool = False) -> Tuple[bool, Optional[Pa
# Find the downloaded VTT file
current_dir = Path(".")
vtt_files = list(current_dir.glob("*.en.vtt"))
if not quiet:
print(f"Debug: Found VTT files: {vtt_files}")
if not vtt_files:
print("No VTT file found after download", file=sys.stderr)
return False, None, 0
@@ -99,6 +116,7 @@ def download_subtitles(url: str, quiet: bool = False) -> Tuple[bool, Optional[Pa
if not quiet:
print(f"Debug: Full exception info: {type(e).__name__}: {str(e)}")
import traceback
traceback.print_exc()
return False, None, 0
@@ -151,7 +169,7 @@ def process_file(input_path: Path, output_path: Path | None = None, quiet: bool
return ""
def save_transcript(text: str, url: str, prompt: str) -> Path:
def save_transcript(text: str, url: str) -> Path:
"""Save transcript with metadata to ~/.yts/transcripts/."""
transcript_dir, _ = ensure_yts_dirs()
filename = sanitize_filename(url) + ".txt"
@@ -160,7 +178,6 @@ def save_transcript(text: str, url: str, prompt: str) -> Path:
metadata = f"""URL: {url}
Script Version: {__version__}
Timestamp: {datetime.now().isoformat()}
Claude Prompt: {prompt}
---
"""
@@ -170,7 +187,7 @@ Claude Prompt: {prompt}
return filepath
def cleanup_files(vtt_path: Optional[Path]):
def cleanup_files(vtt_path: Path | None):
"""Remove downloaded files after processing."""
try:
if vtt_path and vtt_path.exists():
@@ -185,7 +202,7 @@ def cleanup_files(vtt_path: Optional[Path]):
print(f"Warning: Could not clean up temporary files: {str(e)}", file=sys.stderr)
def get_summary_from_claude(text: str, duration_mins: int, prompt: str = None) -> str:
def get_summary_from_claude(text: str, target_words: float) -> str:
"""Send text to Claude API for summarization."""
try:
api_key = os.environ.get("ANTHROPIC_API_KEY")
@@ -193,20 +210,17 @@ def get_summary_from_claude(text: str, duration_mins: int, prompt: str = None) -
raise ValueError("ANTHROPIC_API_KEY environment variable not set")
client = anthropic.Anthropic()
# Calculate target word count based on duration
target_words = max(500, (duration_mins // 10) * 500)
if prompt is None:
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."
prompt = f"Please summarize this transcript in {target_words} or less."
message = client.messages.create(
model="claude-3-sonnet-20240229",
model="claude-3-5-haiku-latest",
max_tokens=2048, # Increased for longer summaries
temperature=0,
system="You are a helpful assistant that summarizes transcripts accurately and concisely.",
system="You are a helpful assistant that summarizes transcripts accurately.",
messages=[{"role": "user", "content": f"{prompt}:\n\n{text}"}],
)
return message.content[0].text
return message.content[0].text # type: ignore
except Exception as e:
print(f"Error getting summary from Claude: {str(e)}", file=sys.stderr)
@@ -214,16 +228,12 @@ def get_summary_from_claude(text: str, duration_mins: int, prompt: str = None) -
def main():
setup_terminal_control()
parser = argparse.ArgumentParser(description="Download YouTube subtitles and get Claude summary")
parser.add_argument("url", help="YouTube video URL", type=str)
parser.add_argument("-o", "--output", help="Output file for summary")
parser.add_argument("-q", "--quiet", action="store_true", help="Suppress status messages")
parser.add_argument("--keep-files", action="store_true", help="Don't delete downloaded VTT files")
parser.add_argument("--transcript", action="store_true", help="Include full transcript in output")
parser.add_argument(
"--prompt", help="Custom prompt for Claude (default: auto-calculated based on video length)"
)
parser.add_argument("-y", "--yes", action="store_true", help="Skip cost confirmation")
args = parser.parse_args()
@@ -234,36 +244,24 @@ def main():
if not success:
cleanup_files(None)
sys.exit(1)
target_words = max(500, (duration_mins // 10) * 500)
# Process the VTT file
cleaned_text = process_file(vtt_path, None, args.quiet)
cleaned_text = process_file(vtt_path, None, args.quiet) # type: ignore
# Save transcript
transcript_path = save_transcript(cleaned_text, args.url, args.prompt)
transcript_path = save_transcript(cleaned_text, args.url)
print(f"\nTranscript saved to: {transcript_path}")
# Estimate and display cost
estimated_cost = estimate_api_cost(cleaned_text)
estimated_cost = estimate_api_cost(cleaned_text, target_words)
print(f"\nEstimated API cost: ${estimated_cost:.4f}")
if not args.yes:
try:
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)
except (EOFError, KeyboardInterrupt):
print("\nOperation cancelled by user.")
cleanup_files(vtt_path)
sys.exit(0)
# Get summary from Claude
if not args.quiet:
print("\nGetting summary from Claude...")
summary = get_summary_from_claude(cleaned_text, duration_mins, args.prompt)
summary = get_summary_from_claude(cleaned_text, target_words)
# Prepare output
output = ""
if args.transcript:
output += "=== Full Transcript ===\n\n"
@@ -275,23 +273,14 @@ def main():
_, summary_dir = ensure_yts_dirs()
summary_filename = sanitize_filename(args.url) + "_summary.txt"
summary_path = summary_dir / summary_filename
with open(summary_path, "w", encoding="utf-8") as f:
f.write(output)
print(f"\nSummary saved to: {summary_path}")
# Print the summary to console
print("\n=== Summary ===\n")
print(summary)
print() # Extra newline for readability
print()
print(f"\nSummary saved to: {summary_path}")
# If output path specified, also save there
if args.output:
with open(args.output, "w", encoding="utf-8") as f:
f.write(output)
print(f"Summary also saved to: {args.output}")
# Cleanup downloaded files unless --keep-files is specified
if not args.keep_files:
cleanup_files(vtt_path)