Podcast Agent

export GOOGLE_APPLICATION_CREDENTIALS="/path/to/service-account-key.json"

from podcast_agent.geminivideo import process_video

# Process a single video
output_path = process_video("path/to/podcast.mp4")
print(f"Transcript saved to: {output_path}")

[
  {
    "speaker": "Andy",
    "content": "Welcome to The Rollup! Today we're discussing..."
  },
  {
    "speaker": "Rob",
    "content": "Thanks for having me. I'm excited to talk about..."
  },
  {
    "speaker": "Guest",
    "content": "It's great to be here. Let me share some insights on..."
  }
]

from podcast_agent.geminivideo import main
import os

# Process all videos in a directory
video_dir = "split_videos"
video_files = [
    os.path.join(video_dir, f)
    for f in os.listdir(video_dir)
    if f.lower().endswith(('.mov', '.mp4', '.avi', '.mkv', '.webm'))
]

for video_path in video_files:
    try:
        process_video(video_path)
    except Exception as e:
        print(f"Failed to process {video_path}: {str(e)}")

from podcast_agent.aiagenteditor import process_video

# Process video with custom instructions
output_path = process_video(
    videopath="podcast_episode.mp4",
    custom_instructions="Focus on technical discussions, remove pauses longer than 2 seconds"
)

if output_path:
    print(f"Edited video saved to: {output_path}")

{
  "edits": [
    {
      "start_time": "00:00:00",
      "end_time": "00:05:30",
      "keep": true,
      "reason": "Strong introduction with key points"
    },
    {
      "start_time": "00:05:30",
      "end_time": "00:06:15",
      "keep": false,
      "reason": "Long pause and technical difficulties"
    },
    {
      "start_time": "00:06:15",
      "end_time": "00:15:00",
      "keep": true,
      "reason": "Main discussion with valuable insights"
    }
  ]
}

# Example 1: Technical content focus
process_video(
    "tech_podcast.mp4",
    custom_instructions="Keep all technical discussions, remove casual banter"
)

# Example 2: Pacing improvement
process_video(
    "interview.mp4",
    custom_instructions="Remove pauses longer than 3 seconds, keep all Q&A segments"
)

# Example 3: Highlight reel
process_video(
    "long_episode.mp4",
    custom_instructions="Extract only the most insightful moments and key takeaways"
)

pip install chromadb sentence-transformers

from podcast_agent.podcast_knowledge_base import PodcastKnowledgeBase

# Initialize with persistent storage
kb = PodcastKnowledgeBase(collection_name="podcast_knowledge")

# Process all JSON transcripts
kb.process_all_json_files(directory="jsonoutputs/")

from podcast_agent.podcast_knowledge_base import PodcastSegment

# Create segments manually
segments = [
    PodcastSegment(
        id="ep1_0",
        speaker="Andy",
        content="Today we're discussing the future of AI...",
        source_file="episode_1.json"
    ),
    PodcastSegment(
        id="ep1_1",
        speaker="Rob",
        content="Machine learning has evolved significantly...",
        source_file="episode_1.json"
    )
]

kb.add_segments(segments)

# Query the knowledge base
results = kb.query_knowledge_base(
    query="What did they say about machine learning?",
    n_results=5
)

# Format and display results
formatted = kb.format_query_results(results)
print(formatted)

[
    {
        "content": "Machine learning has evolved significantly...",
        "metadata": {
            "speaker": "Rob",
            "source_file": "episode_1.json",
            "timestamp": "2024-01-15T10:30:00"
        },
        "relevance_score": 0.89
    },
    {
        "content": "The applications of ML in healthcare are tremendous...",
        "metadata": {
            "speaker": "Guest",
            "source_file": "episode_3.json",
            "timestamp": "2024-01-20T14:15:00"
        },
        "relevance_score": 0.76
    }
]

# Get collection stats
stats = kb.get_collection_stats()
print(f"Total segments: {stats['count']}")
print(f"Last updated: {stats['last_update']}")

# Check processed files
processed = kb.get_processed_files()
print(f"Processed files: {processed}")

# Complex semantic search
queries = [
    "What are the challenges in blockchain scalability?",
    "How does NFT technology work?",
    "What did guests say about DeFi adoption?"
]

for query in queries:
    print(f"\n=== Query: {query} ===")
    results = kb.query_knowledge_base(query, n_results=3)
    
    for i, result in enumerate(results, 1):
        print(f"\n{i}. [{result['metadata']['speaker']}] "
              f"(Score: {result['relevance_score']:.2f})")
        print(f"   {result['content'][:200]}...")

# Clear the entire collection
kb.clear_collection()

# Re-process all files
kb.process_all_json_files()

# Get processed file list
processed_files = kb.get_processed_files()
print(f"Knowledge base contains {len(processed_files)} transcripts")

import asyncio
from podcast_agent.geminivideo import process_video as transcribe
from podcast_agent.aiagenteditor import process_video as edit
from podcast_agent.podcast_knowledge_base import PodcastKnowledgeBase

async def process_podcast_workflow(video_path: str):
    # Step 1: Transcribe the video
    print("Step 1: Transcribing video...")
    transcript_path = transcribe(video_path)
    
    # Step 2: Edit the video (optional)
    print("Step 2: Editing video...")
    edited_path = edit(
        video_path,
        custom_instructions="Keep main discussion, remove technical issues"
    )
    
    # Step 3: Add to knowledge base
    print("Step 3: Adding to knowledge base...")
    kb = PodcastKnowledgeBase()
    kb.process_json_file(transcript_path)
    
    # Step 4: Query the content
    print("Step 4: Testing semantic search...")
    results = kb.query_knowledge_base("main topics discussed", n_results=5)
    print(kb.format_query_results(results))
    
    return {
        "transcript": transcript_path,
        "edited_video": edited_path,
        "segments_added": len(results)
    }

# Run the workflow
result = asyncio.run(process_podcast_workflow("my_podcast.mp4"))
print(f"Workflow complete: {result}")