Configuration Guide

Basic Usage

Here’s how to configure the Podcast LLM system by providing a custom configuration file:

YAML configuration

The system can be configured using the config.yaml file:

The configuration file contains several sections:

LLM Configuration

fast_llm_provider: Provider for quick LLM operations (options: ‘openai’, ‘google’, ‘anthropic’)
long_context_llm_provider: Provider for operations requiring longer context

Text-to-Speech Configuration

tts_provider: Text-to-speech service to use (options: ‘google’, ‘elevenlabs’)
tts_settings: Provider-specific settings including:
- Voice mappings for interviewer and interviewee
- Model settings
- Language codes
- Audio effect profiles

Audio Settings

output_format: Format for generated audio (options: ‘mp3’, ‘wav’)
temp_audio_dir: Directory for temporary audio files
output_dir: Directory for final output files

Checkpointing Settings

checkpoint_dir: Directory for saving generation checkpoints

Rate Limiting

Configure API rate limits per provider:

requests_per_minute: Maximum requests allowed per minute
max_retries: Number of retry attempts
base_delay: Base delay between retries

Content Settings

podcast_name: Name of the podcast
intro: Template for podcast introduction (variables: {podcast_name}, {topic})
outro: Template for podcast conclusion
episode_structure: List defining the structure of episodes

Example Configuration

# LLM Configuration
fast_llm_provider: anthropic
long_context_llm_provider: google

# TTS Configuration
tts_provider: elevenlabs
tts_settings:
  elevenlabs:
    voice_mapping:
      Interviewer: Chris
      Interviewee: Charlie
    model: eleven_multilingual_v2

# Output settings
output_format: mp3
output_dir: ./output

# Content settings
podcast_name: My AI Podcast
intro: "Welcome to {podcast_name}. Today we're exploring {topic}."