podcaster / api_clients.py
marks
Merge branch 'main' of https://huggingface.co/spaces/phxdev/podcaster
a7eb60c
raw
history blame
9.06 kB
from functools import lru_cache
from typing import List, Tuple, Optional, Dict
import aiohttp
import elevenlabs
import time
from contextlib import asynccontextmanager
from logger import setup_logger, log_execution_time, log_async_execution_time
from models import OpenRouterModel
logger = setup_logger("api_clients")
def preprocess_text(text: str) -> str:
"""
Clean and format text by removing unwanted characters and formatting
Args:
text: Raw input text
Returns:
Cleaned text suitable for podcast generation
"""
import re
# Remove markdown-style formatting
text = re.sub(r'\*\*(.+?)\*\*', r'\1', text) # Bold
text = re.sub(r'\*(.+?)\*', r'\1', text) # Italic
text = re.sub(r'__(.+?)__', r'\1', text) # Underline
text = re.sub(r'~~(.+?)~~', r'\1', text) # Strikethrough
# Remove action blocks and special formatting
text = re.sub(r'\[.*?\]', '', text) # Remove [actions]
text = re.sub(r'\{.*?\}', '', text) # Remove {actions}
text = re.sub(r'<.*?>', '', text) # Remove <actions>
# Remove podcast-specific formatting
text = re.sub(r'\((?:pause|break|music|sfx|sound effect|jingle).*?\)', '', text, flags=re.IGNORECASE)
text = re.sub(r'\((host|speaker|guest)\s*\d*\s*:?\)', '', text, flags=re.IGNORECASE)
text = re.sub(r'#\s*\d+\s*[:.-]', '', text) # Remove segment numbers
# Clean up whitespace
return ' '.join(text.split())
class OpenRouterClient:
"""Handles OpenRouter API interactions with comprehensive logging and error tracking"""
def __init__(self, api_key: str):
logger.info("Initializing OpenRouter client")
self.api_key = api_key
self.base_url = "https://openrouter.ai/api/v1"
self.headers = {
"Authorization": f"Bearer {api_key}",
"HTTP-Referer": "https://localhost:7860", # Required by OpenRouter
"X-Title": "URL to Podcast Generator", # Required by OpenRouter
"Content-Type": "application/json"
}
logger.debug("OpenRouter client initialized successfully")
@property
def api_key(self):
return self._api_key
@api_key.setter
def api_key(self, value: str):
if not value or len(value) < 32:
logger.error("Invalid API key format")
raise ValueError("Invalid OpenRouter API key")
self._api_key = value
# Update headers when API key changes
self.headers = {
"Authorization": f"Bearer {value}",
"HTTP-Referer": "https://localhost:7860",
"X-Title": "URL to Podcast Generator",
"Content-Type": "application/json",
}
logger.info("OpenRouter API key updated successfully")
@asynccontextmanager
async def get_session(self):
logger.debug("Creating new aiohttp session")
async with aiohttp.ClientSession(headers=self.headers) as session:
yield session
@lru_cache(maxsize=1)
async def get_models(self) -> List[Tuple[str, str]]:
"""
Fetch available models from OpenRouter API using pydantic models
Returns:
List of tuples containing (model_id, model_id) where both values are the same
"""
logger.info("Fetching available models from OpenRouter")
async with self.get_session() as session:
async with session.get(f"{self.base_url}/models") as response:
response.raise_for_status()
data = await response.json()
models = [OpenRouterModel(**model) for model in data["data"]]
logger.info(f"Successfully fetched {len(models)} models")
return [(model.name, model.id) for model in models]
@log_async_execution_time(logger)
async def generate_script(self, content: str, prompt: str, model_id: str) -> str:
"""
Generate a podcast script with detailed progress tracking and validation
Performance metrics and content analysis are logged at each step.
"""
logger.info(f"Starting script generation with model: {model_id}")
logger.debug(f"Input metrics - Content: {len(content)} chars, Prompt: {len(prompt)} chars")
# Validate inputs
if not content or len(content) < 100:
logger.error("Content too short for meaningful script generation")
raise ValueError("Insufficient content for script generation")
if not prompt or len(prompt) < 10:
logger.error("Prompt too short or missing")
raise ValueError("Please provide a more detailed prompt")
# Clean input text
cleaned_content = preprocess_text(content)
cleaned_prompt = preprocess_text(prompt)
system_prompt = """You are an expert podcast script writer. Your task is to create engaging,
natural-sounding podcast scripts that flow conversationally while being informative and engaging.
Follow these guidelines:
1. Write in a conversational, natural speaking style that sounds authentic
2. Break complex topics into digestible segments with clear transitions
3. Avoid technical jargon unless necessary, explaining complex terms when used
4. Use natural speech patterns:
- Contractions (I'm, we're, let's)
- Casual language
- Rhetorical questions to engage listeners
5. Include brief pauses for emphasis and pacing (but don't mark them explicitly)
6. Incorporate storytelling elements to maintain engagement
7. End with a clear conclusion and call-to-action
8. Keep paragraphs short and focused for easier delivery
9. Use simple sentence structures that flow naturally when spoken
Format the script for natural speech, avoiding any special characters or formatting."""
user_prompt = f"""Create a podcast script based on the following topic and content:
Topic: {cleaned_prompt}
Content to cover: {cleaned_content}
Focus on making it engaging and natural to listen to."""
try:
request = OpenRouterRequest(
model=model_id,
messages=[
Message(role="system", content=system_prompt),
Message(role="user", content=user_prompt)
]
)
async with self.get_session() as session:
async with session.post(
f"{self.base_url}/chat/completions",
json=request_data
) as response:
if response.status != 200:
error_text = await response.text()
logger.error(f"OpenRouter API error: {error_text}")
raise ValueError(f"API request failed: {error_text}")
data = await response.json()
return data['choices'][0]['message']['content']
except Exception as e:
logger.error(f"Script generation failed", exc_info=True)
raise
class ElevenLabsClient:
def __init__(self, api_key: str):
self.api_key = api_key
elevenlabs.set_api_key(api_key)
def get_voices(self) -> List[Tuple[str, str]]:
"""
Synchronously get available voices from ElevenLabs
Returns:
List of tuples containing (voice_id, display_name)
where display_name shows the name and description but not the ID
"""
try:
voices = elevenlabs.voices()
return [(
f"{voice.name} ({voice.labels.get('accent', 'No accent')})" +
(f" - {voice.description[:50]}..." if voice.description else ""),
voice.voice_id # Value (hidden from user)
) for voice in voices]
except Exception as e:
logger.error("Failed to fetch voices from ElevenLabs", exc_info=True)
raise
def generate_audio(self, text: str, voice_id: str):
"""Generate audio synchronously"""
logger.info(f"Starting audio generation with voice: {voice_id}")
logger.debug(f"Input text length: {len(text)} chars")
if len(text) > 5000:
logger.warning(f"Long text detected ({len(text)} chars), may impact performance")
try:
start_time = time.time()
audio = elevenlabs.generate(
text=text,
voice=voice_id,
model="eleven_monolingual_v1"
)
duration = time.time() - start_time
audio_size = len(audio)
logger.info(f"Audio generated: {audio_size} bytes in {duration:.2f} seconds")
logger.debug(f"Audio generation rate: {len(text)/duration:.2f} chars/second")
return audio
except Exception as e:
logger.error("Audio generation failed", exc_info=True)
raise