422 lines
18 KiB
Python
422 lines
18 KiB
Python
"""
|
|
Research Persona Service
|
|
|
|
Handles generation, caching, and retrieval of AI-powered research personas.
|
|
"""
|
|
|
|
from typing import Dict, Any, Optional
|
|
from datetime import datetime, timedelta
|
|
from loguru import logger
|
|
from fastapi import HTTPException
|
|
|
|
from services.database import get_db_session
|
|
from models.onboarding import PersonaData, OnboardingSession
|
|
from models.research_persona_models import ResearchPersona
|
|
from .research_persona_prompt_builder import ResearchPersonaPromptBuilder
|
|
from services.llm_providers.main_text_generation import llm_text_gen
|
|
from services.onboarding.database_service import OnboardingDatabaseService
|
|
from services.persona_data_service import PersonaDataService
|
|
|
|
|
|
class ResearchPersonaService:
|
|
"""Service for generating and managing research personas."""
|
|
|
|
CACHE_TTL_DAYS = 7 # 7-day cache TTL
|
|
|
|
def __init__(self, db_session=None):
|
|
self.db = db_session or get_db_session()
|
|
self.prompt_builder = ResearchPersonaPromptBuilder()
|
|
self.onboarding_service = OnboardingDatabaseService(db=self.db)
|
|
self.persona_data_service = PersonaDataService(db_session=self.db)
|
|
|
|
def get_cached_only(
|
|
self,
|
|
user_id: str
|
|
) -> Optional[ResearchPersona]:
|
|
"""
|
|
Get research persona for user ONLY if it exists in cache.
|
|
This method NEVER generates - it only returns cached personas.
|
|
Use this for config endpoints to avoid triggering rate limit checks.
|
|
|
|
Args:
|
|
user_id: User ID (Clerk string)
|
|
|
|
Returns:
|
|
ResearchPersona if cached and valid, None otherwise
|
|
"""
|
|
try:
|
|
# Get persona data record
|
|
persona_data = self._get_persona_data_record(user_id)
|
|
|
|
if not persona_data:
|
|
logger.debug(f"No persona data found for user {user_id}")
|
|
return None
|
|
|
|
# Only return if cache is valid and persona exists
|
|
if self.is_cache_valid(persona_data) and persona_data.research_persona:
|
|
try:
|
|
logger.debug(f"Returning cached research persona for user {user_id}")
|
|
return ResearchPersona(**persona_data.research_persona)
|
|
except Exception as e:
|
|
logger.warning(f"Failed to parse cached research persona: {e}")
|
|
return None
|
|
|
|
# Cache invalid or persona missing - return None (don't generate)
|
|
logger.debug(f"No valid cached research persona for user {user_id}")
|
|
return None
|
|
|
|
except Exception as e:
|
|
logger.error(f"Error getting cached research persona for user {user_id}: {e}")
|
|
return None
|
|
|
|
def get_or_generate(
|
|
self,
|
|
user_id: str,
|
|
force_refresh: bool = False
|
|
) -> Optional[ResearchPersona]:
|
|
"""
|
|
Get research persona for user, generating if missing or expired.
|
|
|
|
Args:
|
|
user_id: User ID (Clerk string)
|
|
force_refresh: If True, regenerate even if cache is valid
|
|
|
|
Returns:
|
|
ResearchPersona if successful, None otherwise
|
|
"""
|
|
try:
|
|
# Get persona data record
|
|
persona_data = self._get_persona_data_record(user_id)
|
|
|
|
if not persona_data:
|
|
logger.warning(f"No persona data found for user {user_id}, cannot generate research persona")
|
|
return None
|
|
|
|
# Check cache if not forcing refresh
|
|
if not force_refresh and self.is_cache_valid(persona_data):
|
|
if persona_data.research_persona:
|
|
logger.info(f"Using cached research persona for user {user_id}")
|
|
try:
|
|
return ResearchPersona(**persona_data.research_persona)
|
|
except Exception as e:
|
|
logger.warning(f"Failed to parse cached research persona: {e}, regenerating...")
|
|
# Fall through to regeneration
|
|
else:
|
|
logger.info(f"Research persona missing for user {user_id}, generating...")
|
|
else:
|
|
if force_refresh:
|
|
logger.info(f"Forcing refresh of research persona for user {user_id}")
|
|
else:
|
|
logger.info(f"Cache expired for user {user_id}, regenerating...")
|
|
|
|
# Generate new research persona
|
|
try:
|
|
research_persona = self.generate_research_persona(user_id)
|
|
except HTTPException:
|
|
# Re-raise HTTPExceptions (e.g., 429 subscription limit) so they propagate to API
|
|
raise
|
|
|
|
if research_persona:
|
|
# Save to database
|
|
if self.save_research_persona(user_id, research_persona):
|
|
logger.info(f"✅ Research persona generated and saved for user {user_id}")
|
|
else:
|
|
logger.warning(f"Failed to save research persona for user {user_id}")
|
|
|
|
return research_persona
|
|
else:
|
|
# Log detailed error for debugging expensive failures
|
|
logger.error(
|
|
f"❌ Failed to generate research persona for user {user_id} - "
|
|
f"This is an expensive failure (API call consumed). Check logs above for details."
|
|
)
|
|
# Don't return None silently - let the caller know this failed
|
|
return None
|
|
|
|
except HTTPException:
|
|
# Re-raise HTTPExceptions (e.g., 429 subscription limit) so they propagate to API
|
|
raise
|
|
except Exception as e:
|
|
logger.error(f"Error getting/generating research persona for user {user_id}: {e}")
|
|
return None
|
|
|
|
def generate_research_persona(self, user_id: str) -> Optional[ResearchPersona]:
|
|
"""
|
|
Generate a new research persona for the user.
|
|
|
|
Args:
|
|
user_id: User ID (Clerk string)
|
|
|
|
Returns:
|
|
ResearchPersona if successful, None otherwise
|
|
"""
|
|
try:
|
|
logger.info(f"Generating research persona for user {user_id}")
|
|
|
|
# Collect onboarding data
|
|
onboarding_data = self._collect_onboarding_data(user_id)
|
|
|
|
if not onboarding_data:
|
|
logger.warning(f"Insufficient onboarding data for user {user_id}")
|
|
return None
|
|
|
|
# Build prompt
|
|
prompt = self.prompt_builder.build_research_persona_prompt(onboarding_data)
|
|
|
|
# Get JSON schema for structured response
|
|
json_schema = self.prompt_builder.get_json_schema()
|
|
|
|
# Call LLM with structured JSON response
|
|
logger.info(f"Calling LLM for research persona generation (user: {user_id})")
|
|
try:
|
|
response_text = llm_text_gen(
|
|
prompt=prompt,
|
|
json_struct=json_schema,
|
|
user_id=user_id
|
|
)
|
|
except HTTPException:
|
|
# Re-raise HTTPExceptions (e.g., 429 subscription limit) so they propagate to API
|
|
logger.warning(f"HTTPException during LLM call for user {user_id} - re-raising")
|
|
raise
|
|
except RuntimeError as e:
|
|
# Re-raise RuntimeError (subscription limits) as HTTPException
|
|
logger.warning(f"RuntimeError during LLM call for user {user_id}: {e}")
|
|
raise HTTPException(status_code=429, detail=str(e))
|
|
|
|
if not response_text:
|
|
logger.error("Empty response from LLM")
|
|
return None
|
|
|
|
# Parse JSON response
|
|
import json
|
|
try:
|
|
# When json_struct is provided, llm_text_gen may return a dict directly
|
|
if isinstance(response_text, dict):
|
|
# Already parsed, use directly
|
|
persona_dict = response_text
|
|
elif isinstance(response_text, str):
|
|
# Handle case where LLM returns markdown-wrapped JSON or plain JSON string
|
|
response_text = response_text.strip()
|
|
if response_text.startswith("```json"):
|
|
response_text = response_text[7:]
|
|
if response_text.startswith("```"):
|
|
response_text = response_text[3:]
|
|
if response_text.endswith("```"):
|
|
response_text = response_text[:-3]
|
|
response_text = response_text.strip()
|
|
|
|
persona_dict = json.loads(response_text)
|
|
else:
|
|
logger.error(f"Unexpected response type from LLM: {type(response_text)}")
|
|
return None
|
|
|
|
# Add generated_at timestamp
|
|
persona_dict["generated_at"] = datetime.utcnow().isoformat()
|
|
|
|
# Validate and create ResearchPersona
|
|
# Log the dict structure for debugging if validation fails
|
|
try:
|
|
research_persona = ResearchPersona(**persona_dict)
|
|
logger.info(f"✅ Research persona generated successfully for user {user_id}")
|
|
return research_persona
|
|
except Exception as validation_error:
|
|
logger.error(f"Failed to validate ResearchPersona from dict: {validation_error}")
|
|
logger.debug(f"Persona dict keys: {list(persona_dict.keys()) if isinstance(persona_dict, dict) else 'Not a dict'}")
|
|
logger.debug(f"Persona dict sample: {str(persona_dict)[:500]}")
|
|
# Re-raise to be caught by outer exception handler
|
|
raise
|
|
|
|
except json.JSONDecodeError as e:
|
|
logger.error(f"Failed to parse LLM response as JSON: {e}")
|
|
logger.debug(f"Response text: {response_text[:500] if isinstance(response_text, str) else str(response_text)[:500]}")
|
|
return None
|
|
except Exception as e:
|
|
logger.error(f"Failed to create ResearchPersona from response: {e}")
|
|
return None
|
|
|
|
except HTTPException:
|
|
# Re-raise HTTPExceptions (e.g., 429 subscription limit) so they propagate to API
|
|
raise
|
|
except Exception as e:
|
|
logger.error(f"Error generating research persona for user {user_id}: {e}")
|
|
return None
|
|
|
|
def is_cache_valid(self, persona_data: PersonaData) -> bool:
|
|
"""
|
|
Check if cached research persona is still valid (within TTL).
|
|
|
|
Args:
|
|
persona_data: PersonaData database record
|
|
|
|
Returns:
|
|
True if cache is valid, False otherwise
|
|
"""
|
|
if not persona_data.research_persona_generated_at:
|
|
return False
|
|
|
|
# Check if within TTL
|
|
cache_age = datetime.utcnow() - persona_data.research_persona_generated_at
|
|
is_valid = cache_age < timedelta(days=self.CACHE_TTL_DAYS)
|
|
|
|
if not is_valid:
|
|
logger.debug(f"Cache expired (age: {cache_age.days} days, TTL: {self.CACHE_TTL_DAYS} days)")
|
|
|
|
return is_valid
|
|
|
|
def save_research_persona(
|
|
self,
|
|
user_id: str,
|
|
research_persona: ResearchPersona
|
|
) -> bool:
|
|
"""
|
|
Save research persona to database.
|
|
|
|
Args:
|
|
user_id: User ID (Clerk string)
|
|
research_persona: ResearchPersona to save
|
|
|
|
Returns:
|
|
True if successful, False otherwise
|
|
"""
|
|
try:
|
|
persona_data = self._get_persona_data_record(user_id)
|
|
|
|
if not persona_data:
|
|
logger.error(f"No persona data record found for user {user_id}")
|
|
return False
|
|
|
|
# Convert ResearchPersona to dict for JSON storage
|
|
persona_dict = research_persona.dict()
|
|
|
|
# Update database record
|
|
persona_data.research_persona = persona_dict
|
|
persona_data.research_persona_generated_at = datetime.utcnow()
|
|
|
|
self.db.commit()
|
|
|
|
logger.info(f"✅ Research persona saved for user {user_id}")
|
|
return True
|
|
|
|
except Exception as e:
|
|
logger.error(f"Error saving research persona for user {user_id}: {e}")
|
|
self.db.rollback()
|
|
return False
|
|
|
|
def _get_persona_data_record(self, user_id: str) -> Optional[PersonaData]:
|
|
"""Get PersonaData database record for user."""
|
|
try:
|
|
# Ensure research_persona columns exist before querying
|
|
self.onboarding_service._ensure_research_persona_columns(self.db)
|
|
|
|
# Get onboarding session
|
|
session = self.db.query(OnboardingSession).filter(
|
|
OnboardingSession.user_id == user_id
|
|
).first()
|
|
|
|
if not session:
|
|
return None
|
|
|
|
# Get persona data
|
|
persona_data = self.db.query(PersonaData).filter(
|
|
PersonaData.session_id == session.id
|
|
).first()
|
|
|
|
return persona_data
|
|
|
|
except Exception as e:
|
|
logger.error(f"Error getting persona data record for user {user_id}: {e}")
|
|
return None
|
|
|
|
def _collect_onboarding_data(self, user_id: str) -> Optional[Dict[str, Any]]:
|
|
"""
|
|
Collect all onboarding data needed for research persona generation.
|
|
|
|
Returns:
|
|
Dictionary with website_analysis, persona_data, research_preferences, business_info
|
|
"""
|
|
try:
|
|
# Get website analysis
|
|
website_analysis = self.onboarding_service.get_website_analysis(user_id, self.db) or {}
|
|
|
|
# Get persona data
|
|
persona_data_dict = self.onboarding_service.get_persona_data(user_id, self.db) or {}
|
|
|
|
# Get research preferences
|
|
research_prefs = self.onboarding_service.get_research_preferences(user_id, self.db) or {}
|
|
|
|
# Get business info - construct from persona data and website analysis
|
|
business_info = {}
|
|
|
|
# Try to extract from persona data
|
|
if persona_data_dict:
|
|
core_persona = persona_data_dict.get('corePersona') or persona_data_dict.get('core_persona')
|
|
if core_persona:
|
|
if core_persona.get('industry'):
|
|
business_info['industry'] = core_persona['industry']
|
|
if core_persona.get('target_audience'):
|
|
business_info['target_audience'] = core_persona['target_audience']
|
|
|
|
# Fallback to website analysis if not in persona
|
|
if not business_info.get('industry') and website_analysis:
|
|
target_audience_data = website_analysis.get('target_audience', {})
|
|
if isinstance(target_audience_data, dict):
|
|
industry_focus = target_audience_data.get('industry_focus')
|
|
if industry_focus:
|
|
business_info['industry'] = industry_focus
|
|
demographics = target_audience_data.get('demographics')
|
|
if demographics:
|
|
business_info['target_audience'] = demographics if isinstance(demographics, str) else str(demographics)
|
|
|
|
# Check if we have enough data - be more lenient since we can infer from minimal data
|
|
# We need at least some basic information to generate a meaningful persona
|
|
has_basic_data = bool(
|
|
website_analysis or
|
|
persona_data_dict or
|
|
research_prefs.get('content_types') or
|
|
business_info.get('industry')
|
|
)
|
|
|
|
if not has_basic_data:
|
|
logger.warning(f"Insufficient onboarding data for user {user_id} - no basic data found")
|
|
return None
|
|
|
|
# If we have minimal data, add intelligent defaults to help the AI
|
|
if not business_info.get('industry'):
|
|
# Try to infer industry from research preferences or content types
|
|
content_types = research_prefs.get('content_types', [])
|
|
if 'blog' in content_types or 'article' in content_types:
|
|
business_info['industry'] = 'Content Marketing'
|
|
business_info['inferred'] = True
|
|
elif 'social_media' in content_types:
|
|
business_info['industry'] = 'Social Media Marketing'
|
|
business_info['inferred'] = True
|
|
elif 'video' in content_types:
|
|
business_info['industry'] = 'Video Content Creation'
|
|
business_info['inferred'] = True
|
|
|
|
if not business_info.get('target_audience'):
|
|
# Default to professionals for content creators
|
|
business_info['target_audience'] = 'Professionals and content consumers'
|
|
business_info['inferred'] = True
|
|
|
|
# Get competitor analysis data (if available)
|
|
competitor_analysis = None
|
|
try:
|
|
competitor_analysis = self.onboarding_service.get_competitor_analysis(user_id, self.db)
|
|
if competitor_analysis:
|
|
logger.info(f"Found {len(competitor_analysis)} competitors for research persona generation")
|
|
except Exception as e:
|
|
logger.debug(f"Could not retrieve competitor analysis for persona generation: {e}")
|
|
|
|
return {
|
|
"website_analysis": website_analysis,
|
|
"persona_data": persona_data_dict,
|
|
"research_preferences": research_prefs,
|
|
"business_info": business_info,
|
|
"competitor_analysis": competitor_analysis # Add competitor data for better preset generation
|
|
}
|
|
|
|
except Exception as e:
|
|
logger.error(f"Error collecting onboarding data for user {user_id}: {e}")
|
|
return None
|