390 lines
15 KiB
Python
390 lines
15 KiB
Python
from typing import Dict, Any, List, Optional
|
|
from sqlalchemy.orm import Session
|
|
from loguru import logger
|
|
from services.onboarding.data_service import OnboardingDataService
|
|
from services.user_data_service import UserDataService
|
|
from services.llm_providers.gemini_provider import gemini_text_response, gemini_structured_json_response
|
|
|
|
class StrategyCopilotService:
|
|
"""Service for CopilotKit strategy assistance using Gemini."""
|
|
|
|
def __init__(self, db: Session):
|
|
self.db = db
|
|
self.onboarding_service = OnboardingDataService()
|
|
self.user_data_service = UserDataService(db)
|
|
|
|
async def generate_category_data(
|
|
self,
|
|
category: str,
|
|
user_description: str,
|
|
current_form_data: Dict[str, Any]
|
|
) -> Dict[str, Any]:
|
|
"""Generate data for a specific category."""
|
|
try:
|
|
# Get user onboarding data
|
|
user_id = 1 # TODO: Get from auth context
|
|
onboarding_data = self.onboarding_service.get_personalized_ai_inputs(user_id)
|
|
|
|
# Build prompt for category generation
|
|
prompt = self._build_category_generation_prompt(
|
|
category, user_description, current_form_data, onboarding_data
|
|
)
|
|
|
|
# Generate response using Gemini
|
|
response = gemini_text_response(
|
|
prompt=prompt,
|
|
temperature=0.3,
|
|
top_p=0.9,
|
|
n=1,
|
|
max_tokens=2048,
|
|
system_prompt="You are ALwrity's Strategy Assistant. Generate appropriate values for strategy fields."
|
|
)
|
|
|
|
# Parse and validate response
|
|
generated_data = self._parse_category_response(response, category)
|
|
|
|
return generated_data
|
|
|
|
except Exception as e:
|
|
logger.error(f"Error generating category data: {str(e)}")
|
|
raise
|
|
|
|
async def validate_field(self, field_id: str, value: Any) -> Dict[str, Any]:
|
|
"""Validate a specific strategy field."""
|
|
try:
|
|
# Get field definition
|
|
field_definition = self._get_field_definition(field_id)
|
|
|
|
# Build validation prompt
|
|
prompt = self._build_validation_prompt(field_definition, value)
|
|
|
|
# Generate validation response using Gemini
|
|
response = gemini_text_response(
|
|
prompt=prompt,
|
|
temperature=0.2,
|
|
top_p=0.9,
|
|
n=1,
|
|
max_tokens=1024,
|
|
system_prompt="You are ALwrity's Strategy Assistant. Validate field values and provide suggestions."
|
|
)
|
|
|
|
# Parse validation result
|
|
validation_result = self._parse_validation_response(response)
|
|
|
|
return validation_result
|
|
|
|
except Exception as e:
|
|
logger.error(f"Error validating field {field_id}: {str(e)}")
|
|
raise
|
|
|
|
async def analyze_strategy(self, form_data: Dict[str, Any]) -> Dict[str, Any]:
|
|
"""Analyze complete strategy for completeness and coherence."""
|
|
try:
|
|
# Get user data for context
|
|
user_id = 1 # TODO: Get from auth context
|
|
onboarding_data = self.onboarding_service.get_personalized_ai_inputs(user_id)
|
|
|
|
# Build analysis prompt
|
|
prompt = self._build_analysis_prompt(form_data, onboarding_data)
|
|
|
|
# Generate analysis using Gemini
|
|
response = gemini_text_response(
|
|
prompt=prompt,
|
|
temperature=0.3,
|
|
top_p=0.9,
|
|
n=1,
|
|
max_tokens=2048,
|
|
system_prompt="You are ALwrity's Strategy Assistant. Analyze strategies for completeness and coherence."
|
|
)
|
|
|
|
# Parse analysis result
|
|
analysis_result = self._parse_analysis_response(response)
|
|
|
|
return analysis_result
|
|
|
|
except Exception as e:
|
|
logger.error(f"Error analyzing strategy: {str(e)}")
|
|
raise
|
|
|
|
async def generate_field_suggestions(
|
|
self,
|
|
field_id: str,
|
|
current_form_data: Dict[str, Any]
|
|
) -> Dict[str, Any]:
|
|
"""Generate suggestions for a specific field."""
|
|
try:
|
|
# Get field definition
|
|
field_definition = self._get_field_definition(field_id)
|
|
|
|
# Get user data
|
|
user_id = 1 # TODO: Get from auth context
|
|
onboarding_data = self.onboarding_service.get_personalized_ai_inputs(user_id)
|
|
|
|
# Build suggestions prompt
|
|
prompt = self._build_suggestions_prompt(
|
|
field_definition, current_form_data, onboarding_data
|
|
)
|
|
|
|
# Generate suggestions using Gemini
|
|
response = gemini_text_response(
|
|
prompt=prompt,
|
|
temperature=0.4,
|
|
top_p=0.9,
|
|
n=1,
|
|
max_tokens=1024,
|
|
system_prompt="You are ALwrity's Strategy Assistant. Generate helpful suggestions for strategy fields."
|
|
)
|
|
|
|
# Parse suggestions
|
|
suggestions = self._parse_suggestions_response(response)
|
|
|
|
return suggestions
|
|
|
|
except Exception as e:
|
|
logger.error(f"Error generating suggestions for {field_id}: {str(e)}")
|
|
raise
|
|
|
|
def _build_category_generation_prompt(
|
|
self,
|
|
category: str,
|
|
user_description: str,
|
|
current_form_data: Dict[str, Any],
|
|
onboarding_data: Dict[str, Any]
|
|
) -> str:
|
|
"""Build prompt for category data generation."""
|
|
return f"""
|
|
You are ALwrity's Strategy Assistant. Generate data for the {category} category based on the user's description.
|
|
|
|
User Description: {user_description}
|
|
|
|
Current Form Data: {current_form_data}
|
|
|
|
Onboarding Data: {onboarding_data}
|
|
|
|
Category Fields: {self._get_category_fields(category)}
|
|
|
|
Generate appropriate values for all fields in the {category} category. Return only valid JSON with field IDs as keys.
|
|
|
|
Example response format:
|
|
{{
|
|
"field_id": "value",
|
|
"another_field": "value"
|
|
}}
|
|
"""
|
|
|
|
def _build_validation_prompt(self, field_definition: Dict[str, Any], value: Any) -> str:
|
|
"""Build prompt for field validation."""
|
|
return f"""
|
|
Validate the following field value:
|
|
|
|
Field: {field_definition['label']}
|
|
Description: {field_definition['description']}
|
|
Required: {field_definition['required']}
|
|
Type: {field_definition['type']}
|
|
Value: {value}
|
|
|
|
Return JSON with: {{"isValid": boolean, "suggestion": string, "confidence": number}}
|
|
|
|
Example response:
|
|
{{
|
|
"isValid": true,
|
|
"suggestion": "This looks good!",
|
|
"confidence": 0.95
|
|
}}
|
|
"""
|
|
|
|
def _build_analysis_prompt(
|
|
self,
|
|
form_data: Dict[str, Any],
|
|
onboarding_data: Dict[str, Any]
|
|
) -> str:
|
|
"""Build prompt for strategy analysis."""
|
|
return f"""
|
|
Analyze the following content strategy for completeness, coherence, and alignment:
|
|
|
|
Form Data: {form_data}
|
|
Onboarding Data: {onboarding_data}
|
|
|
|
Return JSON with: {{
|
|
"completeness": number,
|
|
"coherence": number,
|
|
"alignment": number,
|
|
"suggestions": [string],
|
|
"missingFields": [string],
|
|
"improvements": [string]
|
|
}}
|
|
|
|
Example response:
|
|
{{
|
|
"completeness": 85,
|
|
"coherence": 90,
|
|
"alignment": 88,
|
|
"suggestions": ["Consider adding more specific metrics"],
|
|
"missingFields": ["content_budget"],
|
|
"improvements": ["Add timeline details"]
|
|
}}
|
|
"""
|
|
|
|
def _build_suggestions_prompt(
|
|
self,
|
|
field_definition: Dict[str, Any],
|
|
current_form_data: Dict[str, Any],
|
|
onboarding_data: Dict[str, Any]
|
|
) -> str:
|
|
"""Build prompt for field suggestions."""
|
|
return f"""
|
|
Generate suggestions for the following field:
|
|
|
|
Field: {field_definition['label']}
|
|
Description: {field_definition['description']}
|
|
Required: {field_definition['required']}
|
|
Type: {field_definition['type']}
|
|
|
|
Current Form Data: {current_form_data}
|
|
Onboarding Data: {onboarding_data}
|
|
|
|
Return JSON with: {{
|
|
"suggestions": [string],
|
|
"reasoning": string,
|
|
"confidence": number
|
|
}}
|
|
|
|
Example response:
|
|
{{
|
|
"suggestions": ["Focus on measurable outcomes", "Align with business goals"],
|
|
"reasoning": "Based on your business context, measurable outcomes will be most effective",
|
|
"confidence": 0.92
|
|
}}
|
|
"""
|
|
|
|
def _get_field_definition(self, field_id: str) -> Dict[str, Any]:
|
|
"""Get field definition from STRATEGIC_INPUT_FIELDS."""
|
|
# This would be imported from the frontend field definitions
|
|
# For now, return a basic structure
|
|
return {
|
|
"id": field_id,
|
|
"label": field_id.replace("_", " ").title(),
|
|
"description": f"Description for {field_id}",
|
|
"required": True,
|
|
"type": "text"
|
|
}
|
|
|
|
def _get_category_fields(self, category: str) -> List[str]:
|
|
"""Get fields for a specific category."""
|
|
# This would be imported from the frontend field definitions
|
|
category_fields = {
|
|
"business_context": [
|
|
"business_objectives", "target_metrics", "content_budget", "team_size",
|
|
"implementation_timeline", "market_share", "competitive_position", "performance_metrics"
|
|
],
|
|
"audience_intelligence": [
|
|
"content_preferences", "consumption_patterns", "audience_pain_points",
|
|
"buying_journey", "seasonal_trends", "engagement_metrics"
|
|
],
|
|
"competitive_intelligence": [
|
|
"top_competitors", "competitor_content_strategies", "market_gaps",
|
|
"industry_trends", "emerging_trends"
|
|
],
|
|
"content_strategy": [
|
|
"preferred_formats", "content_mix", "content_frequency", "optimal_timing",
|
|
"quality_metrics", "editorial_guidelines", "brand_voice"
|
|
],
|
|
"performance_analytics": [
|
|
"traffic_sources", "conversion_rates", "content_roi_targets", "ab_testing_capabilities"
|
|
]
|
|
}
|
|
return category_fields.get(category, [])
|
|
|
|
def _parse_category_response(self, response: str, category: str) -> Dict[str, Any]:
|
|
"""Parse LLM response for category data."""
|
|
try:
|
|
import json
|
|
# Clean up the response to extract JSON
|
|
response = response.strip()
|
|
if response.startswith("```json"):
|
|
response = response[7:]
|
|
if response.endswith("```"):
|
|
response = response[:-3]
|
|
response = response.strip()
|
|
|
|
parsed_data = json.loads(response)
|
|
|
|
# Validate that we have actual data
|
|
if not isinstance(parsed_data, dict) or len(parsed_data) == 0:
|
|
raise Exception("Invalid or empty response data")
|
|
|
|
return parsed_data
|
|
except Exception as e:
|
|
logger.error(f"Error parsing category response: {str(e)}")
|
|
raise Exception(f"Failed to parse category response: {str(e)}")
|
|
|
|
def _parse_validation_response(self, response: str) -> Dict[str, Any]:
|
|
"""Parse LLM response for validation."""
|
|
try:
|
|
import json
|
|
# Clean up the response to extract JSON
|
|
response = response.strip()
|
|
if response.startswith("```json"):
|
|
response = response[7:]
|
|
if response.endswith("```"):
|
|
response = response[:-3]
|
|
response = response.strip()
|
|
|
|
parsed_data = json.loads(response)
|
|
|
|
# Validate required fields
|
|
if not isinstance(parsed_data, dict) or 'isValid' not in parsed_data:
|
|
raise Exception("Invalid validation response format")
|
|
|
|
return parsed_data
|
|
except Exception as e:
|
|
logger.error(f"Error parsing validation response: {str(e)}")
|
|
raise Exception(f"Failed to parse validation response: {str(e)}")
|
|
|
|
def _parse_analysis_response(self, response: str) -> Dict[str, Any]:
|
|
"""Parse LLM response for analysis."""
|
|
try:
|
|
import json
|
|
# Clean up the response to extract JSON
|
|
response = response.strip()
|
|
if response.startswith("```json"):
|
|
response = response[7:]
|
|
if response.endswith("```"):
|
|
response = response[:-3]
|
|
response = response.strip()
|
|
|
|
parsed_data = json.loads(response)
|
|
|
|
# Validate required fields
|
|
required_fields = ['completeness', 'coherence', 'alignment']
|
|
if not isinstance(parsed_data, dict) or not all(field in parsed_data for field in required_fields):
|
|
raise Exception("Invalid analysis response format")
|
|
|
|
return parsed_data
|
|
except Exception as e:
|
|
logger.error(f"Error parsing analysis response: {str(e)}")
|
|
raise Exception(f"Failed to parse analysis response: {str(e)}")
|
|
|
|
def _parse_suggestions_response(self, response: str) -> Dict[str, Any]:
|
|
"""Parse LLM response for suggestions."""
|
|
try:
|
|
import json
|
|
# Clean up the response to extract JSON
|
|
response = response.strip()
|
|
if response.startswith("```json"):
|
|
response = response[7:]
|
|
if response.endswith("```"):
|
|
response = response[:-3]
|
|
response = response.strip()
|
|
|
|
parsed_data = json.loads(response)
|
|
|
|
# Validate required fields
|
|
if not isinstance(parsed_data, dict) or 'suggestions' not in parsed_data:
|
|
raise Exception("Invalid suggestions response format")
|
|
|
|
return parsed_data
|
|
except Exception as e:
|
|
logger.error(f"Error parsing suggestions response: {str(e)}")
|
|
raise Exception(f"Failed to parse suggestions response: {str(e)}")
|