nagur-shareef-shaik's picture
Added Plan Recommendations Functionality
28baf2e
Raw History Blame Contribute Delete
4.08 kB
import logging
from enum import Enum
from typing import List, Optional, Any, Type, Dict
from pydantic import BaseModel, Field
class IntentType(str, Enum):
"""
Enum for the types of user query intents.
This provides a controlled vocabulary for our RAG strategies.
"""
SIMPLE = "Simple"
AMBIGUOUS = "Ambiguous (RAG-Fusion)"
COMPLEX = "Complex (Decomposition)"
CONCISE = "Concise (Step-Back)"
class QueryIntent(BaseModel):
"""
Pydantic model for the structured output of the query intent classifier.
This model ensures the LLM's response is predictable and directly usable.
"""
intent: IntentType = Field(
...,
description="The classified intent of the user's query."
)
reasoning: str = Field(
...,
description="A brief explanation of why the query was classified with this intent."
)
transformed_queries: List[str] = Field(
...,
description=(
"The transformed query/queries suitable for the classified intent. "
"For SIMPLE, this is the original query. "
"For AMBIGUOUS, these are multiple variations for RAG-Fusion. "
"For COMPLEX, these are the decomposed sub-queries. "
"For CONCISE, this is a single 'step-back' question."
)
)
class TransformedQueries(BaseModel):
"""
Pydantic model for the structured output of the query transformation step.
"""
transformed_queries: List[str] = Field(
...,
description=(
"A list of transformed queries suitable for the classified intent. "
"For SIMPLE, this should be the original query in a list. "
"For AMBIGUOUS, these should be multiple variations for RAG-Fusion. "
"For COMPLEX, these should be the decomposed sub-queries. "
"For SPECULATIVE, this should be a single hypothetical document in a list. "
"For CONCISE, this should be a single 'step-back' question in a list."
)
)
class UserProfile(BaseModel):
"""
Pydantic model for the user's 360-degree profile.
"""
# --- Mandatory fields provided initially ---
zip_code: str = Field(..., description="User's 5-digit ZIP code.")
county: str = Field(..., description="User's county, derived from ZIP code.")
state: str = Field(..., description="User's state, derived from ZIP code.")
age: int = Field(..., description="User's age in years.")
gender: str = Field(..., description="User's gender (e.g., 'Male', 'Female', 'Other').")
household_size: int = Field(..., description="Number of people in the user's household.")
income: int = Field(..., description="Annual household income.")
employment_status: str = Field(..., description="User's employment status.")
citizenship: str = Field(..., description="User's citizenship status.")
# --- Fields to be populated by the ProfileAgent ---
medical_history: Optional[List[str]] = Field(
None, description="List of chronic conditions or significant medical history."
)
medications: Optional[List[str]] = Field(
None, description="List of current prescription medications."
)
special_cases: Optional[List[str]] = Field(
None, description="Special circumstances like pregnancy, tobacco use, etc."
)
class GeoDataResponse(BaseModel):
"""
The response model for the /geodata endpoint.
"""
zip_code: str
county: str
city: str
state: str
state_abbreviation: str
class ChatRequest(BaseModel):
"""
The request model for the main /chat endpoint.
"""
thread_id: str
user_profile: Dict[str, Optional[Any]]
message: str
conversation_history: List[str] = Field(default_factory=list)
is_profile_complete: bool
class ChatResponse(BaseModel):
"""
The response model for the /chat endpoint.
"""
agent_response: str
updated_profile: Dict[str, Any]
updated_history: List[str]
is_profile_complete: bool
plan_recommendations: Optional[Dict[str, Any]]