- Introduction of PARTNER_RAG retriever, PARTNER_RAG_SPECIALIST and linked Agent and Task, to support documentation inquiries in the management app (eveai_app)
- Addition of a tenant_partner_services view to show partner services from the viewpoint of a tenant - Addition of domain model diagrams - Addition of license_periods views and form
This commit is contained in:
Josako
@@ -0,0 +1,243 @@
|
||||
import json
|
||||
from os import wait
|
||||
from typing import Optional, List, Dict, Any
|
||||
|
||||
from crewai.flow.flow import start, listen, and_
|
||||
from flask import current_app
|
||||
from pydantic import BaseModel, Field
|
||||
|
||||
from common.extensions import db
|
||||
from common.models.interaction import SpecialistRetriever
|
||||
from common.models.user import Partner, PartnerService, PartnerTenant, PartnerRAGRetriever
|
||||
from common.services.utils.translation_services import TranslationServices
|
||||
from common.utils.business_event_context import current_event
|
||||
from eveai_chat_workers.retrievers.base_retriever import BaseRetriever, get_retriever_class
|
||||
from eveai_chat_workers.retrievers.retriever_typing import RetrieverArguments
|
||||
from eveai_chat_workers.specialists.crewai_base_specialist import CrewAIBaseSpecialistExecutor
|
||||
from eveai_chat_workers.specialists.specialist_typing import SpecialistResult, SpecialistArguments
|
||||
from eveai_chat_workers.outputs.globals.rag.rag_v1_0 import RAGOutput
|
||||
from eveai_chat_workers.specialists.crewai_base_classes import EveAICrewAICrew, EveAICrewAIFlow, EveAIFlowState
|
||||
|
||||
INSUFFICIENT_INFORMATION_MESSAGE = (
|
||||
"We do not have the necessary information to provide you with the requested answers. "
|
||||
"Please accept our apologies. Don't hesitate to ask other questions, and I'll do my best to answer them.")
|
||||
|
||||
|
||||
class SpecialistExecutor(CrewAIBaseSpecialistExecutor):
|
||||
"""
|
||||
type: RAG_SPECIALIST
|
||||
type_version: 1.0
|
||||
RAG Specialist Executor class
|
||||
"""
|
||||
|
||||
def __init__(self, tenant_id, specialist_id, session_id, task_id, **kwargs):
|
||||
self.rag_crew = None
|
||||
|
||||
super().__init__(tenant_id, specialist_id, session_id, task_id)
|
||||
|
||||
@property
|
||||
def type(self) -> str:
|
||||
return "RAG_SPECIALIST"
|
||||
|
||||
@property
|
||||
def type_version(self) -> str:
|
||||
return "1.1"
|
||||
|
||||
def _initialize_retrievers(self) -> List[BaseRetriever]:
|
||||
"""Initialize all retrievers associated with this specialist"""
|
||||
retrievers = []
|
||||
|
||||
partner_ids = (
|
||||
db.session.query(Partner.id)
|
||||
.join(PartnerService, Partner.id == PartnerService.partner_id)
|
||||
.join(PartnerTenant, PartnerService.id == PartnerTenant.partner_service_id)
|
||||
.filter(PartnerTenant.tenant_id == self.tenant_id)
|
||||
.distinct()
|
||||
.all()
|
||||
)
|
||||
|
||||
# Extract the actual partner IDs from the query result
|
||||
partner_ids_list = [partner_id[0] for partner_id in partner_ids]
|
||||
|
||||
# Get all corresponding PartnerRagRetrievers for the partner_ids list
|
||||
partner_rag_retrievers = (
|
||||
PartnerRAGRetriever.query
|
||||
.filter(PartnerRAGRetriever.partner_id.in_(partner_ids_list))
|
||||
.filter(PartnerRAGRetriever.tenant_id == self.tenant_id)
|
||||
.all()
|
||||
)
|
||||
|
||||
retriever_executor_class = get_retriever_class("PARTNER_RAG", "1.0")
|
||||
# Get retriever associations from database
|
||||
|
||||
self.log_tuning("_initialize_retrievers", {"Nr of partner retrievers": len(partner_rag_retrievers)})
|
||||
|
||||
for partner_rag_retriever in partner_rag_retrievers :
|
||||
# Get retriever configuration from database
|
||||
self.log_tuning("_initialize_retrievers", {
|
||||
"Partner id": partner_rag_retriever.partner_id,
|
||||
"Tenant id": partner_rag_retriever.tenant_id,
|
||||
"Retriever id": partner_rag_retriever.retriever_id,
|
||||
})
|
||||
|
||||
retriever_executor = retriever_executor_class(partner_rag_retriever.tenant_id,
|
||||
partner_rag_retriever.retriever_id)
|
||||
# Initialize retriever with its configuration
|
||||
retrievers.append(retriever_executor)
|
||||
|
||||
return retrievers
|
||||
|
||||
def _config_task_agents(self):
|
||||
self._add_task_agent("rag_task", "rag_agent")
|
||||
|
||||
def _config_pydantic_outputs(self):
|
||||
self._add_pydantic_output("rag_task", RAGOutput, "rag_output")
|
||||
|
||||
def _config_state_result_relations(self):
|
||||
self._add_state_result_relation("rag_output")
|
||||
self._add_state_result_relation("citations")
|
||||
|
||||
def _instantiate_specialist(self):
|
||||
verbose = self.tuning
|
||||
|
||||
rag_agents = [self.rag_agent]
|
||||
rag_tasks = [self.rag_task]
|
||||
self.rag_crew = EveAICrewAICrew(
|
||||
self,
|
||||
"Rag Crew",
|
||||
agents=rag_agents,
|
||||
tasks=rag_tasks,
|
||||
verbose=verbose,
|
||||
)
|
||||
|
||||
self.flow = RAGFlow(
|
||||
self,
|
||||
self.rag_crew,
|
||||
)
|
||||
|
||||
def execute(self, arguments: SpecialistArguments, formatted_context, citations) -> SpecialistResult:
|
||||
self.log_tuning("RAG Specialist execution started", {})
|
||||
|
||||
if not self._cached_session.interactions:
|
||||
specialist_phase = "initial"
|
||||
else:
|
||||
specialist_phase = self._cached_session.interactions[-1].specialist_results.get('phase', 'initial')
|
||||
|
||||
results = None
|
||||
|
||||
match specialist_phase:
|
||||
case "initial":
|
||||
results = self.execute_initial_state(arguments, formatted_context, citations)
|
||||
case "rag":
|
||||
results = self.execute_rag_state(arguments, formatted_context, citations)
|
||||
|
||||
self.log_tuning(f"RAG Specialist execution ended", {"Results": results.model_dump()})
|
||||
|
||||
return results
|
||||
|
||||
def execute_initial_state(self, arguments: SpecialistArguments, formatted_context, citations) -> SpecialistResult:
|
||||
self.log_tuning("RAG Specialist initial_state execution started", {})
|
||||
|
||||
welcome_message = self.specialist.configuration.get('welcome_message', 'Welcome! You can start asking questions')
|
||||
welcome_message = TranslationServices.translate(self.tenant_id, welcome_message, arguments.language)
|
||||
|
||||
self.flow.state.answer = welcome_message
|
||||
self.flow.state.phase = "rag"
|
||||
|
||||
results = RAGSpecialistResult.create_for_type(self.type, self.type_version)
|
||||
|
||||
return results
|
||||
|
||||
def execute_rag_state(self, arguments: SpecialistArguments, formatted_context, citations) -> SpecialistResult:
|
||||
self.log_tuning("RAG Specialist rag_state execution started", {})
|
||||
|
||||
insufficient_info_message = TranslationServices.translate(self.tenant_id,
|
||||
INSUFFICIENT_INFORMATION_MESSAGE,
|
||||
arguments.language)
|
||||
|
||||
formatted_context, citations = self._retrieve_context(arguments)
|
||||
|
||||
if formatted_context:
|
||||
flow_inputs = {
|
||||
"language": arguments.language,
|
||||
"question": arguments.question,
|
||||
"context": formatted_context,
|
||||
"history": self.formatted_history,
|
||||
"name": self.specialist.configuration.get('name', ''),
|
||||
"welcome_message": self.specialist.configuration.get('welcome_message', '')
|
||||
}
|
||||
|
||||
flow_results = self.flow.kickoff(inputs=flow_inputs)
|
||||
|
||||
if flow_results.rag_output.insufficient_info:
|
||||
flow_results.rag_output.answer = insufficient_info_message
|
||||
|
||||
rag_output = flow_results.rag_output
|
||||
else:
|
||||
rag_output = RAGOutput(answer=insufficient_info_message, insufficient_info=True)
|
||||
|
||||
self.flow.state.rag_output = rag_output
|
||||
self.flow.state.citations = citations
|
||||
self.flow.state.answer = rag_output.answer
|
||||
self.flow.state.phase = "rag"
|
||||
|
||||
results = RAGSpecialistResult.create_for_type(self.type, self.type_version)
|
||||
|
||||
return results
|
||||
|
||||
|
||||
class RAGSpecialistInput(BaseModel):
|
||||
language: Optional[str] = Field(None, alias="language")
|
||||
question: Optional[str] = Field(None, alias="question")
|
||||
context: Optional[str] = Field(None, alias="context")
|
||||
history: Optional[str] = Field(None, alias="history")
|
||||
name: Optional[str] = Field(None, alias="name")
|
||||
welcome_message: Optional[str] = Field(None, alias="welcome_message")
|
||||
|
||||
|
||||
class RAGSpecialistResult(SpecialistResult):
|
||||
rag_output: Optional[RAGOutput] = Field(None, alias="Rag Output")
|
||||
|
||||
|
||||
class RAGFlowState(EveAIFlowState):
|
||||
"""Flow state for RAG specialist that automatically updates from task outputs"""
|
||||
input: Optional[RAGSpecialistInput] = None
|
||||
rag_output: Optional[RAGOutput] = None
|
||||
citations: Optional[List[Dict[str, Any]]] = None
|
||||
|
||||
|
||||
class RAGFlow(EveAICrewAIFlow[RAGFlowState]):
|
||||
def __init__(self,
|
||||
specialist_executor: CrewAIBaseSpecialistExecutor,
|
||||
rag_crew: EveAICrewAICrew,
|
||||
**kwargs):
|
||||
super().__init__(specialist_executor, "RAG Specialist Flow", **kwargs)
|
||||
self.specialist_executor = specialist_executor
|
||||
self.rag_crew = rag_crew
|
||||
self.exception_raised = False
|
||||
|
||||
@start()
|
||||
def process_inputs(self):
|
||||
return ""
|
||||
|
||||
@listen(process_inputs)
|
||||
async def execute_rag(self):
|
||||
inputs = self.state.input.model_dump()
|
||||
try:
|
||||
crew_output = await self.rag_crew.kickoff_async(inputs=inputs)
|
||||
self.specialist_executor.log_tuning("RAG Crew Output", crew_output.model_dump())
|
||||
output_pydantic = crew_output.pydantic
|
||||
if not output_pydantic:
|
||||
raw_json = json.loads(crew_output.raw)
|
||||
output_pydantic = RAGOutput.model_validate(raw_json)
|
||||
self.state.rag_output = output_pydantic
|
||||
return crew_output
|
||||
except Exception as e:
|
||||
current_app.logger.error(f"CREW rag_crew Kickoff Error: {str(e)}")
|
||||
self.exception_raised = True
|
||||
raise e
|
||||
|
||||
async def kickoff_async(self, inputs=None):
|
||||
self.state.input = RAGSpecialistInput.model_validate(inputs)
|
||||
result = await super().kickoff_async(inputs)
|
||||
return self.state
|
||||
@@ -1,239 +0,0 @@
|
||||
from datetime import datetime
|
||||
from typing import Dict, Any, List
|
||||
from flask import current_app
|
||||
from langchain_core.exceptions import LangChainException
|
||||
from langchain_core.output_parsers import StrOutputParser
|
||||
from langchain_core.prompts import ChatPromptTemplate
|
||||
from langchain_core.runnables import RunnableParallel, RunnablePassthrough
|
||||
|
||||
from common.langchain.outputs.base import OutputRegistry
|
||||
from common.langchain.outputs.rag import RAGOutput
|
||||
from common.utils.business_event_context import current_event
|
||||
from eveai_chat_workers.specialists.specialist_typing import SpecialistArguments, SpecialistResult
|
||||
from eveai_chat_workers.chat_session_cache import get_chat_history
|
||||
from common.models.interaction import Specialist
|
||||
from common.utils.model_utils import create_language_template, replace_variable_in_template, \
|
||||
get_template
|
||||
from eveai_chat_workers.specialists.base_specialist import BaseSpecialistExecutor
|
||||
from eveai_chat_workers.retrievers.retriever_typing import RetrieverArguments
|
||||
|
||||
|
||||
class SpecialistExecutor(BaseSpecialistExecutor):
|
||||
"""
|
||||
type: STANDARD_RAG
|
||||
type_version: 1.0
|
||||
Standard Q&A RAG Specialist implementation that combines retriever results
|
||||
with LLM processing to generate answers.
|
||||
"""
|
||||
|
||||
def __init__(self, tenant_id: int, specialist_id: int, session_id: str, task_id: str):
|
||||
super().__init__(tenant_id, specialist_id, session_id, task_id)
|
||||
|
||||
# Check and load the specialist
|
||||
specialist = Specialist.query.get_or_404(specialist_id)
|
||||
# Set the specific configuration for the RAG Specialist
|
||||
self.specialist_context = specialist.configuration.get('specialist_context', '')
|
||||
self.temperature = specialist.configuration.get('temperature', 0.3)
|
||||
self.tuning = specialist.tuning
|
||||
|
||||
# Initialize retrievers
|
||||
self.retrievers = self._initialize_retrievers()
|
||||
|
||||
@property
|
||||
def type(self) -> str:
|
||||
return "STANDARD_RAG_SPECIALIST"
|
||||
|
||||
@property
|
||||
def type_version(self) -> str:
|
||||
return "1.0"
|
||||
|
||||
@property
|
||||
def required_templates(self) -> List[str]:
|
||||
"""List of required templates for this specialist"""
|
||||
return ['rag', 'history']
|
||||
|
||||
def _detail_question(self, language: str, question: str) -> str:
|
||||
"""Detail question based on conversation history"""
|
||||
try:
|
||||
# Get cached session history
|
||||
cached_session = get_chat_history(self.session_id)
|
||||
|
||||
# Format history for the prompt
|
||||
formatted_history = "\n\n".join([
|
||||
f"HUMAN:\n{interaction.specialist_results.get('detailed_query')}\n\n"
|
||||
f"AI:\n{interaction.specialist_results.get('answer')}"
|
||||
for interaction in cached_session.interactions
|
||||
])
|
||||
|
||||
# Get LLM and template
|
||||
template, llm = get_template("history", temperature=0.3)
|
||||
language_template = create_language_template(template, language)
|
||||
|
||||
# Create prompt
|
||||
history_prompt = ChatPromptTemplate.from_template(language_template)
|
||||
|
||||
# Create chain
|
||||
chain = (
|
||||
history_prompt |
|
||||
llm |
|
||||
StrOutputParser()
|
||||
)
|
||||
|
||||
# Execute chain
|
||||
detailed_question = chain.invoke({
|
||||
"history": formatted_history,
|
||||
"question": question
|
||||
})
|
||||
|
||||
if self.tuning:
|
||||
self.log_tuning("_detail_question", {
|
||||
"cached_session_id": cached_session.session_id,
|
||||
"cached_session.interactions": str(cached_session.interactions),
|
||||
"original_question": question,
|
||||
"history_used": formatted_history,
|
||||
"detailed_question": detailed_question,
|
||||
})
|
||||
|
||||
return detailed_question
|
||||
|
||||
except Exception as e:
|
||||
current_app.logger.error(f"Error detailing question: {e}")
|
||||
return question # Fallback to original question
|
||||
|
||||
def execute(self, arguments: SpecialistArguments) -> SpecialistResult:
|
||||
"""
|
||||
Execute the RAG specialist to generate an answer
|
||||
"""
|
||||
start_time = datetime.now()
|
||||
|
||||
try:
|
||||
with current_event.create_span("Specialist Detail Question"):
|
||||
self.update_progress("Detail Question Start", {})
|
||||
# Get required arguments
|
||||
language = arguments.language
|
||||
query = arguments.query
|
||||
detailed_question = self._detail_question(language, query)
|
||||
self.update_progress("Detail Question End", {})
|
||||
|
||||
# Log the start of retrieval process if tuning is enabled
|
||||
with current_event.create_span("Specialist Retrieval"):
|
||||
self.log_tuning("Starting context retrieval", {
|
||||
"num_retrievers": len(self.retrievers),
|
||||
"all arguments": arguments.model_dump(),
|
||||
})
|
||||
self.update_progress("EveAI Retriever Start", {})
|
||||
|
||||
# Get retriever-specific arguments
|
||||
retriever_arguments = arguments.retriever_arguments
|
||||
|
||||
# Collect context from all retrievers
|
||||
all_context = []
|
||||
for retriever in self.retrievers:
|
||||
# Get arguments for this specific retriever
|
||||
retriever_id = str(retriever.retriever_id)
|
||||
if retriever_id not in retriever_arguments:
|
||||
current_app.logger.error(f"Missing arguments for retriever {retriever_id}")
|
||||
continue
|
||||
|
||||
# Get the retriever's arguments and update the query
|
||||
current_retriever_args = retriever_arguments[retriever_id]
|
||||
if isinstance(retriever_arguments[retriever_id], RetrieverArguments):
|
||||
updated_args = current_retriever_args.model_dump()
|
||||
updated_args['query'] = detailed_question
|
||||
retriever_args = RetrieverArguments(**updated_args)
|
||||
else:
|
||||
# Create a new RetrieverArguments instance from the dictionary
|
||||
current_retriever_args['query'] = detailed_question
|
||||
retriever_args = RetrieverArguments(**current_retriever_args)
|
||||
|
||||
# Each retriever gets its own specific arguments
|
||||
retriever_result = retriever.retrieve(retriever_args)
|
||||
all_context.extend(retriever_result)
|
||||
|
||||
# Sort by similarity if available and get unique contexts
|
||||
all_context.sort(key=lambda x: x.similarity, reverse=True)
|
||||
unique_contexts = []
|
||||
seen_chunks = set()
|
||||
for ctx in all_context:
|
||||
if ctx.chunk not in seen_chunks:
|
||||
unique_contexts.append(ctx)
|
||||
seen_chunks.add(ctx.chunk)
|
||||
|
||||
self.log_tuning("Context retrieval completed", {
|
||||
"total_contexts": len(all_context),
|
||||
"unique_contexts": len(unique_contexts),
|
||||
"average_similarity": sum(ctx.similarity for ctx in unique_contexts) / len(
|
||||
unique_contexts) if unique_contexts else 0
|
||||
})
|
||||
self.update_progress("EveAI Retriever Complete", {})
|
||||
|
||||
# Prepare context for LLM
|
||||
formatted_context = "\n\n".join([
|
||||
f"SOURCE: {ctx.metadata.document_id}\n\n{ctx.chunk}"
|
||||
for ctx in unique_contexts
|
||||
])
|
||||
|
||||
with current_event.create_span("Specialist RAG invocation"):
|
||||
try:
|
||||
self.update_progress(self.task_id, "EveAI Chain Start", {})
|
||||
template, llm = get_template("rag", self.temperature)
|
||||
language_template = create_language_template(template, language)
|
||||
full_template = replace_variable_in_template(
|
||||
language_template,
|
||||
"{tenant_context}",
|
||||
self.specialist_context
|
||||
)
|
||||
|
||||
if self.tuning:
|
||||
self.log_tuning("Template preparation completed", {
|
||||
"template": full_template,
|
||||
"context": formatted_context,
|
||||
"tenant_context": self.specialist_context,
|
||||
})
|
||||
|
||||
# Create prompt
|
||||
rag_prompt = ChatPromptTemplate.from_template(full_template)
|
||||
|
||||
# Setup chain components
|
||||
setup_and_retrieval = RunnableParallel({
|
||||
"context": lambda x: formatted_context,
|
||||
"question": lambda x: x
|
||||
})
|
||||
|
||||
# Get output schema for structured output
|
||||
output_schema = OutputRegistry.get_schema(self.type)
|
||||
structured_llm = llm.with_structured_output(output_schema)
|
||||
chain = setup_and_retrieval | rag_prompt | structured_llm
|
||||
|
||||
raw_result = chain.invoke(detailed_question)
|
||||
result = SpecialistResult.create_for_type(
|
||||
self.type,
|
||||
self.type_version,
|
||||
detailed_question=detailed_question,
|
||||
answer=raw_result.answer,
|
||||
citations=[ctx.metadata.document_id for ctx in unique_contexts
|
||||
if ctx.id in raw_result.citations],
|
||||
insufficient_info=raw_result.insufficient_info
|
||||
)
|
||||
|
||||
if self.tuning:
|
||||
self.log_tuning("LLM chain execution completed", {
|
||||
"Result": result.model_dump()
|
||||
})
|
||||
self.update_progress("EveAI Chain Complete", {})
|
||||
|
||||
except Exception as e:
|
||||
current_app.logger.error(f"Error in LLM processing: {e}")
|
||||
if self.tuning:
|
||||
self.log_tuning("LLM processing error", {"error": str(e)})
|
||||
raise
|
||||
|
||||
return result
|
||||
|
||||
except Exception as e:
|
||||
current_app.logger.error(f'Error in RAG specialist execution: {str(e)}')
|
||||
raise
|
||||
|
||||
|
||||
# Register the specialist type
|
||||
OutputRegistry.register("STANDARD_RAG_SPECIALIST", RAGOutput)
|
||||
Reference in New Issue
Block a user