Feat:dataset retiever resource (#1123)

Co-authored-by: jyong <jyong@dify.ai>
Co-authored-by: StyleZhang <jasonapring2015@outlook.com>
This commit is contained in:
Jyong
2023-09-10 15:17:43 +08:00
committed by GitHub
parent e161c511af
commit 642842d61b
32 changed files with 442 additions and 33 deletions

View File

@@ -1,6 +1,6 @@
import decimal
import json
from typing import Optional, Union
from typing import Optional, Union, List
from core.callback_handler.entity.agent_loop import AgentLoop
from core.callback_handler.entity.dataset_query import DatasetQueryObj
@@ -15,7 +15,8 @@ from events.message_event import message_was_created
from extensions.ext_database import db
from extensions.ext_redis import redis_client
from models.dataset import DatasetQuery
from models.model import AppModelConfig, Conversation, Account, Message, EndUser, App, MessageAgentThought, MessageChain
from models.model import AppModelConfig, Conversation, Account, Message, EndUser, App, MessageAgentThought, \
MessageChain, DatasetRetrieverResource
class ConversationMessageTask:
@@ -41,6 +42,8 @@ class ConversationMessageTask:
self.message = None
self.retriever_resource = None
self.model_dict = self.app_model_config.model_dict
self.provider_name = self.model_dict.get('provider')
self.model_name = self.model_dict.get('name')
@@ -157,7 +160,8 @@ class ConversationMessageTask:
self.message.message_tokens = message_tokens
self.message.message_unit_price = message_unit_price
self.message.message_price_unit = message_price_unit
self.message.answer = PromptBuilder.process_template(llm_message.completion.strip()) if llm_message.completion else ''
self.message.answer = PromptBuilder.process_template(
llm_message.completion.strip()) if llm_message.completion else ''
self.message.answer_tokens = answer_tokens
self.message.answer_unit_price = answer_unit_price
self.message.answer_price_unit = answer_price_unit
@@ -256,7 +260,36 @@ class ConversationMessageTask:
db.session.add(dataset_query)
def on_dataset_query_finish(self, resource: List):
if resource and len(resource) > 0:
for item in resource:
dataset_retriever_resource = DatasetRetrieverResource(
message_id=self.message.id,
position=item.get('position'),
dataset_id=item.get('dataset_id'),
dataset_name=item.get('dataset_name'),
document_id=item.get('document_id'),
document_name=item.get('document_name'),
data_source_type=item.get('data_source_type'),
segment_id=item.get('segment_id'),
score=item.get('score') if 'score' in item else None,
hit_count=item.get('hit_count') if 'hit_count' else None,
word_count=item.get('word_count') if 'word_count' in item else None,
segment_position=item.get('segment_position') if 'segment_position' in item else None,
index_node_hash=item.get('index_node_hash') if 'index_node_hash' in item else None,
content=item.get('content'),
retriever_from=item.get('retriever_from'),
created_by=self.user.id
)
db.session.add(dataset_retriever_resource)
db.session.flush()
self.retriever_resource = resource
def message_end(self):
self._pub_handler.pub_message_end(self.retriever_resource)
def end(self):
self._pub_handler.pub_message_end(self.retriever_resource)
self._pub_handler.pub_end()
@@ -350,6 +383,23 @@ class PubHandler:
self.pub_end()
raise ConversationTaskStoppedException()
def pub_message_end(self, retriever_resource: List):
content = {
'event': 'message_end',
'data': {
'task_id': self._task_id,
'message_id': self._message.id,
'mode': self._conversation.mode,
'conversation_id': self._conversation.id
}
}
if retriever_resource:
content['data']['retriever_resources'] = retriever_resource
redis_client.publish(self._channel, json.dumps(content))
if self._is_stopped():
self.pub_end()
raise ConversationTaskStoppedException()
def pub_end(self):
content = {