|
|
@@ -1,6 +1,7 @@
|
|
|
import asyncio
|
|
|
import io
|
|
|
import json
|
|
|
+import uuid
|
|
|
|
|
|
import fitz
|
|
|
from fastapi import HTTPException
|
|
|
@@ -10,7 +11,7 @@ from app.config.agent_base_url import RG_CHAT_DIALOG, DF_CHAT_AGENT, DF_CHAT_PAR
|
|
|
DF_CHAT_WORKFLOW, DF_UPLOAD_FILE, RG_ORIGINAL_URL
|
|
|
from app.config.config import settings
|
|
|
from app.config.const import *
|
|
|
-from app.models import DialogModel, ApiTokenModel, UserTokenModel
|
|
|
+from app.models import DialogModel, ApiTokenModel, UserTokenModel, ComplexChatSessionDao, ChatDataRequest
|
|
|
from app.models.v2.session_model import ChatSessionDao, ChatData
|
|
|
from app.service.v2.app_driver.chat_agent import ChatAgent
|
|
|
from app.service.v2.app_driver.chat_data import ChatBaseApply
|
|
|
@@ -47,6 +48,7 @@ async def add_session_log(db, session_id: str, question: str, chat_id: str, user
|
|
|
logger.error(e)
|
|
|
return None
|
|
|
|
|
|
+
|
|
|
async def get_app_token(db, app_id):
|
|
|
app_token = db.query(UserTokenModel).filter_by(id=app_id).first()
|
|
|
if app_token:
|
|
|
@@ -54,7 +56,6 @@ async def get_app_token(db, app_id):
|
|
|
return ""
|
|
|
|
|
|
|
|
|
-
|
|
|
async def get_chat_token(db, app_id):
|
|
|
app_token = db.query(ApiTokenModel).filter_by(app_id=app_id).first()
|
|
|
if app_token:
|
|
|
@@ -71,7 +72,6 @@ async def add_chat_token(db, data):
|
|
|
logger.error(e)
|
|
|
|
|
|
|
|
|
-
|
|
|
async def get_chat_info(db, chat_id: str):
|
|
|
return db.query(DialogModel).filter_by(id=chat_id, status=Dialog_STATSU_ON).first()
|
|
|
|
|
|
@@ -134,6 +134,7 @@ async def service_chat_dialog(db, chat_id: str, question: str, session_id: str,
|
|
|
message["role"] = "assistant"
|
|
|
await update_session_log(db, session_id, message, conversation_id)
|
|
|
|
|
|
+
|
|
|
async def data_process(data):
|
|
|
if isinstance(data, str):
|
|
|
return data.replace("dify", "smart")
|
|
|
@@ -170,8 +171,9 @@ async def service_chat_workflow(db, chat_id: str, chat_data: ChatData, session_i
|
|
|
if hasattr(chat_data, "query"):
|
|
|
query = chat_data.query
|
|
|
else:
|
|
|
- query = "start new workflow"
|
|
|
- session = await add_session_log(db, session_id,query if query else "start new conversation", chat_id, user_id, mode, conversation_id, 3)
|
|
|
+ query = "start new conversation"
|
|
|
+ session = await add_session_log(db, session_id, query if query else "start new conversation", chat_id, user_id,
|
|
|
+ mode, conversation_id, 3)
|
|
|
if session:
|
|
|
conversation_id = session.conversation_id
|
|
|
try:
|
|
|
@@ -215,7 +217,7 @@ async def service_chat_workflow(db, chat_id: str, chat_data: ChatData, session_i
|
|
|
[workflow_started, node_started, node_finished].index(ans.get("event"))]
|
|
|
elif ans.get("event") == workflow_finished:
|
|
|
data = ans.get("data", {})
|
|
|
- answer_workflow = data.get("outputs", {}).get("output")
|
|
|
+ answer_workflow = data.get("outputs", {}).get("output", data.get("outputs", {}).get("answer"))
|
|
|
download_url = data.get("outputs", {}).get("download_url")
|
|
|
event = smart_workflow_finished
|
|
|
if data.get("status") == "failed":
|
|
|
@@ -242,8 +244,9 @@ async def service_chat_workflow(db, chat_id: str, chat_data: ChatData, session_i
|
|
|
except:
|
|
|
...
|
|
|
finally:
|
|
|
- await update_session_log(db, session_id, {"role": "assistant", "answer": answer_event or answer_agent or answer_workflow or error,
|
|
|
- "download_url":download_url,
|
|
|
+ await update_session_log(db, session_id, {"role": "assistant",
|
|
|
+ "answer": answer_event or answer_agent or answer_workflow or error,
|
|
|
+ "download_url": download_url,
|
|
|
"node_list": node_list, "task_id": task_id, "id": message_id,
|
|
|
"error": error}, conversation_id)
|
|
|
|
|
|
@@ -258,6 +261,7 @@ async def service_chat_parameters(db, chat_id, user_id):
|
|
|
return {}
|
|
|
return chat_info.parameters
|
|
|
|
|
|
+
|
|
|
async def service_chat_sessions(db, chat_id, name):
|
|
|
token = await get_chat_token(db, rg_api_token)
|
|
|
# print(token)
|
|
|
@@ -276,14 +280,12 @@ async def service_chat_sessions_list(db, chat_id, current, page_size, user_id, k
|
|
|
page=current,
|
|
|
page_size=page_size
|
|
|
)
|
|
|
- return json.dumps({"total":total, "rows": [session.to_dict() for session in session_list]})
|
|
|
-
|
|
|
+ return json.dumps({"total": total, "rows": [session.to_dict() for session in session_list]})
|
|
|
|
|
|
|
|
|
async def service_chat_session_log(db, session_id):
|
|
|
session_log = await ChatSessionDao(db).get_session_by_id(session_id)
|
|
|
- return json.dumps(session_log.log_to_json())
|
|
|
-
|
|
|
+ return json.dumps(session_log.log_to_json() if session_log else {})
|
|
|
|
|
|
|
|
|
async def service_chat_upload(db, chat_id, file, user_id):
|
|
|
@@ -316,6 +318,7 @@ async def get_str_token(input_str):
|
|
|
tokens = tokenizer.encode(input_str)
|
|
|
return len(tokens)
|
|
|
|
|
|
+
|
|
|
async def read_pdf(pdf_stream):
|
|
|
text = ""
|
|
|
with fitz.open(stream=pdf_stream, filetype="pdf") as pdf_document:
|
|
|
@@ -335,6 +338,7 @@ async def read_word(word_stream):
|
|
|
|
|
|
return text
|
|
|
|
|
|
+
|
|
|
async def read_file(file, filename, content_type):
|
|
|
text = ""
|
|
|
if content_type == "application/pdf" or filename.endswith('.pdf'):
|
|
|
@@ -349,7 +353,7 @@ async def read_file(file, filename, content_type):
|
|
|
|
|
|
|
|
|
async def service_chunk_retrieval(query, knowledge_id, top_k, similarity_threshold, api_key):
|
|
|
- print(query)
|
|
|
+ # print(query)
|
|
|
|
|
|
try:
|
|
|
request_data = json.loads(query)
|
|
|
@@ -357,7 +361,7 @@ async def service_chunk_retrieval(query, knowledge_id, top_k, similarity_thresho
|
|
|
"question": request_data.get("query", ""),
|
|
|
"dataset_ids": request_data.get("dataset_ids", []),
|
|
|
"page_size": top_k,
|
|
|
- "similarity_threshold": similarity_threshold
|
|
|
+ "similarity_threshold": similarity_threshold if similarity_threshold else 0.2
|
|
|
}
|
|
|
except json.JSONDecodeError as e:
|
|
|
fixed_json = query.replace("'", '"')
|
|
|
@@ -367,15 +371,16 @@ async def service_chunk_retrieval(query, knowledge_id, top_k, similarity_thresho
|
|
|
"question": request_data.get("query", ""),
|
|
|
"dataset_ids": request_data.get("dataset_ids", []),
|
|
|
"page_size": top_k,
|
|
|
- "similarity_threshold": similarity_threshold
|
|
|
+ "similarity_threshold": similarity_threshold if similarity_threshold else 0.2
|
|
|
}
|
|
|
except Exception:
|
|
|
payload = {
|
|
|
- "question":query,
|
|
|
- "dataset_ids":[knowledge_id],
|
|
|
+ "question": query,
|
|
|
+ "dataset_ids": [knowledge_id],
|
|
|
"page_size": top_k,
|
|
|
- "similarity_threshold": similarity_threshold
|
|
|
+ "similarity_threshold": similarity_threshold if similarity_threshold else 0.2
|
|
|
}
|
|
|
+ # print(payload)
|
|
|
url = settings.fwr_base_url + RG_ORIGINAL_URL
|
|
|
chat = ChatBaseApply()
|
|
|
response = await chat.chat_post(url, payload, await chat.get_headers(api_key))
|
|
|
@@ -388,12 +393,16 @@ async def service_chunk_retrieval(query, knowledge_id, top_k, similarity_thresho
|
|
|
"title": chunk.get("document_keyword", "Unknown Document"),
|
|
|
"metadata": {"document_id": chunk["document_id"],
|
|
|
"path": f"{settings.fwr_base_url}/document/{chunk['document_id']}?ext={chunk.get('document_keyword').split('.')[-1]}&prefix=document",
|
|
|
- 'highlight': chunk.get("highlight") , "image_id": chunk.get("image_id"), "positions": chunk.get("positions"),}
|
|
|
+ 'highlight': chunk.get("highlight"), "image_id": chunk.get("image_id"),
|
|
|
+ "positions": chunk.get("positions"), }
|
|
|
}
|
|
|
for chunk in response.get("data", {}).get("chunks", [])
|
|
|
]
|
|
|
+ # print(len(records))
|
|
|
+ # print(records)
|
|
|
return records
|
|
|
|
|
|
+
|
|
|
async def service_base_chunk_retrieval(query, knowledge_id, top_k, similarity_threshold, api_key):
|
|
|
# request_data = json.loads(query)
|
|
|
payload = {
|
|
|
@@ -420,15 +429,170 @@ async def service_base_chunk_retrieval(query, knowledge_id, top_k, similarity_th
|
|
|
return records
|
|
|
|
|
|
|
|
|
+async def add_complex_log(db, message_id, chat_id, session_id, chat_mode, query, user_id, mode, agent_type, message_type, conversation_id="", node_data=None, query_data=None):
|
|
|
+ if not node_data:
|
|
|
+ node_data = []
|
|
|
+ if not query_data:
|
|
|
+ query_data = {}
|
|
|
+ try:
|
|
|
+ complex_log = ComplexChatSessionDao(db)
|
|
|
+ if not conversation_id:
|
|
|
+ session = await complex_log.get_session_by_session_id(session_id, chat_id)
|
|
|
+ if session:
|
|
|
+ conversation_id = session.conversation_id
|
|
|
+ await complex_log.create_session(message_id,
|
|
|
+ chat_id=chat_id,
|
|
|
+ session_id=session_id,
|
|
|
+ chat_mode=chat_mode,
|
|
|
+ message_type=message_type,
|
|
|
+ content=query,
|
|
|
+ event_type=mode,
|
|
|
+ tenant_id=user_id,
|
|
|
+ conversation_id=conversation_id,
|
|
|
+ node_data=json.dumps(node_data),
|
|
|
+ query=json.dumps(query_data),
|
|
|
+ agent_type=agent_type)
|
|
|
+ return conversation_id, True
|
|
|
+
|
|
|
+ except Exception as e:
|
|
|
+ logger.error(e)
|
|
|
+ return conversation_id, False
|
|
|
+
|
|
|
+
|
|
|
+
|
|
|
+async def service_complex_chat(db, chat_id, mode, user_id, chat_request: ChatDataRequest):
|
|
|
+ answer_event = ""
|
|
|
+ answer_agent = ""
|
|
|
+ answer_workflow = ""
|
|
|
+ download_url = ""
|
|
|
+ message_id = ""
|
|
|
+ task_id = ""
|
|
|
+ error = ""
|
|
|
+ files = []
|
|
|
+ node_list = []
|
|
|
+ token = await get_chat_token(db, chat_id)
|
|
|
+ chat, url = await get_chat_object(mode)
|
|
|
+ conversation_id, message = await add_complex_log(db, str(uuid.uuid4()),chat_id, chat_request.sessionId, chat_request.chatMode, chat_request.query, user_id, mode, DF_TYPE, 1, query_data=chat_request.to_dict())
|
|
|
+ if not message:
|
|
|
+ yield "data: " + json.dumps({"message": smart_message_error,
|
|
|
+ "error": "\n**ERROR**: 创建会话失败!", "status": http_500},
|
|
|
+ ensure_ascii=False) + "\n\n"
|
|
|
+ return
|
|
|
+ inputs = {"is_deep": chat_request.isDeep}
|
|
|
+ if chat_request.chatMode == complex_knowledge_chat:
|
|
|
+ inputs["query_json"] = json.dumps({"query": chat_request.query, "dataset_ids": chat_request.knowledgeId})
|
|
|
+
|
|
|
+ try:
|
|
|
+ async for ans in chat.chat_completions(url,
|
|
|
+ await chat.complex_request_data(chat_request.query, conversation_id, str(user_id), files=chat_request.files, inputs=inputs),
|
|
|
+ await chat.get_headers(token)):
|
|
|
+ # print(ans)
|
|
|
+ data = {}
|
|
|
+ status = http_200
|
|
|
+ conversation_id = ans.get("conversation_id")
|
|
|
+ task_id = ans.get("task_id")
|
|
|
+ if ans.get("event") == message_error:
|
|
|
+ error = ans.get("message", "参数异常!")
|
|
|
+ status = http_400
|
|
|
+ event = smart_message_error
|
|
|
+ elif ans.get("event") == message_agent:
|
|
|
+ data = {"answer": ans.get("answer", ""), "id": ans.get("message_id", "")}
|
|
|
+ answer_agent += ans.get("answer", "")
|
|
|
+ message_id = ans.get("message_id", "")
|
|
|
+ event = smart_message_stream
|
|
|
+ elif ans.get("event") == message_event:
|
|
|
+ data = {"answer": ans.get("answer", ""), "id": ans.get("message_id", "")}
|
|
|
+ answer_event += ans.get("answer", "")
|
|
|
+ message_id = ans.get("message_id", "")
|
|
|
+ event = smart_message_stream
|
|
|
+ elif ans.get("event") == message_file:
|
|
|
+ data = {"url": ans.get("url", ""), "id": ans.get("id", ""),
|
|
|
+ "type": ans.get("type", "")}
|
|
|
+ files.append(data)
|
|
|
+ event = smart_message_file
|
|
|
+ elif ans.get("event") in [workflow_started, node_started, node_finished]:
|
|
|
+ data = ans.get("data", {})
|
|
|
+ data["inputs"] = await data_process(data.get("inputs", {}))
|
|
|
+ data["outputs"] = await data_process(data.get("outputs", {}))
|
|
|
+ data["files"] = await data_process(data.get("files", []))
|
|
|
+ data["process_data"] = ""
|
|
|
+ if data.get("status") == "failed":
|
|
|
+ status = http_500
|
|
|
+ error = data.get("error", "")
|
|
|
+ node_list.append(ans)
|
|
|
+ event = [smart_workflow_started, smart_node_started, smart_node_finished][
|
|
|
+ [workflow_started, node_started, node_finished].index(ans.get("event"))]
|
|
|
+ elif ans.get("event") == workflow_finished:
|
|
|
+ data = ans.get("data", {})
|
|
|
+ answer_workflow = data.get("outputs", {}).get("output", data.get("outputs", {}).get("answer"))
|
|
|
+ download_url = data.get("outputs", {}).get("download_url")
|
|
|
+ event = smart_workflow_finished
|
|
|
+ if data.get("status") == "failed":
|
|
|
+ status = http_500
|
|
|
+ error = data.get("error", "")
|
|
|
+ node_list.append(ans)
|
|
|
+
|
|
|
+ elif ans.get("event") == message_end:
|
|
|
+ event = smart_message_end
|
|
|
+ else:
|
|
|
+ continue
|
|
|
+
|
|
|
+ yield "data: " + json.dumps(
|
|
|
+ {"event": event, "data": data, "error": error, "status": status, "task_id": task_id, "message_id":message_id,
|
|
|
+ "session_id": chat_request.sessionId},
|
|
|
+ ensure_ascii=False) + "\n\n"
|
|
|
+
|
|
|
+ except Exception as e:
|
|
|
+ logger.error(e)
|
|
|
+ try:
|
|
|
+ yield "data: " + json.dumps({"message": smart_message_error,
|
|
|
+ "error": "\n**ERROR**: " + str(e), "status": http_500},
|
|
|
+ ensure_ascii=False) + "\n\n"
|
|
|
+ except:
|
|
|
+ ...
|
|
|
+ finally:
|
|
|
+ # await update_session_log(db, session_id, {"role": "assistant",
|
|
|
+ # "answer": answer_event or answer_agent or answer_workflow or error,
|
|
|
+ # "download_url": download_url,
|
|
|
+ # "node_list": node_list, "task_id": task_id, "id": message_id,
|
|
|
+ # "error": error}, conversation_id)
|
|
|
+ if message_id:
|
|
|
+ await add_complex_log(db, message_id, chat_id, chat_request.sessionId, chat_request.chatMode, answer_event or answer_agent or answer_workflow or error, user_id, mode, DF_TYPE, 2, conversation_id, node_data=node_list, query_data=chat_request.to_dict())
|
|
|
+
|
|
|
+async def service_complex_upload(db, chat_id, file, user_id):
|
|
|
+ files = []
|
|
|
+ token = await get_chat_token(db, chat_id)
|
|
|
+ if not token:
|
|
|
+ return files
|
|
|
+ url = settings.dify_base_url + DF_UPLOAD_FILE
|
|
|
+ chat = ChatBaseApply()
|
|
|
+ for f in file:
|
|
|
+ try:
|
|
|
+ file_content = await f.read()
|
|
|
+ file_upload = await chat.chat_upload(url, {"file": (f.filename, file_content)}, {"user": str(user_id)},
|
|
|
+ {'Authorization': f'Bearer {token}'})
|
|
|
+ # try:
|
|
|
+ # tokens = await read_file(file_content, f.filename, f.content_type)
|
|
|
+ # file_upload["tokens"] = tokens
|
|
|
+ # except:
|
|
|
+ # ...
|
|
|
+ files.append(file_upload)
|
|
|
+ except Exception as e:
|
|
|
+ logger.error(e)
|
|
|
+ return json.dumps(files) if files else ""
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
q = json.dumps({"query": "设备", "dataset_ids": ["fc68db52f43111efb94a0242ac120004"]})
|
|
|
top_k = 2
|
|
|
similarity_threshold = 0.5
|
|
|
api_key = "ragflow-Y4MGYwY2JlZjM2YjExZWY4ZWU5MDI0Mm"
|
|
|
+
|
|
|
+
|
|
|
# a = service_chunk_retrieval(q, top_k, similarity_threshold, api_key)
|
|
|
# print(a)
|
|
|
async def a():
|
|
|
b = await service_chunk_retrieval(q, top_k, similarity_threshold, api_key)
|
|
|
print(b)
|
|
|
- asyncio.run(a())
|
|
|
+
|
|
|
+
|
|
|
+ asyncio.run(a())
|