zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-07-16 04:37:21 +08:00
Code Issues Packages Projects Releases Wiki Activity

Revert "Refa: Chats /chat API to RESTFul (#13871)" (#13877)

Browse Source 
### What problem does this PR solve?

This reverts commit 1a608ac411.

### Type of change

- [x] Other (please describe):
This commit is contained in:
Liu An
2026-04-01 11:05:29 +08:00
committed by GitHub
parent 1a608ac411
commit b1d28b5898
52 changed files with 3584 additions and 2044 deletions
Download Patch File Download Diff File Expand all files Collapse all files

253
api/apps/dialog_app.py Normal file
View File

@@ -0,0 +1,253 @@
#
# Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
from quart import request
from api.db.services import duplicate_name
from api.db.services.dialog_service import DialogService
from common.constants import StatusEnum
from api.db.services.tenant_llm_service import TenantLLMService
from api.db.services.knowledgebase_service import KnowledgebaseService
from api.db.services.user_service import TenantService, UserTenantService
from api.utils.api_utils import get_data_error_result, get_json_result, get_request_json, server_error_response, validate_request
from api.utils.tenant_utils import ensure_tenant_model_id_for_params
from common.misc_utils import get_uuid
from common.constants import RetCode
from api.apps import login_required, current_user
import logging
@manager.route('/set', methods=['POST']) # noqa: F821
@validate_request("prompt_config")
@login_required
async def set_dialog():
req = await get_request_json()
dialog_info = ensure_tenant_model_id_for_params(current_user.id, req)
dialog_id = dialog_info.get("dialog_id", "")
is_create = not dialog_id
name = dialog_info.get("name", "New Dialog")
if not isinstance(name, str):
return get_data_error_result(message="Dialog name must be string.")
if name.strip() == "":
return get_data_error_result(message="Dialog name can't be empty.")
if len(name.encode("utf-8")) > 255:
return get_data_error_result(message=f"Dialog name length is {len(name)} which is larger than 255")
name = name.strip()
if is_create:
# only for chat creating
existing_names = {
d.name.casefold()
for d in DialogService.query(tenant_id=current_user.id, status=StatusEnum.VALID.value)
if d.name
}
if name.casefold() in existing_names:
def _name_exists(name: str, **_kwargs) -> bool:
return name.casefold() in existing_names
name = duplicate_name(_name_exists, name=name)
description = dialog_info.get("description", "A helpful dialog")
icon = dialog_info.get("icon", "")
top_n = dialog_info.get("top_n", 6)
top_k = dialog_info.get("top_k", 1024)
rerank_id = dialog_info.get("rerank_id", "")
if not rerank_id:
dialog_info["rerank_id"] = ""
similarity_threshold = dialog_info.get("similarity_threshold", 0.1)
vector_similarity_weight = dialog_info.get("vector_similarity_weight", 0.3)
llm_setting = dialog_info.get("llm_setting", {})
meta_data_filter = dialog_info.get("meta_data_filter", {})
prompt_config = dialog_info["prompt_config"]
# Set default parameters for datasets with knowledge retrieval
# All datasets with {knowledge} in system prompt need "knowledge" parameter to enable retrieval
kb_ids = dialog_info.get("kb_ids", [])
parameters = prompt_config.get("parameters")
logging.debug(f"set_dialog: kb_ids={kb_ids}, parameters={parameters}, is_create={not is_create}")
# Check if parameters is missing, None, or empty list
if kb_ids and not parameters:
# Check if system prompt uses {knowledge} placeholder
if "{knowledge}" in prompt_config.get("system", ""):
# Set default parameters for any dataset with knowledge placeholder
prompt_config["parameters"] = [{"key": "knowledge", "optional": False}]
logging.debug(f"Set default parameters for datasets with knowledge placeholder: {kb_ids}")
if not is_create:
# only for chat updating
if not dialog_info.get("kb_ids", []) and not prompt_config.get("tavily_api_key") and "{knowledge}" in prompt_config.get("system", ""):
return get_data_error_result(message="Please remove `{knowledge}` in system prompt since no dataset / Tavily used here.")
for p in prompt_config.get("parameters", []):
if p["optional"]:
continue
if prompt_config.get("system", "").find("{%s}" % p["key"]) < 0:
return get_data_error_result(
message="Parameter '{}' is not used".format(p["key"]))
try:
e, tenant = TenantService.get_by_id(current_user.id)
if not e:
return get_data_error_result(message="Tenant not found!")
kbs = KnowledgebaseService.get_by_ids(dialog_info.get("kb_ids", []))
embd_ids = [TenantLLMService.split_model_name_and_factory(kb.embd_id)[0] for kb in kbs] # remove vendor suffix for comparison
embd_count = len(set(embd_ids))
if embd_count > 1:
return get_data_error_result(message=f'Datasets use different embedding models: {[kb.embd_id for kb in kbs]}"')
llm_id = dialog_info.get("llm_id", tenant.llm_id)
tenant_llm_id = dialog_info.get("tenant_llm_id", tenant.tenant_llm_id)
if not dialog_id:
dia = {
"id": get_uuid(),
"tenant_id": current_user.id,
"name": name,
"kb_ids": dialog_info.get("kb_ids", []),
"description": description,
"llm_id": llm_id,
"tenant_llm_id": tenant_llm_id,
"llm_setting": llm_setting,
"prompt_config": prompt_config,
"meta_data_filter": meta_data_filter,
"top_n": top_n,
"top_k": top_k,
"rerank_id": rerank_id,
"tenant_rerank_id": dialog_info.get("tenant_rerank_id", 0),
"similarity_threshold": similarity_threshold,
"vector_similarity_weight": vector_similarity_weight,
"icon": icon
}
if not DialogService.save(**dia):
return get_data_error_result(message="Fail to new a dialog!")
return get_json_result(data=dia)
else:
del dialog_info["dialog_id"]
if "kb_names" in dialog_info:
del dialog_info["kb_names"]
if not DialogService.update_by_id(dialog_id, dialog_info):
return get_data_error_result(message="Dialog not found!")
e, dia = DialogService.get_by_id(dialog_id)
if not e:
return get_data_error_result(message="Fail to update a dialog!")
dia = dia.to_dict()
dia.update(dialog_info)
dia["kb_ids"], dia["kb_names"] = get_kb_names(dia["kb_ids"])
return get_json_result(data=dia)
except Exception as e:
return server_error_response(e)
@manager.route('/get', methods=['GET']) # noqa: F821
@login_required
def get():
dialog_id = request.args["dialog_id"]
try:
e, dia = DialogService.get_by_id(dialog_id)
if not e:
return get_data_error_result(message="Dialog not found!")
dia = dia.to_dict()
dia["kb_ids"], dia["kb_names"] = get_kb_names(dia["kb_ids"])
return get_json_result(data=dia)
except Exception as e:
return server_error_response(e)
def get_kb_names(kb_ids):
ids, nms = [], []
for kid in kb_ids:
e, kb = KnowledgebaseService.get_by_id(kid)
if not e or kb.status != StatusEnum.VALID.value:
continue
ids.append(kid)
nms.append(kb.name)
return ids, nms
@manager.route('/list', methods=['GET']) # noqa: F821
@login_required
def list_dialogs():
try:
conversations = DialogService.query(
tenant_id=current_user.id,
status=StatusEnum.VALID.value,
reverse=True,
order_by=DialogService.model.create_time)
conversations = [d.to_dict() for d in conversations]
for conversation in conversations:
conversation["kb_ids"], conversation["kb_names"] = get_kb_names(conversation["kb_ids"])
return get_json_result(data=conversations)
except Exception as e:
return server_error_response(e)
@manager.route('/next', methods=['POST']) # noqa: F821
@login_required
async def list_dialogs_next():
args = request.args
keywords = args.get("keywords", "")
page_number = int(args.get("page", 0))
items_per_page = int(args.get("page_size", 0))
parser_id = args.get("parser_id")
orderby = args.get("orderby", "create_time")
if args.get("desc", "true").lower() == "false":
desc = False
else:
desc = True
req = await get_request_json()
owner_ids = req.get("owner_ids", [])
try:
if not owner_ids:
# tenants = TenantService.get_joined_tenants_by_user_id(current_user.id)
# tenants = [tenant["tenant_id"] for tenant in tenants]
tenants = [] # keep it here
dialogs, total = DialogService.get_by_tenant_ids(
tenants, current_user.id, page_number,
items_per_page, orderby, desc, keywords, parser_id)
else:
tenants = owner_ids
dialogs, total = DialogService.get_by_tenant_ids(
tenants, current_user.id, 0,
0, orderby, desc, keywords, parser_id)
dialogs = [dialog for dialog in dialogs if dialog["tenant_id"] in tenants]
total = len(dialogs)
if page_number and items_per_page:
dialogs = dialogs[(page_number-1)*items_per_page:page_number*items_per_page]
return get_json_result(data={"dialogs": dialogs, "total": total})
except Exception as e:
return server_error_response(e)
@manager.route('/rm', methods=['POST']) # noqa: F821
@login_required
@validate_request("dialog_ids")
async def rm():
req = await get_request_json()
dialog_list=[]
tenants = UserTenantService.query(user_id=current_user.id)
try:
for id in req["dialog_ids"]:
for tenant in tenants:
if DialogService.query(tenant_id=tenant.tenant_id, id=id):
break
else:
return get_json_result(
data=False, message='Only owner of dialog authorized for this operation.',
code=RetCode.OPERATING_ERROR)
dialog_list.append({"id": id,"status":StatusEnum.INVALID.value})
DialogService.update_many_by_id(dialog_list)
return get_json_result(data=True)
except Exception as e:
return server_error_response(e)

568
api/apps/restful_apis/chat_api.py
View File

@@ -1,568 +0,0 @@
#
# Copyright 2026 The InfiniFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
from copy import deepcopy
from quart import request
from api.apps import current_user, login_required
from api.db.services.dialog_service import DialogService
from api.db.services.knowledgebase_service import KnowledgebaseService
from api.db.services.tenant_llm_service import TenantLLMService
from api.db.services.user_service import TenantService, UserTenantService
from api.utils.api_utils import (
check_duplicate_ids,
get_data_error_result,
get_json_result,
get_request_json,
server_error_response,
)
from api.utils.tenant_utils import ensure_tenant_model_id_for_params
from common.constants import RetCode, StatusEnum
from common.misc_utils import get_uuid
_DEFAULT_PROMPT_CONFIG = {
"system": (
'You are an intelligent assistant. Please summarize the content of the dataset to answer the question. '
'Please list the data in the dataset and answer in detail. When all dataset content is irrelevant to the '
'question, your answer must include the sentence "The answer you are looking for is not found in the dataset!" '
"Answers need to consider chat history.\n"
" Here is the knowledge base:\n"
" {knowledge}\n"
" The above is the knowledge base."
),
"prologue": "Hi! I'm your assistant. What can I do for you?",
"parameters": [{"key": "knowledge", "optional": False}],
"empty_response": "Sorry! No relevant content was found in the knowledge base!",
"quote": True,
"tts": False,
"refine_multiturn": True,
}
_DEFAULT_RERANK_MODELS = {"BAAI/bge-reranker-v2-m3", "maidalun1020/bce-reranker-base_v1"}
_READONLY_FIELDS = {"id", "tenant_id", "created_by", "create_time", "create_date", "update_time", "update_date"}
_PERSISTED_FIELDS = set(DialogService.model._meta.fields)
def _build_chat_response(chat):
data = chat.to_dict() if hasattr(chat, "to_dict") else dict(chat)
kb_ids, kb_names = _resolve_kb_names(data.get("kb_ids", []))
data["dataset_ids"] = kb_ids
data.pop("kb_ids", None)
data["kb_names"] = kb_names
return data
def _resolve_kb_names(kb_ids):
ids, names = [], []
for kb_id in kb_ids or []:
ok, kb = KnowledgebaseService.get_by_id(kb_id)
if not ok or kb.status != StatusEnum.VALID.value:
continue
ids.append(kb_id)
names.append(kb.name)
return ids, names
def _has_knowledge_placeholder(prompt_config):
return "{knowledge}" in (prompt_config or {}).get("system", "")
def _validate_name(name, *, required=True):
if name is None:
if required:
return None, "`name` is required."
return None, None
if not isinstance(name, str):
return None, "Chat name must be a string."
name = name.strip()
if not name:
return None, "Chat name can't be empty." if required else "`name` cannot be empty."
if len(name.encode("utf-8")) > 255:
return None, f"Chat name length is {len(name.encode('utf-8'))} which is larger than 255."
return name, None
def _ensure_owned_chat(chat_id):
return DialogService.query(
tenant_id=current_user.id, id=chat_id, status=StatusEnum.VALID.value
)
def _validate_llm_id(llm_id, tenant_id, llm_setting=None):
if not llm_id:
return None
llm_name, llm_factory = TenantLLMService.split_model_name_and_factory(llm_id)
model_type = (llm_setting or {}).get("model_type")
if model_type not in {"chat", "image2text"}:
model_type = "chat"
if not TenantLLMService.query(
tenant_id=tenant_id,
llm_name=llm_name,
llm_factory=llm_factory,
model_type=model_type,
):
return f"`llm_id` {llm_id} doesn't exist"
return None
def _validate_rerank_id(rerank_id, tenant_id):
if not rerank_id:
return None
llm_name, llm_factory = TenantLLMService.split_model_name_and_factory(rerank_id)
if llm_name in _DEFAULT_RERANK_MODELS:
return None
if TenantLLMService.query(
tenant_id=tenant_id,
llm_name=llm_name,
llm_factory=llm_factory,
model_type="rerank",
):
return None
return f"`rerank_id` {rerank_id} doesn't exist"
def _validate_prompt_config(prompt_config):
for parameter in prompt_config.get("parameters", []):
if parameter.get("optional"):
continue
if prompt_config.get("system", "").find("{%s}" % parameter["key"]) < 0:
return f"Parameter '{parameter['key']}' is not used"
return None
def _validate_dataset_ids(dataset_ids, tenant_id):
if dataset_ids is None:
return []
if not isinstance(dataset_ids, list):
return f"`dataset_ids` should be a list."
normalized_ids = [dataset_id for dataset_id in dataset_ids if dataset_id]
kbs = []
for dataset_id in normalized_ids:
if not KnowledgebaseService.accessible(kb_id=dataset_id, user_id=tenant_id):
return f"You don't own the dataset {dataset_id}"
matches = KnowledgebaseService.query(id=dataset_id)
if not matches:
return f"You don't own the dataset {dataset_id}"
kb = matches[0]
if kb.chunk_num == 0:
return f"The dataset {dataset_id} doesn't own parsed file"
kbs.append(kb)
embd_ids = [TenantLLMService.split_model_name_and_factory(kb.embd_id)[0] for kb in kbs]
if len(set(embd_ids)) > 1:
return f'Datasets use different embedding models: {[kb.embd_id for kb in kbs]}'
return normalized_ids
def _apply_prompt_defaults(req):
prompt_config = req.setdefault("prompt_config", {})
for key, value in _DEFAULT_PROMPT_CONFIG.items():
temp = prompt_config.get(key)
if (key == "system" and not temp) or key not in prompt_config:
prompt_config[key] = deepcopy(value)
if req.get("kb_ids") and not prompt_config.get("parameters") and "{knowledge}" in prompt_config.get("system", ""):
prompt_config["parameters"] = [{"key": "knowledge", "optional": False}]
@manager.route("/chats", methods=["POST"]) # noqa: F821
@login_required
async def create():
try:
req = await get_request_json()
ok, tenant = TenantService.get_by_id(current_user.id)
if not ok:
return get_data_error_result(message="Tenant not found!")
# Validate tenant_id should not be provided
if req.get("tenant_id"):
return get_data_error_result(message="`tenant_id` must not be provided.")
# Validate name
name, err = _validate_name(req.get("name"), required=True)
if err:
return get_data_error_result(message=err)
req["name"] = name
if "dataset_ids" in req:
kb_ids = _validate_dataset_ids(req.get("dataset_ids"), current_user.id)
if isinstance(kb_ids, str):
return get_data_error_result(message=kb_ids)
req["kb_ids"] = kb_ids
req.pop("dataset_ids", None)
if "llm_id" in req:
err = _validate_llm_id(req.get("llm_id"), current_user.id, req.get("llm_setting"))
if err:
return get_data_error_result(message=err)
if "rerank_id" in req:
err = _validate_rerank_id(req.get("rerank_id"), current_user.id)
if err:
return get_data_error_result(message=err)
if "prompt_config" in req:
if not isinstance(req["prompt_config"], dict):
return get_data_error_result(message="`prompt_config` should be an object.")
err = _validate_prompt_config(req["prompt_config"])
if err:
return get_data_error_result(message=err)
req.setdefault("kb_ids", [])
req.setdefault("llm_id", tenant.llm_id)
if req["llm_id"] is None:
req["llm_id"] = tenant.llm_id
req.setdefault("llm_setting", {})
req.setdefault("description", "A helpful Assistant")
req.setdefault("top_n", 6)
req.setdefault("top_k", 1024)
req.setdefault("rerank_id", "")
req.setdefault("similarity_threshold", 0.1)
req.setdefault("vector_similarity_weight", 0.3)
req.setdefault("icon", "")
_apply_prompt_defaults(req)
err = _validate_prompt_config(req["prompt_config"])
if err:
return get_data_error_result(message=err)
req = ensure_tenant_model_id_for_params(current_user.id, req)
req = {field: value for field, value in req.items() if field in _PERSISTED_FIELDS}
for field in _READONLY_FIELDS:
req.pop(field, None)
if DialogService.query(
name=req["name"],
tenant_id=current_user.id,
status=StatusEnum.VALID.value,
):
return get_data_error_result(message="Duplicated chat name in creating chat.")
req["id"] = get_uuid()
req["tenant_id"] = current_user.id
if not DialogService.save(**req):
return get_data_error_result(message="Failed to create chat.")
ok, chat = DialogService.get_by_id(req["id"])
if not ok:
return get_data_error_result(message="Failed to retrieve created chat.")
return get_json_result(data=_build_chat_response(chat))
except Exception as ex:
return server_error_response(ex)
@manager.route("/chats", methods=["GET"]) # noqa: F821
@login_required
def list_chats():
chat_id = request.args.get("id")
name = request.args.get("name")
keywords = request.args.get("keywords", "")
orderby = request.args.get("orderby", "create_time")
desc = request.args.get("desc", "true").lower() != "false"
owner_ids = request.args.getlist("owner_ids")
exact_filters = {"id": chat_id, "name": name}
if chat_id or name:
keywords = ""
try:
page_number = int(request.args.get("page", 1))
items_per_page = int(request.args.get("page_size", 0))
if owner_ids:
chats, total = DialogService.get_by_tenant_ids(
owner_ids, current_user.id, 0, 0, orderby, desc, keywords, **exact_filters
)
chats = [chat for chat in chats if chat["tenant_id"] in owner_ids]
total = len(chats)
if page_number and items_per_page:
start = (page_number - 1) * items_per_page
chats = chats[start : start + items_per_page]
else:
chats, total = DialogService.get_by_tenant_ids(
[], current_user.id, page_number, items_per_page, orderby, desc, keywords, **exact_filters
)
return get_json_result(
data={"chats": [_build_chat_response(chat) for chat in chats], "total": total}
)
except Exception as ex:
return server_error_response(ex)
@manager.route("/chats/<chat_id>", methods=["GET"]) # noqa: F821
@login_required
def get_chat(chat_id):
try:
tenants = UserTenantService.query(user_id=current_user.id)
for tenant in tenants:
if DialogService.query(
tenant_id=tenant.tenant_id, id=chat_id, status=StatusEnum.VALID.value
):
break
else:
return get_json_result(
data=False,
message="No authorization.",
code=RetCode.AUTHENTICATION_ERROR,
)
ok, chat = DialogService.get_by_id(chat_id)
if not ok:
return get_data_error_result(message="Chat not found!")
return get_json_result(data=_build_chat_response(chat))
except Exception as ex:
return server_error_response(ex)
@manager.route("/chats/<chat_id>", methods=["PUT"]) # noqa: F821
@login_required
async def update_chat(chat_id):
if not _ensure_owned_chat(chat_id):
return get_json_result(
data=False, message="No authorization.", code=RetCode.AUTHENTICATION_ERROR
)
try:
req = await get_request_json()
ok, tenant = TenantService.get_by_id(current_user.id)
if not ok:
return get_data_error_result(message="Tenant not found!")
ok, current_chat = DialogService.get_by_id(chat_id)
if not ok:
return get_data_error_result(message="Chat not found!")
current_chat = current_chat.to_dict()
if req.get("tenant_id"):
return get_data_error_result(message="`tenant_id` must not be provided.")
if "name" in req:
name, err = _validate_name(req.get("name"), required=True)
if err:
return get_data_error_result(message=err)
req["name"] = name
if "dataset_ids" in req:
kb_ids = _validate_dataset_ids(req.get("dataset_ids"), current_user.id)
if isinstance(kb_ids, str):
return get_data_error_result(message=kb_ids)
req["kb_ids"] = kb_ids
req.pop("dataset_ids", None)
if "llm_id" in req:
err = _validate_llm_id(req.get("llm_id"), current_user.id, req.get("llm_setting"))
if err:
return get_data_error_result(message=err)
if "rerank_id" in req:
err = _validate_rerank_id(req.get("rerank_id"), current_user.id)
if err:
return get_data_error_result(message=err)
if "prompt_config" in req:
if not isinstance(req["prompt_config"], dict):
return get_data_error_result(message="`prompt_config` should be an object.")
err = _validate_prompt_config(req["prompt_config"])
if err:
return get_data_error_result(message=err)
prompt_config = req.get("prompt_config", {})
if not prompt_config:
prompt_config = current_chat.get("prompt_config", {})
kb_ids = req.get("kb_ids", current_chat.get("kb_ids", []))
if not kb_ids and not prompt_config.get("tavily_api_key") and _has_knowledge_placeholder(prompt_config):
return get_data_error_result(message="Please remove `{knowledge}` in system prompt since no dataset / Tavily used here.")
req = ensure_tenant_model_id_for_params(current_user.id, req)
req = {field: value for field, value in req.items() if field in _PERSISTED_FIELDS}
for field in _READONLY_FIELDS:
req.pop(field, None)
if (
"name" in req
and req["name"].lower() != current_chat["name"].lower()
and DialogService.query(
name=req["name"],
tenant_id=current_user.id,
status=StatusEnum.VALID.value,
)
):
return get_data_error_result(message="Duplicated chat name.")
if not DialogService.update_by_id(chat_id, req):
return get_data_error_result(message="Chat not found!")
ok, chat = DialogService.get_by_id(chat_id)
if not ok:
return get_data_error_result(message="Failed to retrieve updated chat.")
return get_json_result(data=_build_chat_response(chat))
except Exception as ex:
return server_error_response(ex)
@manager.route("/chats/<chat_id>", methods=["PATCH"]) # noqa: F821
@login_required
async def patch_chat(chat_id):
if not _ensure_owned_chat(chat_id):
return get_json_result(
data=False, message="No authorization.", code=RetCode.AUTHENTICATION_ERROR
)
try:
req = await get_request_json()
ok, tenant = TenantService.get_by_id(current_user.id)
if not ok:
return get_data_error_result(message="Tenant not found!")
ok, current_chat = DialogService.get_by_id(chat_id)
if not ok:
return get_data_error_result(message="Chat not found!")
current_chat = current_chat.to_dict()
if req.get("tenant_id"):
return get_data_error_result(message="`tenant_id` must not be provided.")
if "name" in req:
name, err = _validate_name(req.get("name"), required=False)
if err:
return get_data_error_result(message=err)
if name is not None:
req["name"] = name
if "dataset_ids" in req:
kb_ids = _validate_dataset_ids(req.get("dataset_ids"), current_user.id)
if isinstance(kb_ids, str):
return get_data_error_result(message=kb_ids)
req["kb_ids"] = kb_ids
req.pop("dataset_ids", None)
if "llm_id" in req:
err = _validate_llm_id(req.get("llm_id"), current_user.id, req.get("llm_setting"))
if err:
return get_data_error_result(message=err)
if "rerank_id" in req:
err = _validate_rerank_id(req.get("rerank_id"), current_user.id)
if err:
return get_data_error_result(message=err)
if "prompt_config" in req:
if not isinstance(req["prompt_config"], dict):
return get_data_error_result(message="`prompt_config` should be an object.")
prompt_config = deepcopy(current_chat.get("prompt_config", {}))
prompt_config.update(req["prompt_config"])
req["prompt_config"] = prompt_config
err = _validate_prompt_config(prompt_config)
if err:
return get_data_error_result(message=err)
if "llm_setting" in req:
llm_setting = deepcopy(current_chat.get("llm_setting", {}))
llm_setting.update(req["llm_setting"])
req["llm_setting"] = llm_setting
if "prompt_config" in req or "kb_ids" in req:
prompt_config = req.get("prompt_config", current_chat.get("prompt_config", {}))
kb_ids = req.get("kb_ids", current_chat.get("kb_ids", []))
if not kb_ids and not prompt_config.get("tavily_api_key") and _has_knowledge_placeholder(prompt_config):
return get_data_error_result(message="Please remove `{knowledge}` in system prompt since no dataset / Tavily used here.")
req = ensure_tenant_model_id_for_params(current_user.id, req)
req = {field: value for field, value in req.items() if field in _PERSISTED_FIELDS}
for field in _READONLY_FIELDS:
req.pop(field, None)
if (
"name" in req
and req["name"].lower() != current_chat["name"].lower()
and DialogService.query(
name=req["name"],
tenant_id=current_user.id,
status=StatusEnum.VALID.value,
)
):
return get_data_error_result(message="Duplicated chat name.")
if not DialogService.update_by_id(chat_id, req):
return get_data_error_result(message="Failed to update chat.")
ok, chat = DialogService.get_by_id(chat_id)
if not ok:
return get_data_error_result(message="Failed to retrieve updated chat.")
return get_json_result(data=_build_chat_response(chat))
except Exception as ex:
return server_error_response(ex)
@manager.route("/chats/<chat_id>", methods=["DELETE"]) # noqa: F821
@login_required
def delete_chat(chat_id):
if not _ensure_owned_chat(chat_id):
return get_json_result(
data=False, message="No authorization.", code=RetCode.AUTHENTICATION_ERROR
)
try:
if not DialogService.update_by_id(chat_id, {"status": StatusEnum.INVALID.value}):
return get_data_error_result(message=f"Failed to delete chat {chat_id}")
return get_json_result(data=True)
except Exception as ex:
return server_error_response(ex)
@manager.route("/chats", methods=["DELETE"]) # noqa: F821
@login_required
async def bulk_delete_chats():
req = await get_request_json()
if not req:
return get_json_result(data={})
ids = req.get("ids")
if not ids:
if req.get("delete_all") is True:
ids = [
chat.id
for chat in DialogService.query(
tenant_id=current_user.id, status=StatusEnum.VALID.value
)
]
if not ids:
return get_json_result(data={})
else:
return get_json_result(data={})
errors = []
success_count = 0
unique_ids, duplicate_messages = check_duplicate_ids(ids, "chat")
for chat_id in unique_ids:
if not _ensure_owned_chat(chat_id):
errors.append(f"Chat({chat_id}) not found.")
continue
success_count += DialogService.update_by_id(chat_id, {"status": StatusEnum.INVALID.value})
all_errors = errors + duplicate_messages
if all_errors:
if success_count > 0:
return get_json_result(
data={"success_count": success_count, "errors": all_errors},
message=f"Partially deleted {success_count} chats with {len(all_errors)} errors",
)
return get_data_error_result(message="; ".join(all_errors))
return get_json_result(data={"success_count": success_count})

329
api/apps/sdk/chat.py Normal file
View File

@@ -0,0 +1,329 @@
#
# Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import logging
from quart import request
from api.db.services.dialog_service import DialogService
from api.db.services.knowledgebase_service import KnowledgebaseService
from api.db.services.tenant_llm_service import TenantLLMService
from api.db.services.user_service import TenantService
from common.misc_utils import get_uuid
from common.constants import RetCode, StatusEnum
from api.utils.api_utils import check_duplicate_ids, get_error_data_result, get_result, token_required, get_request_json
@manager.route("/chats", methods=["POST"]) # noqa: F821
@token_required
async def create(tenant_id):
req = await get_request_json()
ids = [i for i in req.get("dataset_ids", []) if i]
for kb_id in ids:
kbs = KnowledgebaseService.accessible(kb_id=kb_id, user_id=tenant_id)
if not kbs:
return get_error_data_result(f"You don't own the dataset {kb_id}")
kbs = KnowledgebaseService.query(id=kb_id)
kb = kbs[0]
if kb.chunk_num == 0:
return get_error_data_result(f"The dataset {kb_id} doesn't own parsed file")
kbs = KnowledgebaseService.get_by_ids(ids) if ids else []
embd_ids = [TenantLLMService.split_model_name_and_factory(kb.embd_id)[0] for kb in kbs] # remove vendor suffix for comparison
embd_count = list(set(embd_ids))
if len(embd_count) > 1:
return get_result(message='Datasets use different embedding models."', code=RetCode.AUTHENTICATION_ERROR)
req["kb_ids"] = ids
# llm
llm = req.get("llm")
if llm:
if "model_name" in llm:
req["llm_id"] = llm.pop("model_name")
if req.get("llm_id") is not None:
llm_name, llm_factory = TenantLLMService.split_model_name_and_factory(req["llm_id"])
model_type = llm.get("model_type")
model_type = model_type if model_type in ["chat", "image2text"] else "chat"
if not TenantLLMService.query(tenant_id=tenant_id, llm_name=llm_name, llm_factory=llm_factory, model_type=model_type):
return get_error_data_result(f"`model_name` {req.get('llm_id')} doesn't exist")
req["llm_setting"] = req.pop("llm")
e, tenant = TenantService.get_by_id(tenant_id)
if not e:
return get_error_data_result(message="Tenant not found!")
# prompt
prompt = req.get("prompt")
key_mapping = {"parameters": "variables", "prologue": "opener", "quote": "show_quote", "system": "prompt", "rerank_id": "rerank_model", "vector_similarity_weight": "keywords_similarity_weight"}
key_list = ["similarity_threshold", "vector_similarity_weight", "top_n", "rerank_id", "top_k"]
if prompt:
for new_key, old_key in key_mapping.items():
if old_key in prompt:
prompt[new_key] = prompt.pop(old_key)
for key in key_list:
if key in prompt:
req[key] = prompt.pop(key)
req["prompt_config"] = req.pop("prompt")
# init
req["id"] = get_uuid()
req["description"] = req.get("description", "A helpful Assistant")
req["icon"] = req.get("avatar", "")
req["top_n"] = req.get("top_n", 6)
req["top_k"] = req.get("top_k", 1024)
req["rerank_id"] = req.get("rerank_id", "")
if req.get("rerank_id"):
value_rerank_model = ["BAAI/bge-reranker-v2-m3", "maidalun1020/bce-reranker-base_v1"]
if req["rerank_id"] not in value_rerank_model and not TenantLLMService.query(tenant_id=tenant_id, llm_name=req.get("rerank_id"), model_type="rerank"):
return get_error_data_result(f"`rerank_model` {req.get('rerank_id')} doesn't exist")
if not req.get("llm_id"):
req["llm_id"] = tenant.llm_id
if not req.get("name"):
return get_error_data_result(message="`name` is required.")
if DialogService.query(name=req["name"], tenant_id=tenant_id, status=StatusEnum.VALID.value):
return get_error_data_result(message="Duplicated chat name in creating chat.")
# tenant_id
if req.get("tenant_id"):
return get_error_data_result(message="`tenant_id` must not be provided.")
req["tenant_id"] = tenant_id
# prompt more parameter
default_prompt = {
"system": """You are an intelligent assistant. Please summarize the content of the dataset to answer the question. Please list the data in the dataset and answer in detail. When all dataset content is irrelevant to the question, your answer must include the sentence "The answer you are looking for is not found in the dataset!" Answers need to consider chat history.
Here is the knowledge base:
{knowledge}
The above is the knowledge base.""",
"prologue": "Hi! I'm your assistant. What can I do for you?",
"parameters": [{"key": "knowledge", "optional": False}],
"empty_response": "Sorry! No relevant content was found in the knowledge base!",
"quote": True,
"tts": False,
"refine_multiturn": True,
}
key_list_2 = ["system", "prologue", "parameters", "empty_response", "quote", "tts", "refine_multiturn"]
if "prompt_config" not in req:
req["prompt_config"] = {}
for key in key_list_2:
temp = req["prompt_config"].get(key)
if (not temp and key == "system") or (key not in req["prompt_config"]):
req["prompt_config"][key] = default_prompt[key]
for p in req["prompt_config"]["parameters"]:
if p["optional"]:
continue
if req["prompt_config"]["system"].find("{%s}" % p["key"]) < 0:
return get_error_data_result(message="Parameter '{}' is not used".format(p["key"]))
# save
if not DialogService.save(**req):
return get_error_data_result(message="Fail to new a chat!")
# response
e, res = DialogService.get_by_id(req["id"])
if not e:
return get_error_data_result(message="Fail to new a chat!")
res = res.to_json()
renamed_dict = {}
for key, value in res["prompt_config"].items():
new_key = key_mapping.get(key, key)
renamed_dict[new_key] = value
res["prompt"] = renamed_dict
del res["prompt_config"]
new_dict = {"similarity_threshold": res["similarity_threshold"], "keywords_similarity_weight": 1 - res["vector_similarity_weight"], "top_n": res["top_n"], "rerank_model": res["rerank_id"]}
res["prompt"].update(new_dict)
for key in key_list:
del res[key]
res["llm"] = res.pop("llm_setting")
res["llm"]["model_name"] = res.pop("llm_id")
del res["kb_ids"]
res["dataset_ids"] = req.get("dataset_ids", [])
res["avatar"] = res.pop("icon")
return get_result(data=res)
@manager.route("/chats/<chat_id>", methods=["PUT"]) # noqa: F821
@token_required
async def update(tenant_id, chat_id):
if not DialogService.query(tenant_id=tenant_id, id=chat_id, status=StatusEnum.VALID.value):
return get_error_data_result(message="You do not own the chat")
req = await get_request_json()
ids = req.get("dataset_ids", [])
if "show_quotation" in req:
req["do_refer"] = req.pop("show_quotation")
if ids:
for kb_id in ids:
kbs = KnowledgebaseService.accessible(kb_id=kb_id, user_id=tenant_id)
if not kbs:
return get_error_data_result(f"You don't own the dataset {kb_id}")
kbs = KnowledgebaseService.query(id=kb_id)
kb = kbs[0]
if kb.chunk_num == 0:
return get_error_data_result(f"The dataset {kb_id} doesn't own parsed file")
kbs = KnowledgebaseService.get_by_ids(ids)
embd_ids = [TenantLLMService.split_model_name_and_factory(kb.embd_id)[0] for kb in kbs] # remove vendor suffix for comparison
embd_count = list(set(embd_ids))
if len(embd_count) > 1:
return get_result(message='Datasets use different embedding models."', code=RetCode.AUTHENTICATION_ERROR)
req["kb_ids"] = ids
else:
req["kb_ids"] = []
llm = req.get("llm")
if llm:
if "model_name" in llm:
req["llm_id"] = llm.pop("model_name")
if req.get("llm_id") is not None:
llm_name, llm_factory = TenantLLMService.split_model_name_and_factory(req["llm_id"])
model_type = llm.get("model_type")
model_type = model_type if model_type in ["chat", "image2text"] else "chat"
if not TenantLLMService.query(tenant_id=tenant_id, llm_name=llm_name, llm_factory=llm_factory, model_type=model_type):
return get_error_data_result(f"`model_name` {req.get('llm_id')} doesn't exist")
req["llm_setting"] = req.pop("llm")
e, tenant = TenantService.get_by_id(tenant_id)
if not e:
return get_error_data_result(message="Tenant not found!")
# prompt
prompt = req.get("prompt")
key_mapping = {"parameters": "variables", "prologue": "opener", "quote": "show_quote", "system": "prompt", "rerank_id": "rerank_model", "vector_similarity_weight": "keywords_similarity_weight"}
key_list = ["similarity_threshold", "vector_similarity_weight", "top_n", "rerank_id", "top_k"]
if prompt:
for new_key, old_key in key_mapping.items():
if old_key in prompt:
prompt[new_key] = prompt.pop(old_key)
for key in key_list:
if key in prompt:
req[key] = prompt.pop(key)
req["prompt_config"] = req.pop("prompt")
e, res = DialogService.get_by_id(chat_id)
res = res.to_json()
if req.get("rerank_id"):
value_rerank_model = ["BAAI/bge-reranker-v2-m3", "maidalun1020/bce-reranker-base_v1"]
if req["rerank_id"] not in value_rerank_model and not TenantLLMService.query(tenant_id=tenant_id, llm_name=req.get("rerank_id"), model_type="rerank"):
return get_error_data_result(f"`rerank_model` {req.get('rerank_id')} doesn't exist")
if "name" in req:
if not req.get("name"):
return get_error_data_result(message="`name` cannot be empty.")
if req["name"].lower() != res["name"].lower() and len(DialogService.query(name=req["name"], tenant_id=tenant_id, status=StatusEnum.VALID.value)) > 0:
return get_error_data_result(message="Duplicated chat name in updating chat.")
if "prompt_config" in req:
res["prompt_config"].update(req["prompt_config"])
for p in res["prompt_config"]["parameters"]:
if p["optional"]:
continue
if res["prompt_config"]["system"].find("{%s}" % p["key"]) < 0:
return get_error_data_result(message="Parameter '{}' is not used".format(p["key"]))
if "llm_setting" in req:
res["llm_setting"].update(req["llm_setting"])
req["prompt_config"] = res["prompt_config"]
req["llm_setting"] = res["llm_setting"]
# avatar
if "avatar" in req:
req["icon"] = req.pop("avatar")
if "dataset_ids" in req:
req.pop("dataset_ids")
if not DialogService.update_by_id(chat_id, req):
return get_error_data_result(message="Chat not found!")
return get_result()
@manager.route("/chats", methods=["DELETE"]) # noqa: F821
@token_required
async def delete_chats(tenant_id):
errors = []
success_count = 0
req = await get_request_json()
if not req:
return get_result()
ids = req.get("ids")
if not ids:
if req.get("delete_all") is True:
ids = [d.id for d in DialogService.query(tenant_id=tenant_id, status=StatusEnum.VALID.value)]
if not ids:
return get_result()
else:
return get_result()
id_list = ids
unique_id_list, duplicate_messages = check_duplicate_ids(id_list, "assistant")
for id in unique_id_list:
if not DialogService.query(tenant_id=tenant_id, id=id, status=StatusEnum.VALID.value):
errors.append(f"Assistant({id}) not found.")
continue
temp_dict = {"status": StatusEnum.INVALID.value}
success_count += DialogService.update_by_id(id, temp_dict)
if errors:
if success_count > 0:
return get_result(data={"success_count": success_count, "errors": errors}, message=f"Partially deleted {success_count} chats with {len(errors)} errors")
else:
return get_error_data_result(message="; ".join(errors))
if duplicate_messages:
if success_count > 0:
return get_result(message=f"Partially deleted {success_count} chats with {len(duplicate_messages)} errors", data={"success_count": success_count, "errors": duplicate_messages})
else:
return get_error_data_result(message=";".join(duplicate_messages))
return get_result()
@manager.route("/chats", methods=["GET"]) # noqa: F821
@token_required
def list_chat(tenant_id):
id = request.args.get("id")
name = request.args.get("name")
if id or name:
chat = DialogService.query(id=id, name=name, status=StatusEnum.VALID.value, tenant_id=tenant_id)
if not chat:
return get_error_data_result(message="The chat doesn't exist")
page_number = int(request.args.get("page", 1))
items_per_page = int(request.args.get("page_size", 30))
orderby = request.args.get("orderby", "create_time")
if request.args.get("desc") == "False" or request.args.get("desc") == "false":
desc = False
else:
desc = True
chats = DialogService.get_list(tenant_id, page_number, items_per_page, orderby, desc, id, name)
if not chats:
return get_result(data=[])
list_assistants = []
key_mapping = {
"parameters": "variables",
"prologue": "opener",
"quote": "show_quote",
"system": "prompt",
"rerank_id": "rerank_model",
"vector_similarity_weight": "keywords_similarity_weight",
"do_refer": "show_quotation",
}
key_list = ["similarity_threshold", "vector_similarity_weight", "top_n", "rerank_id"]
for res in chats:
renamed_dict = {}
for key, value in res["prompt_config"].items():
new_key = key_mapping.get(key, key)
renamed_dict[new_key] = value
res["prompt"] = renamed_dict
del res["prompt_config"]
new_dict = {"similarity_threshold": res["similarity_threshold"], "keywords_similarity_weight": 1 - res["vector_similarity_weight"], "top_n": res["top_n"], "rerank_model": res["rerank_id"]}
res["prompt"].update(new_dict)
for key in key_list:
del res[key]
res["llm"] = res.pop("llm_setting")
res["llm"]["model_name"] = res.pop("llm_id")
kb_list = []
for kb_id in res["kb_ids"]:
kb = KnowledgebaseService.query(id=kb_id)
if not kb:
logging.warning(f"The kb {kb_id} does not exist.")
continue
kb_list.append(kb[0].to_json())
del res["kb_ids"]
res["datasets"] = kb_list
res["avatar"] = res.pop("icon")
list_assistants.append(res)
return get_result(data=list_assistants)

44
api/db/services/dialog_service.py
View File

@@ -105,18 +105,7 @@ class DialogService(CommonService):
@classmethod
@DB.connection_context()
def get_by_tenant_ids(
cls,
joined_tenant_ids,
user_id,
page_number,
items_per_page,
orderby,
desc,
keywords,
id=None,
name=None,
):
def get_by_tenant_ids(cls, joined_tenant_ids, user_id, page_number, items_per_page, orderby, desc, keywords, parser_id=None):
from api.db.db_models import User
fields = [
@@ -143,20 +132,25 @@ class DialogService(CommonService):
cls.model.update_time,
cls.model.create_time,
]
dialogs = (
cls.model.select(*fields)
.join(User, on=(cls.model.tenant_id == User.id))
.where(
(cls.model.tenant_id.in_(joined_tenant_ids) | (cls.model.tenant_id == user_id))
& (cls.model.status == StatusEnum.VALID.value),
)
)
if id:
dialogs = dialogs.where(cls.model.id == id)
if name:
dialogs = dialogs.where(cls.model.name == name)
if keywords:
dialogs = dialogs.where(fn.LOWER(cls.model.name).contains(keywords.lower()))
dialogs = (
cls.model.select(*fields)
.join(User, on=(cls.model.tenant_id == User.id))
.where(
(cls.model.tenant_id.in_(joined_tenant_ids) | (cls.model.tenant_id == user_id)) & (cls.model.status == StatusEnum.VALID.value),
(fn.LOWER(cls.model.name).contains(keywords.lower())),
)
)
else:
dialogs = (
cls.model.select(*fields)
.join(User, on=(cls.model.tenant_id == User.id))
.where(
(cls.model.tenant_id.in_(joined_tenant_ids) | (cls.model.tenant_id == user_id)) & (cls.model.status == StatusEnum.VALID.value),
)
)
if parser_id:
dialogs = dialogs.where(cls.model.parser_id == parser_id)
if desc:
dialogs = dialogs.order_by(cls.model.getter_by(orderby).desc())
else: