mirror of
https://github.com/OpenSPG/KAG.git
synced 2025-06-27 03:20:08 +00:00
update examples
This commit is contained in:
锦呈
parent
59376537ae
commit
9d8592bd73
@ -1065,6 +1065,5 @@ if __name__ == "__main__":
|
||||
file_path = os.path.join(
|
||||
dir_path, "../../../../tests/unit/builder/data", "需求内容test.md"
|
||||
)
|
||||
file_path = "/Users/zhangxinhong.zxh/workspace/KAG/dep/KAG/kag/examples/AFAC2024/builder/data/BY04.md"
|
||||
chunks = reader.invoke(file_path, write_ckpt=False)
|
||||
print(chunks)
|
||||
|
||||
1
kag/examples/EastElectric/.gitignore
vendored
1
kag/examples/EastElectric/.gitignore
vendored
@ -1 +0,0 @@
|
||||
ckpt/
|
||||
@ -1,14 +0,0 @@
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
|
||||
"""
|
||||
Builder Dir.
|
||||
"""
|
||||
@ -1,15 +0,0 @@
|
||||
from kag.common.checkpointer import CheckPointer, CheckpointerManager
|
||||
|
||||
checkpointer: CheckPointer = CheckpointerManager.get_checkpointer(
|
||||
{
|
||||
"type": "diskcache",
|
||||
# "ckpt_dir": "ckpt/SchemaFreeExtractor",
|
||||
"ckpt_dir": "ckpt/OutlineExtractor",
|
||||
}
|
||||
)
|
||||
|
||||
if checkpointer.size() > 0:
|
||||
tmp_key = checkpointer.keys()[-1]
|
||||
print(checkpointer.read_from_ckpt(tmp_key))
|
||||
else:
|
||||
print("checkpoint is empty")
|
||||
@ -1,14 +0,0 @@
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
|
||||
"""
|
||||
Place the files to be used for building the index in this directory.
|
||||
"""
|
||||
File diff suppressed because it is too large
Load Diff
@ -1,50 +0,0 @@
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
import asyncio
|
||||
import logging
|
||||
import argparse
|
||||
import os
|
||||
|
||||
from kag.common.registry import import_modules_from_path
|
||||
|
||||
from kag.builder.runner import BuilderChainRunner
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
async def buildKB(file_path):
|
||||
from kag.common.conf import KAG_CONFIG
|
||||
|
||||
runner = BuilderChainRunner.from_config(
|
||||
KAG_CONFIG.all_config["kag_builder_pipeline"]
|
||||
)
|
||||
await runner.ainvoke(file_path)
|
||||
|
||||
logger.info(f"\n\nbuildKB successfully for {file_path}\n\n")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
import_modules_from_path(".")
|
||||
parser = argparse.ArgumentParser(description="args")
|
||||
parser.add_argument(
|
||||
"--corpus_file",
|
||||
type=str,
|
||||
help="test file name in /data",
|
||||
default="./data/附件1 10kV～110kV线路保护及辅助装置标准化设计规范.md",
|
||||
)
|
||||
|
||||
args = parser.parse_args()
|
||||
file_path = args.corpus_file
|
||||
|
||||
dir_path = os.path.dirname(__file__)
|
||||
file_path = os.path.join(dir_path, file_path)
|
||||
|
||||
asyncio.run(buildKB(file_path))
|
||||
@ -1,14 +0,0 @@
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
|
||||
"""
|
||||
Place the prompts to be used for building the index in this directory.
|
||||
"""
|
||||
@ -1,146 +0,0 @@
|
||||
#------------project configuration start----------------#
|
||||
openie_llm: &openie_llm
|
||||
type: maas
|
||||
base_url: https://dashscope.aliyuncs.com/compatible-mode/v1
|
||||
api_key: key
|
||||
model: qwen2.5-7b-instruct-1m
|
||||
enable_check: false
|
||||
max_tokens: 8192
|
||||
|
||||
chat_llm: &chat_llm
|
||||
type: maas
|
||||
base_url: https://dashscope.aliyuncs.com/compatible-mode/v1
|
||||
api_key: key
|
||||
model: qwen2.5-72b-instruct
|
||||
enable_check: false
|
||||
max_tokens: 8192
|
||||
|
||||
ner_llm:
|
||||
type: maas
|
||||
base_url: https://dashscope.aliyuncs.com/compatible-mode/v1
|
||||
api_key: key
|
||||
model: qwen2.5-72b-instruct
|
||||
enable_check: false
|
||||
max_tokens: 8192
|
||||
|
||||
vectorize_model: &vectorize_model
|
||||
api_key: key
|
||||
base_url: https://api.siliconflow.cn/v1
|
||||
model: BAAI/bge-m3
|
||||
type: openai
|
||||
vector_dimensions: 1024
|
||||
vectorizer: *vectorize_model
|
||||
|
||||
log:
|
||||
level: INFO
|
||||
|
||||
project:
|
||||
biz_scene: default
|
||||
host_addr: http://127.0.0.1:8887
|
||||
id: '5'
|
||||
language: en
|
||||
namespace: EastElectric
|
||||
#------------project configuration end----------------#
|
||||
|
||||
#------------kag-builder configuration start----------------#
|
||||
|
||||
e1: &e1
|
||||
type: schema_free_extractor # kag.builder.component.extractor.schema_free_extractor.SchemaFreeExtractor
|
||||
llm: *openie_llm
|
||||
ner_prompt:
|
||||
type: default_ner # kag.builder.prompt.default.ner.OpenIENERPrompt
|
||||
std_prompt:
|
||||
type: default_std # kag.builder.prompt.default.std.OpenIEEntitystandardizationdPrompt
|
||||
triple_prompt:
|
||||
type: default_triple # kag.builder.prompt.default.triple.OpenIETriplePrompt
|
||||
|
||||
e2: &e2
|
||||
type: outline_extractor # kag.builder.component.extractor.outline_extractor.OutlineExtractor
|
||||
|
||||
e3: &e3
|
||||
type: summary_extractor # kag.builder.component.extractor.summary_extractor.SummaryExtractor
|
||||
|
||||
e4: &e4
|
||||
type: chunk_extractor # kag.builder.component.extractor.chunk_extractor.ChunkExtractor
|
||||
|
||||
kag_builder_pipeline:
|
||||
chain:
|
||||
type: unstructured_builder_chain # kag.builder.default_chain.DefaultUnstructuredBuilderChain
|
||||
extractor:
|
||||
- *e2
|
||||
- *e3
|
||||
- *e4
|
||||
reader:
|
||||
type: md_reader # kag.builder.component.reader.dict_reader.DictReader
|
||||
cut_depth: 6
|
||||
post_processor:
|
||||
type: kag_post_processor # kag.builder.component.postprocessor.kag_postprocessor.KAGPostProcessor
|
||||
splitter:
|
||||
type: length_splitter # kag.builder.component.splitter.length_splitter.LengthSplitter
|
||||
split_length: 1000
|
||||
window_length: 0
|
||||
vectorizer:
|
||||
type: batch_vectorizer # kag.builder.component.vectorizer.batch_vectorizer.BatchVectorizer
|
||||
vectorize_model: *vectorize_model
|
||||
writer:
|
||||
type: kg_writer # kag.builder.component.writer.kg_writer.KGWriter
|
||||
num_threads_per_chain: 1
|
||||
num_chains: 16
|
||||
scanner:
|
||||
type: file_scanner # kag.builder.component.scanner.dataset_scanner.HotpotqaCorpusScanner
|
||||
#------------kag-builder configuration end----------------#
|
||||
|
||||
#------------kag-solver configuration start----------------#
|
||||
search_api: &search_api
|
||||
type: openspg_search_api #kag.solver.tools.search_api.impl.openspg_search_api.OpenSPGSearchAPI
|
||||
|
||||
graph_api: &graph_api
|
||||
type: openspg_graph_api #kag.solver.tools.graph_api.impl.openspg_graph_api.OpenSPGGraphApi
|
||||
|
||||
r1: &r1
|
||||
type: vector_chunk_retriever
|
||||
vectorize_model: *vectorize_model
|
||||
search_api: *search_api
|
||||
top_k: 10
|
||||
|
||||
r2: &r2
|
||||
type: text_chunk_retriever
|
||||
search_api: *search_api
|
||||
top_k: 10
|
||||
|
||||
r3: &r3
|
||||
type: outline_chunk_retriever # recall chunk through outline
|
||||
search_api: *search_api
|
||||
graph_api: *graph_api
|
||||
vectorize_model: *vectorize_model
|
||||
top_k: 5
|
||||
|
||||
r4: &r4
|
||||
type: summary_chunk_retriever # recall chunk through summary
|
||||
search_api: *search_api
|
||||
graph_api: *graph_api
|
||||
vectorize_model: *vectorize_model
|
||||
top_k: 5
|
||||
|
||||
retriever_executor: &retriever_executor
|
||||
type: kag_hybrid_retrieval_executor
|
||||
retrievers:
|
||||
- *r3
|
||||
merger:
|
||||
type: kag_merger
|
||||
|
||||
kag_solver_pipeline:
|
||||
type: kag_static_pipeline
|
||||
planner:
|
||||
type: kag_static_planner
|
||||
llm: *chat_llm
|
||||
plan_prompt:
|
||||
type: default_retriever_static_planning
|
||||
rewrite_prompt:
|
||||
type: default_query_rewrite
|
||||
executors:
|
||||
- *retriever_executor
|
||||
generator:
|
||||
type: llm_generator_with_thought
|
||||
llm_client: *chat_llm
|
||||
#------------kag-solver configuration end----------------#
|
||||
@ -1,20 +0,0 @@
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
|
||||
"""
|
||||
Place the DSL file for graph reasoning in this directory.
|
||||
For example:
|
||||
|
||||
```company.dsl
|
||||
MATCH (s:DEFAULT.Company)
|
||||
RETURN s.id, s.address
|
||||
```
|
||||
"""
|
||||
@ -1,153 +0,0 @@
|
||||
namespace EastElectric
|
||||
|
||||
Chunk(文本块): EntityType
|
||||
properties:
|
||||
content(内容): Text
|
||||
index: TextAndVector
|
||||
|
||||
Outline(标题大纲): EntityType
|
||||
properties:
|
||||
content(内容): Text
|
||||
index: TextAndVector
|
||||
relations:
|
||||
sourceChunk(关联): Chunk
|
||||
childOf(子标题): Outline
|
||||
|
||||
Diagram(标题大纲): EntityType
|
||||
properties:
|
||||
content(内容): Text
|
||||
index: TextAndVector
|
||||
relations:
|
||||
sourceChunk(关联): Chunk
|
||||
|
||||
Summary(文本摘要): EntityType
|
||||
properties:
|
||||
content(内容): Text
|
||||
index: TextAndVector
|
||||
relations:
|
||||
sourceChunk(关联): Chunk
|
||||
childOf(子摘要): Summary
|
||||
|
||||
ArtificialObject(人造物体): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Astronomy(天文学): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Building(建筑): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Creature(生物): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Concept(概念): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Date(日期): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
GeographicLocation(地理位置): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Keyword(关键词): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Medicine(药物): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
|
||||
NaturalScience(自然科学): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Organization(组织机构): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Person(人物): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Transport(运输): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Works(作品): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Others(其他): EntityType
|
||||
properties:
|
||||
desc(描述): Text
|
||||
index: TextAndVector
|
||||
semanticType(语义类型): Text
|
||||
index: Text
|
||||
|
||||
Event(事件): EventType
|
||||
properties:
|
||||
subject(主体): Person
|
||||
participants(参与者): Person
|
||||
constraint: MultiValue
|
||||
time(时间): Date
|
||||
location(地点): GeographicLocation
|
||||
abstract(摘要): Text
|
||||
index: TextAndVector
|
||||
type(事件类型): Text
|
||||
index: Text
|
||||
|
||||
|
||||
SemanticConcept(语义概念): EntityType
|
||||
properties:
|
||||
desc(内容): Text
|
||||
index: Text
|
||||
@ -1,18 +0,0 @@
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
|
||||
"""
|
||||
{{namespace}}.schema:
|
||||
The MarkLang file for the schema of this project.
|
||||
You can execute `kag schema commit` to commit your schema to SPG server.
|
||||
|
||||
|
||||
"""
|
||||
@ -1,14 +0,0 @@
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
|
||||
"""
|
||||
Place the files to be used for building the index in this directory.
|
||||
"""
|
||||
@ -1,14 +0,0 @@
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
|
||||
"""
|
||||
Place the prompts to be used for solving problems in this directory.
|
||||
"""
|
||||
@ -1,115 +0,0 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
# flake8: noqa
|
||||
import asyncio
|
||||
import json
|
||||
from typing import List
|
||||
from kag.interface import (
|
||||
ExecutorABC,
|
||||
LLMClient,
|
||||
Task,
|
||||
Context,
|
||||
RetrieverOutputMerger,
|
||||
RetrieverABC,
|
||||
)
|
||||
|
||||
|
||||
@ExecutorABC.register("evidence_based_reasoner")
|
||||
class EvidenceBasedReasoner(ExecutorABC):
|
||||
def __init__(
|
||||
self,
|
||||
llm: LLMClient,
|
||||
retrievers: List[RetrieverABC],
|
||||
merger: RetrieverOutputMerger,
|
||||
):
|
||||
self.llm = llm
|
||||
self.retrievers = retrievers
|
||||
self.merger = merger
|
||||
|
||||
async def ainvoke(self, query: str, task: Task, context: Context, **kwargs):
|
||||
# retrieve_task = Task(
|
||||
# executor=self.retriever.schema()["name"],
|
||||
# arguments=task.arguments,
|
||||
# id=task.id,
|
||||
# )
|
||||
|
||||
retrieval_futures = []
|
||||
for retriever in self.retrievers:
|
||||
retrieval_futures.append(
|
||||
asyncio.create_task(retriever.ainvoke(task, **kwargs))
|
||||
)
|
||||
outputs = await asyncio.gather(*retrieval_futures)
|
||||
merged = await self.merger.ainvoke(task, outputs, **kwargs)
|
||||
|
||||
retrieved_docs = []
|
||||
for chunk in merged.chunks:
|
||||
retrieved_docs.append(chunk.content)
|
||||
retrieved_docs = "\n\n".join(retrieved_docs)
|
||||
|
||||
system_instruction = """
|
||||
作为解决复杂多跳问题的专家，我需要您协助我解答一个多跳问题。该问题被拆分成多个简单的单跳查询，每个问题可能依赖于前面问题的答案，也就是说，问题正文可能包含诸如“{{i.output}}”之类的内容，表示第i个子问题的答案。我将为您提供一些关于如何解答这些初步问题（或答案本身）的见解，这些见解对于准确解决问题至关重要。此外，我将提供与当前问题相关的文本摘录，建议您仔细阅读并彻底理解。您的回复请以“思考：”开头，概述逐步得出结论的思考过程。最后以“答案：”结尾，以便清晰准确地给出答案，无需任何额外的注释。
|
||||
|
||||
召回文档：
|
||||
Sylvester
|
||||
Sylvester 这个名字源自拉丁语形容词 silvestris，意为“树木繁茂的”或“荒野的”，而 silvestris 又源自名词 silva，意为“林地”。古典拉丁语将其拼写为 i。在古典拉丁语中，y 代表与 i 不同的独立发音，这并非拉丁语的固有发音，而是用于转录外来词的发音。古典时期之后，y 的发音开始为 i。用 Sylv（代替 Silv）拼写的拼写可以追溯到古典时期之后。
|
||||
|
||||
伊利诺伊州香槟县斯坦顿镇
|
||||
斯坦顿镇是美国伊利诺伊州香槟县的一个镇区。根据 2010 年人口普查，其人口为 505 人，共有 202 个住房单元。
|
||||
|
||||
纽约州蒙特贝罗
|
||||
蒙特贝罗（意大利语：美丽的山峰）是美国纽约州罗克兰县拉马波镇的一个建制村。它位于萨弗恩以北、希尔伯恩以东、韦斯利山以南、艾尔蒙特以西。2010 年人口普查时，人口为 4,526 人。
|
||||
|
||||
埃里克·霍特
|
||||
埃里克·霍特（1987 年 2 月 16 日出生于纽约州蒙特贝罗）是一名美国足球运动员，目前是自由球员。
|
||||
|
||||
问题：
|
||||
0：公元 800 年，谁被加冕为西方皇帝？
|
||||
思考：提供的关于查理曼大帝的一段文字表明，他于 800 年被加冕为神圣罗马帝国皇帝。答案：查理曼大帝。
|
||||
|
||||
1：{{0.output}} 后来被称为什么？
|
||||
思考：为了确定 {{0.oputput}}（查理曼大帝）后来被称为什么，我需要复习一下提供的关于查理曼大帝的文章。文章表明查理曼大帝也被称为“查理大帝”。答案：查理大帝
|
||||
|
||||
2：在 {{0.output}} 时代，Sylvester 这个姓氏起源于什么语言？
|
||||
思考：这个问题询问的是 {{0.output}} 查理曼大帝统治时期，Sylvester 这个姓氏的起源，当时正值中世纪早期。关于 Sylvester 这个名字的文章指出，它源于拉丁语。答案：拉丁语
|
||||
"""
|
||||
query = f"{task.id}: {task.arguments['query']}"
|
||||
subqa = []
|
||||
for pt in task.parents:
|
||||
subq = f"{pt.id}: {pt.arguments['query']}"
|
||||
result = json.loads(pt.result)
|
||||
suba = str(result["response"])
|
||||
subqa.append(f"{subq}\n{suba}")
|
||||
subqa = "\n\n".join(subqa)
|
||||
request = f"{system_instruction}\nDocs:\n{retrieved_docs}\nQuestions:\n{subqa}\n{query}"
|
||||
|
||||
# print(f"Reasoner request = {request}")
|
||||
response = await self.llm.acall(request)
|
||||
# print(f"Reasoner response = {response}")
|
||||
task.update_memory("retriever", merged)
|
||||
task.result = json.dumps(
|
||||
{"query": task.arguments["query"], "response": response}, ensure_ascii=False
|
||||
)
|
||||
|
||||
return response
|
||||
|
||||
def schema(self, func_name: str = None):
|
||||
return {
|
||||
"name": "Retriever",
|
||||
"description": "Synthesizes precise, evidence-backed answers to user queries by analyzing provided contextual documents. Note: Contextual documents are pre-loaded and processed implicitly; no explicit context parameter is required.",
|
||||
"parameters": {
|
||||
"query": {
|
||||
"type": "string",
|
||||
"description": "User-provided query.",
|
||||
"optional": False,
|
||||
},
|
||||
},
|
||||
}
|
||||
@ -1,76 +0,0 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
# Copyright 2023 OpenSPG Authors
|
||||
#
|
||||
# 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.
|
||||
# flake8: noqa
|
||||
import json
|
||||
|
||||
from kag.common.tools.algorithm_tool.rerank.rerank_by_vector import RerankByVector
|
||||
from kag.interface import GeneratorABC, LLMClient
|
||||
from kag.solver.executor.retriever.local_knowledge_base.kag_retriever.kag_hybrid_executor import (
|
||||
to_reference_list,
|
||||
)
|
||||
|
||||
|
||||
@GeneratorABC.register("llm_generator_with_thought")
|
||||
class LLMGeneratorWithThought(GeneratorABC):
|
||||
def __init__(
|
||||
self,
|
||||
llm_client: LLMClient,
|
||||
chunk_reranker: RerankByVector = None,
|
||||
**kwargs,
|
||||
):
|
||||
super().__init__(**kwargs)
|
||||
self.llm_client = llm_client
|
||||
self.chunk_reranker = chunk_reranker or RerankByVector.from_config(
|
||||
{
|
||||
"type": "rerank_by_vector",
|
||||
}
|
||||
)
|
||||
|
||||
def invoke(self, query, context, **kwargs):
|
||||
rerank_queries = []
|
||||
chunks = []
|
||||
thoughts = []
|
||||
for task in context.gen_task(False):
|
||||
print(f"task.result = {task.result}")
|
||||
task_result = json.loads(task.result)
|
||||
subq = task_result["query"]
|
||||
suba = task_result["response"]
|
||||
thoughts.append(f"Sub-Query: {subq}\n{suba}")
|
||||
retrieved_docs = task.memory.get("retriever")
|
||||
if retrieved_docs and self.chunk_reranker:
|
||||
rerank_queries.append(task.arguments["query"])
|
||||
chunks.append(retrieved_docs.chunks)
|
||||
rerank_chunks = self.chunk_reranker.invoke(query, rerank_queries, chunks)
|
||||
total_reference_source = rerank_chunks
|
||||
refer_data = to_reference_list(
|
||||
prefix_id=0, retrieved_datas=total_reference_source
|
||||
)
|
||||
|
||||
refer_data = [f"Title:{x['document_name']}\n{x['content']}" for x in refer_data]
|
||||
refer_data = "\n\n".join(refer_data)
|
||||
thoughts = "\n\n".join(thoughts)
|
||||
|
||||
system_instruction = """
|
||||
作为一名高级阅读理解助手，你的任务是根据我提供的上下文回答复杂的多跳问题。我提供的上下文包含两部分：一组有助于回答问题的文档，以及对问题的逐步分解和分析性思维过程。请结合这两部分上下文来回答问题。你的回答应从“思考：”之后开始，逐步系统地分解推理过程，并说明你是如何得出结论的。最后以“答案：”结尾，给出简洁、明确的答案，无需额外的阐述。\n
|
||||
注意：
|
||||
1. 我希望你的答案与召回文档完全一致。
|
||||
2. 如果您认为所提供的文件无法回答问题，请回答“未知”。
|
||||
"""
|
||||
|
||||
prompt = (
|
||||
f"{system_instruction}\n\n召回文档:\n{refer_data}\n思考:\n{thoughts}问题: {query}"
|
||||
)
|
||||
response = self.llm_client(prompt)
|
||||
if "答案：" not in response:
|
||||
raise ValueError(f"no answer found in response: {response}")
|
||||
answer = response.split("答案：")[1].strip()
|
||||
return answer
|
||||
Some files were not shown because too many files have changed in this diff Show More
Loading…
x
Reference in New Issue
Block a user