GraphRAG本地部署使用及兼容千帆通义
文章目录
- 前言
- 一、GraphRAG本地安装
- 1.创建环境并安装
- 2.准备demo数据
- 3.初始化demo目录
- 二、GraphRAG兼容千帆通义等大模型
- 1.安装 graphrag-more
- 2.准备Demo数据
- 3.初始化demo目录
- 4.移动和修改 settings.yaml 文件
- 三、知识库构建与使用
- 1.知识库构建
- 2.执行查询
前言
GraphRAG是一种基于知识图谱的检索增强生成（RAG）应用，不同于传统基于向量检索的RAG应用，其允许进行更深层、更细致与上下文感知的检索，从而帮助获得更高质量的输出。
GitHub地址：https://github.com/microsoft/graphrag
文档地址：https://microsoft.github.io/graphrag/get_started/
graphrag-more地址：https://github.com/guoyao/graphrag-more
一、GraphRAG本地安装
1.创建环境并安装
#使用conda创建graphrag虚拟环境
conda create -n graphrag python=3.10
#安装graphrag
pip install graphrag
2.准备demo数据
# 创建demo目录
mkdir -p ./ragtest/input
# 下载微软官方demo数据
curl https://www.gutenberg.org/cache/epub/24022/pg24022.txt -o ./ragtest/input/book.txt
3.初始化demo目录
要初始化工作区，请先运行graphrag init命令。由于我们在上一步中已经配置了一个名为./ragtest的目录，请运行以下命令：
graphrag init --root ./ragtest
这将在./ragtest目录中创建两个文件：.env和settings.yaml。
- .env包含运行GraphRAG所需的环境变量。如果您检查文件，您将看到一个定义的环境变量，GRAPHRAG_API_KEY=<API_KEY>。这是OpenAI API或Azure OpenAI端点的API密钥。您可以用您自己的API密钥替换它。如果您正在使用另一种形式的身份验证（即托管身份），请删除此文件。
- settings.yaml包含GraphRAG的设置。您可以修改此文件来更改管道的设置。
二、GraphRAG兼容千帆通义等大模型
国内使用 OpenAI / AzureOpenAI 诸多不便，因此fork了官方代码创建了个新代码库：graphrag-more在其基础上做了小部分修改，支持使用百度千帆、阿里通义、Ollama。
安装步骤如下
1.安装 graphrag-more
如需二次开发或者调试的话，也可以直接使用源码的方式，步骤如下：
下载 graphrag-more 代码库
git clone https://github.com/guoyao/graphrag-more.git
安装依赖包
这里使用 poetry 来管理python虚拟环境
安装 poetry 参考：https://python-poetry.org/docs/#installation
cd graphrag-more
poetry install
conda create -n graphrag python=3.10
pip install graphrag-more
#or 如果使用了国内的镜像源需要改为官方的镜像源下载
pip install -i https://pypi.org/simple graphrag-more
2.准备Demo数据
# 创建demo目录
mkdir -p ./ragtest-more/input
# 下载微软官方demo数据
curl https://www.gutenberg.org/cache/epub/24022/pg24022.txt -o ./ragtest-more/input/book.txt
3.初始化demo目录
python -m graphrag.index --init --root ./ragtest-more
4.移动和修改 settings.yaml 文件
根据选用的模型（千帆、通义、Ollama）将 example_settings 文件夹对应模型的 settings.yaml 文件复制到 ragtest 目录，覆盖初始化过程生成的 settings.yaml 文件。
每个settings.yaml里面都设置了默认的 llm 和 embeddings 模型，根据你自己要使用的模型修改 settings.yaml 文件的 model 配置
千帆默认使用 qianfan.ERNIE-3.5-128K 和 qianfan.bge-large-zh ，注意：必须带上 qianfan. 前缀 ！！！
通义默认使用 tongyi.qwen-plus 和 tongyi.text-embedding-v2 ，注意：必须带上 tongyi. 前缀 ！！！
Ollama默认使用 ollama.mistral:latest 和 ollama.quentinz/bge-large-zh-v1.5:latest ，注意：<=0.3.0版本时，其llm模型不用带前缀，>=0.3.1版本时，其llm模型必须带上 ollama. 前缀，embeddings模型必须带 ollama. 前缀 ！！！
以下是Ollama、千帆、通义千问的配置文件内容
ollama
encoding_model: cl100k_base
skip_workflows: []
llm:
api_key: ${GRAPHRAG_API_KEY}
type: openai_chat # or azure_openai_chat
model: ollama.mistral:latest
model_supports_json: true # recommended if this is available for your model.
# max_tokens: 4000
# request_timeout: 180.0
# api_base: http://localhost:11434/v1
# api_version: 2024-02-15-preview
# organization: <organization_id>
# deployment_name: <azure_model_deployment_name>
# tokens_per_minute: 150_000 # set a leaky bucket throttle
# requests_per_minute: 10_000 # set a leaky bucket throttle
# max_retries: 10
# max_retry_wait: 10.0
# sleep_on_rate_limit_recommendation: true # whether to sleep when azure suggests wait-times
# concurrent_requests: 25 # the number of parallel inflight requests that may be made
# temperature: 0 # temperature for sampling
# top_p: 1 # top-p sampling
# n: 1 # Number of completions to generate
parallelization:
stagger: 0.3
# num_threads: 50 # the number of threads to use for parallel processing
async_mode: threaded # or asyncio
embeddings:
## parallelization: override the global parallelization settings for embeddings
async_mode: threaded # or asyncio
# target: required # or all
llm:
api_key: ${GRAPHRAG_API_KEY}
type: openai_embedding # or azure_openai_embedding
model: ollama.quentinz/bge-large-zh-v1.5:latest
# api_base: http://localhost:11434/api
# api_version: 2024-02-15-preview
# organization: <organization_id>
# deployment_name: <azure_model_deployment_name>
# tokens_per_minute: 150_000 # set a leaky bucket throttle
# requests_per_minute: 10_000 # set a leaky bucket throttle
# max_retries: 10
# max_retry_wait: 10.0
# sleep_on_rate_limit_recommendation: true # whether to sleep when azure suggests wait-times
# concurrent_requests: 25 # the number of parallel inflight requests that may be made
# batch_size: 16 # the number of documents to send in a single request
# batch_max_tokens: 8191 # the maximum number of tokens to send in a single request
# target: required # or optional
chunks:
size: 300
overlap: 100
group_by_columns: [id] # by default, we don't allow chunks to cross documents
input:
type: file # or blob
file_type: text # or csv
base_dir: "input"
file_encoding: utf-8
file_pattern: ".*\\.txt$"
cache:
type: file # or blob
base_dir: "cache"
# connection_string: <azure_blob_storage_connection_string>
# container_name: <azure_blob_storage_container_name>
storage:
type: file # or blob
base_dir: "output" # output/${timestamp}/artifacts
# connection_string: <azure_blob_storage_connection_string>
# container_name: <azure_blob_storage_container_name>
reporting:
type: file # or console, blob
base_dir: "output" # output/${timestamp}/reports
# connection_string: <azure_blob_storage_connection_string>
# container_name: <azure_blob_storage_container_name>
entity_extraction:
## strategy: fully override the entity extraction strategy.
## type: one of graph_intelligence, graph_intelligence_json and nltk
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
prompt: "prompts/entity_extraction.txt"
entity_types: [organization,person,geo,event]
max_gleanings: 1
summarize_descriptions:
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
prompt: "prompts/summarize_descriptions.txt"
max_length: 500
claim_extraction:
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
# enabled: true
prompt: "prompts/claim_extraction.txt"
description: "Any claims or facts that could be relevant to information discovery."
max_gleanings: 1
community_reports:
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
prompt: "prompts/community_report.txt"
max_length: 2000
max_input_length: 8000
cluster_graph:
max_cluster_size: 10
embed_graph:
enabled: false # if true, will generate node2vec embeddings for nodes
# num_walks: 10
# walk_length: 40
# window_size: 2
# iterations: 3
# random_seed: 597832
umap:
enabled: false # if true, will generate UMAP embeddings for nodes
snapshots:
graphml: false
raw_entities: false
top_level_nodes: false
local_search:
# text_unit_prop: 0.5
# community_prop: 0.1
# conversation_history_max_turns: 5
# top_k_mapped_entities: 10
# top_k_relationships: 10
# llm_temperature: 0 # temperature for sampling
# llm_top_p: 1 # top-p sampling
# llm_n: 1 # Number of completions to generate
# max_tokens: 12000
global_search:
# llm_temperature: 0 # temperature for sampling
# llm_top_p: 1 # top-p sampling
# llm_n: 1 # Number of completions to generate
# max_tokens: 12000
# data_max_tokens: 12000
# map_max_tokens: 1000
# reduce_max_tokens: 2000
# concurrency: 32
千帆
encoding_model: cl100k_base
skip_workflows: []
llm:
api_key: ${GRAPHRAG_API_KEY}
type: openai_chat # or azure_openai_chat
model: qianfan.ERNIE-3.5-128K
model_supports_json: false # recommended if this is available for your model, original default is true
# max_tokens: 4000
# request_timeout: 180.0
# api_base: https://<instance>.openai.azure.com
# api_version: 2024-02-15-preview
# organization: <organization_id>
# deployment_name: <azure_model_deployment_name>
# tokens_per_minute: 150_000 # set a leaky bucket throttle
# requests_per_minute: 120 # set a leaky bucket throttle，original default is 10_000
# max_retries: 10
# max_retry_wait: 10.0
# sleep_on_rate_limit_recommendation: true # whether to sleep when azure suggests wait-times
# concurrent_requests: 2 # the number of parallel inflight requests that may be made，original default is 25
temperature: 1e-10 # temperature for sampling, original default is 0
# top_p: 1 # top-p sampling
# n: 1 # Number of completions to generate
parallelization:
stagger: 0.3
# num_threads: 50 # the number of threads to use for parallel processing
async_mode: asyncio # or threaded
embeddings:
## parallelization: override the global parallelization settings for embeddings
async_mode: asyncio # or threaded
# target: required # or all
llm:
api_key: ${GRAPHRAG_API_KEY}
type: openai_embedding # or azure_openai_embedding
model: qianfan.bge-large-zh
# api_base: https://<instance>.openai.azure.com
# api_version: 2024-02-15-preview
# organization: <organization_id>
# deployment_name: <azure_model_deployment_name>
# tokens_per_minute: 150_000 # set a leaky bucket throttle
# requests_per_minute: 10_000 # set a leaky bucket throttle
# max_retries: 10
# max_retry_wait: 10.0
# sleep_on_rate_limit_recommendation: true # whether to sleep when azure suggests wait-times
# concurrent_requests: 25 # the number of parallel inflight requests that may be made
# batch_size: 16 # the number of documents to send in a single request
# batch_max_tokens: 8191 # the maximum number of tokens to send in a single request
# target: required # or optional
chunks:
size: 1200
overlap: 100
group_by_columns: [id] # by default, we don't allow chunks to cross documents
input:
type: file # or blob
file_type: text # or csv
base_dir: "input"
file_encoding: utf-8
file_pattern: ".*\\.txt$"
cache:
type: file # or blob
base_dir: "cache"
# connection_string: <azure_blob_storage_connection_string>
# container_name: <azure_blob_storage_container_name>
storage:
type: file # or blob
base_dir: "output" # output/${timestamp}/artifacts
# connection_string: <azure_blob_storage_connection_string>
# container_name: <azure_blob_storage_container_name>
reporting:
type: file # or console, blob
base_dir: "output" # output/${timestamp}/reports
# connection_string: <azure_blob_storage_connection_string>
# container_name: <azure_blob_storage_container_name>
entity_extraction:
## strategy: fully override the entity extraction strategy.
## type: one of graph_intelligence, graph_intelligence_json and nltk
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
prompt: "prompts/entity_extraction.txt"
entity_types: [organization,person,geo,event]
max_gleanings: 1
summarize_descriptions:
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
prompt: "prompts/summarize_descriptions.txt"
max_length: 500
claim_extraction:
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
# enabled: true
prompt: "prompts/claim_extraction.txt"
description: "Any claims or facts that could be relevant to information discovery."
max_gleanings: 1
community_reports:
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
prompt: "prompts/community_report.txt"
max_length: 2000
max_input_length: 8000
cluster_graph:
max_cluster_size: 10
embed_graph:
enabled: false # if true, will generate node2vec embeddings for nodes
# num_walks: 10
# walk_length: 40
# window_size: 2
# iterations: 3
# random_seed: 597832
umap:
enabled: false # if true, will generate UMAP embeddings for nodes
snapshots:
graphml: false
raw_entities: false
top_level_nodes: false
local_search:
# text_unit_prop: 0.5
# community_prop: 0.1
# conversation_history_max_turns: 5
# top_k_mapped_entities: 10
# top_k_relationships: 10
llm_temperature: 1e-10 # temperature for sampling, original default is 0
# llm_top_p: 1 # top-p sampling
# llm_n: 1 # Number of completions to generate
# max_tokens: 5000 # original default is 12000
global_search:
llm_temperature: 1e-10 # temperature for sampling, original default is 0
# llm_top_p: 1 # top-p sampling
# llm_n: 1 # Number of completions to generate
# max_tokens: 5000 # original default is 12000
# data_max_tokens: 12000
# map_max_tokens: 1000
# reduce_max_tokens: 2000
# concurrency: 32
通义千问
encoding_model: cl100k_base
skip_workflows: []
llm:
api_key: ${GRAPHRAG_API_KEY}
type: openai_chat # or azure_openai_chat
model: tongyi.qwen-plus
model_supports_json: false # recommended if this is available for your model, original default is true
# max_tokens: 4000
# request_timeout: 180.0
# api_base: https://<instance>.openai.azure.com
# api_version: 2024-02-15-preview
# organization: <organization_id>
# deployment_name: <azure_model_deployment_name>
# tokens_per_minute: 150_000 # set a leaky bucket throttle
# requests_per_minute: 10_000 # set a leaky bucket throttle
# max_retries: 10
# max_retry_wait: 10.0
# sleep_on_rate_limit_recommendation: true # whether to sleep when azure suggests wait-times
# concurrent_requests: 25 # the number of parallel inflight requests that may be made
# temperature: 0 # temperature for sampling
# top_p: 1 # top-p sampling
# n: 1 # Number of completions to generate
parallelization:
stagger: 0.3
# num_threads: 50 # the number of threads to use for parallel processing
async_mode: threaded # or asyncio
embeddings:
## parallelization: override the global parallelization settings for embeddings
async_mode: threaded # or asyncio
# target: required # or all
llm:
api_key: ${GRAPHRAG_API_KEY}
type: openai_embedding # or azure_openai_embedding
model: tongyi.text-embedding-v2
# api_base: https://<instance>.openai.azure.com
# api_version: 2024-02-15-preview
# organization: <organization_id>
# deployment_name: <azure_model_deployment_name>
# tokens_per_minute: 150_000 # set a leaky bucket throttle
# requests_per_minute: 10_000 # set a leaky bucket throttle
# max_retries: 10
# max_retry_wait: 10.0
# sleep_on_rate_limit_recommendation: true # whether to sleep when azure suggests wait-times
# concurrent_requests: 25 # the number of parallel inflight requests that may be made
# batch_size: 16 # the number of documents to send in a single request
# batch_max_tokens: 8191 # the maximum number of tokens to send in a single request
# target: required # or optional
chunks:
size: 1200
overlap: 100
group_by_columns: [id] # by default, we don't allow chunks to cross documents
input:
type: file # or blob
file_type: text # or csv
base_dir: "input"
file_encoding: utf-8
file_pattern: ".*\\.txt$"
cache:
type: file # or blob
base_dir: "cache"
# connection_string: <azure_blob_storage_connection_string>
# container_name: <azure_blob_storage_container_name>
storage:
type: file # or blob
base_dir: "output" # output/${timestamp}/artifacts
# connection_string: <azure_blob_storage_connection_string>
# container_name: <azure_blob_storage_container_name>
reporting:
type: file # or console, blob
base_dir: "output" # output/${timestamp}/reports
# connection_string: <azure_blob_storage_connection_string>
# container_name: <azure_blob_storage_container_name>
entity_extraction:
## strategy: fully override the entity extraction strategy.
## type: one of graph_intelligence, graph_intelligence_json and nltk
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
prompt: "prompts/entity_extraction.txt"
entity_types: [organization,person,geo,event]
max_gleanings: 1
summarize_descriptions:
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
prompt: "prompts/summarize_descriptions.txt"
max_length: 500
claim_extraction:
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
# enabled: true
prompt: "prompts/claim_extraction.txt"
description: "Any claims or facts that could be relevant to information discovery."
max_gleanings: 1
community_reports:
## llm: override the global llm settings for this task
## parallelization: override the global parallelization settings for this task
## async_mode: override the global async_mode settings for this task
prompt: "prompts/community_report.txt"
max_length: 2000
max_input_length: 8000
cluster_graph:
max_cluster_size: 10
embed_graph:
enabled: false # if true, will generate node2vec embeddings for nodes
# num_walks: 10
# walk_length: 40
# window_size: 2
# iterations: 3
# random_seed: 597832
umap:
enabled: false # if true, will generate UMAP embeddings for nodes
snapshots:
graphml: false
raw_entities: false
top_level_nodes: false
local_search:
# text_unit_prop: 0.5
# community_prop: 0.1
# conversation_history_max_turns: 5
# top_k_mapped_entities: 10
# top_k_relationships: 10
# llm_temperature: 0 # temperature for sampling
# llm_top_p: 1 # top-p sampling
# llm_n: 1 # Number of completions to generate
# max_tokens: 12000
global_search:
# llm_temperature: 0 # temperature for sampling
# llm_top_p: 1 # top-p sampling
# llm_n: 1 # Number of completions to generate
# max_tokens: 12000
# data_max_tokens: 12000
# map_max_tokens: 1000
# reduce_max_tokens: 2000
# concurrency: 32
三、知识库构建与使用
1.知识库构建
构建过程可能会触发 rate limit （限速）导致构建失败，重复执行几次，或者尝试调小 settings.yaml 中 的 requests_per_minute 和 concurrent_requests 配置，然后重试
#如果使用是graphrag
graphrag index --root ./ragtest
#如果使用的是graphrag-more
python -m graphrag.index --root ./ragtest-more
需要注意的是如果使用graphrag-more需要将秘钥配置环境变量
根据选用的模型，配置对应的环境变量，若使用Ollama需要安装并下载对应模型
1.千帆：需配置环境变量 QIANFAN_AK、QIANFAN_SK（注意是应用的AK/SK，不是安全认证的Access Key/Secret Key），如何获取请参考官方文档
2.通义：需配置环境变量 TONGYI_API_KEY（从0.3.6.1版本开始，也支持使用 DASHSCOPE_API_KEY，同时都配置的情况下 TONGYI_API_KEY 优先级高于 DASHSCOPE_API_KEY），如何获取请参考官方文档
3.Ollama：
安装：https://ollama.com/download ，安装后启动
下载模型
ollama pull mistral:latest
ollama pull quentinz/bge-large-zh-v1.5:latest
配置环境变量
linux: export QIANFAN_AK=“value”
2.执行查询
graphrag
# global query
graphrag query \
--root ./ragtest \
--method global \
--query "What are the top themes in this story?"
# local query
graphrag query \
--root ./ragtest \
--method local \
--query "Who is Scrooge and what are his main relationships?"
graphrag-more
# global query
python -m graphrag.query \
--root ./ragtest-more \
--method global \
"What are the top themes in this story?"
# local query
python -m graphrag.query \
--root ./ragtest-more \
--method local \
"Who is Scrooge, and what are his main relationships?"
运行结果如下
