yujunjun/graphrag
1
0
Fork 0
mirror of https://github.com/microsoft/graphrag.git synced 2025-09-17 20:24:20 +00:00
Code Issues Packages Projects Releases Wiki Activity
graphrag/posts/query/notebooks/global_search_nb/index.html

201 lines
26 KiB
HTML
Raw Blame History

<div style="position: relative">
<pre class="language-python"><code id="code-0" class="language-python"><span class="token comment"># Copyright (c) 2024 Microsoft Corporation. All rights reserved.</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-0" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<p>'\nCopyright (c) Microsoft Corporation. All rights reserved.\n'</p>
<div style="position: relative">
<pre class="language-python"><code id="code-4" class="language-python"><span class="token keyword">import</span> os
<span class="token keyword">from</span> pathlib <span class="token keyword">import</span> Path
<span class="token keyword">import</span> pandas <span class="token keyword">as</span> pd
<span class="token keyword">import</span> tiktoken
<span class="token keyword">from</span> graphrag<span class="token punctuation">.</span>query<span class="token punctuation">.</span><span class="token builtin">input</span><span class="token punctuation">.</span>loaders<span class="token punctuation">.</span>dfs <span class="token keyword">import</span> read_community_reports
<span class="token keyword">from</span> graphrag<span class="token punctuation">.</span>query<span class="token punctuation">.</span>llm<span class="token punctuation">.</span>oai<span class="token punctuation">.</span>chat_openai <span class="token keyword">import</span> ChatOpenAI
<span class="token keyword">from</span> graphrag<span class="token punctuation">.</span>query<span class="token punctuation">.</span>llm<span class="token punctuation">.</span>oai<span class="token punctuation">.</span>typing <span class="token keyword">import</span> OpenaiApiType
<span class="token keyword">from</span> graphrag<span class="token punctuation">.</span>query<span class="token punctuation">.</span>structured_search<span class="token punctuation">.</span>global_search<span class="token punctuation">.</span>community_context <span class="token keyword">import</span> <span class="token punctuation">(</span>
GlobalCommunityContext<span class="token punctuation">,</span>
<span class="token punctuation">)</span>
<span class="token keyword">from</span> graphrag<span class="token punctuation">.</span>query<span class="token punctuation">.</span>structured_search<span class="token punctuation">.</span>global_search<span class="token punctuation">.</span>search <span class="token keyword">import</span> GlobalSearch
<span class="token keyword">print</span><span class="token punctuation">(</span>Path<span class="token punctuation">.</span>cwd<span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">)</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-4" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<h2>Global Search example</h2>
<p>Global search method generates answers by searching over all AI-generated community reports in a map-reduce fashion. This is a resource-intensive method, but often gives good responses for questions that require an understanding of the dataset as a whole (e.g. What are the most significant values of the herbs mentioned in this notebook?).</p>
<h3>LLM setup</h3>
<div style="position: relative">
<pre class="language-python"><code id="code-14" class="language-python">api_key <span class="token operator">=</span> os<span class="token punctuation">.</span>environ<span class="token punctuation">[</span><span class="token string">"GRAPHRAG_API_KEY"</span><span class="token punctuation">]</span>
llm_model <span class="token operator">=</span> os<span class="token punctuation">.</span>environ<span class="token punctuation">[</span><span class="token string">"GRAPHRAG_EMBEDDING_MODEL"</span><span class="token punctuation">]</span>
llm <span class="token operator">=</span> ChatOpenAI<span class="token punctuation">(</span>
api_key<span class="token operator">=</span>api_key<span class="token punctuation">,</span>
model<span class="token operator">=</span>llm_model<span class="token punctuation">,</span>
api_type<span class="token operator">=</span>OpenaiApiType<span class="token punctuation">.</span>OpenAI<span class="token punctuation">,</span> <span class="token comment"># OpenaiApiType.OpenAI or OpenaiApiType.AzureOpenAI</span>
max_retries<span class="token operator">=</span><span class="token number">20</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span>
token_encoder <span class="token operator">=</span> tiktoken<span class="token punctuation">.</span>get_encoding<span class="token punctuation">(</span><span class="token string">"cl100k_base"</span><span class="token punctuation">)</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-14" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<h3>Load community reports as context for global search</h3>
<ul>
<li>Load all community reports from <strong>create_final_community_reports</strong> table from the ire-indexing engine.</li>
</ul>
<div style="position: relative">
<pre class="language-python"><code id="code-25" class="language-python"><span class="token comment"># parquet files generated from indexing pipeline</span>
INPUT_DIR <span class="token operator">=</span> <span class="token string">"./inputs/operation dulce"</span>
COMMUNITY_REPORT_TABLE <span class="token operator">=</span> <span class="token string">"create_final_community_reports"</span>
ENTITY_TABLE <span class="token operator">=</span> <span class="token string">"create_final_nodes"</span>
<span class="token comment"># community level in the Leiden community hierarchy from which we will load the community reports</span>
<span class="token comment"># higher value means we use reports on smaller communities (and thus will have more reports to query aga</span>
COMMUNITY_LEVEL <span class="token operator">=</span> <span class="token number">2</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-25" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<div style="position: relative">
<pre class="language-python"><code id="code-26" class="language-python">entity_df <span class="token operator">=</span> pd<span class="token punctuation">.</span>read_parquet<span class="token punctuation">(</span><span class="token string-interpolation"><span class="token string">f"</span><span class="token interpolation"><span class="token punctuation">{</span>INPUT_DIR<span class="token punctuation">}</span></span><span class="token string">/</span><span class="token interpolation"><span class="token punctuation">{</span>ENTITY_TABLE<span class="token punctuation">}</span></span><span class="token string">.parquet"</span></span><span class="token punctuation">)</span>
entity_df <span class="token operator">=</span> entity_df<span class="token punctuation">[</span>
<span class="token punctuation">(</span>entity_df<span class="token punctuation">.</span><span class="token builtin">type</span> <span class="token operator">==</span> <span class="token string">"entity"</span><span class="token punctuation">)</span> <span class="token operator">&amp;</span> <span class="token punctuation">(</span>entity_df<span class="token punctuation">.</span>level <span class="token operator">&lt;=</span> <span class="token string-interpolation"><span class="token string">f"level_</span><span class="token interpolation"><span class="token punctuation">{</span>COMMUNITY_LEVEL<span class="token punctuation">}</span></span><span class="token string">"</span></span><span class="token punctuation">)</span>
<span class="token punctuation">]</span>
entity_df<span class="token punctuation">[</span><span class="token string">"community"</span><span class="token punctuation">]</span> <span class="token operator">=</span> entity_df<span class="token punctuation">[</span><span class="token string">"community"</span><span class="token punctuation">]</span><span class="token punctuation">.</span>fillna<span class="token punctuation">(</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">)</span>
entity_df<span class="token punctuation">[</span><span class="token string">"community"</span><span class="token punctuation">]</span> <span class="token operator">=</span> entity_df<span class="token punctuation">[</span><span class="token string">"community"</span><span class="token punctuation">]</span><span class="token punctuation">.</span>astype<span class="token punctuation">(</span><span class="token builtin">int</span><span class="token punctuation">)</span>
entity_df <span class="token operator">=</span> entity_df<span class="token punctuation">.</span>groupby<span class="token punctuation">(</span><span class="token punctuation">[</span><span class="token string">"title"</span><span class="token punctuation">]</span><span class="token punctuation">)</span><span class="token punctuation">.</span>agg<span class="token punctuation">(</span><span class="token punctuation">{</span><span class="token string">"community"</span><span class="token punctuation">:</span> <span class="token string">"max"</span><span class="token punctuation">}</span><span class="token punctuation">)</span><span class="token punctuation">.</span>reset_index<span class="token punctuation">(</span><span class="token punctuation">)</span>
entity_df<span class="token punctuation">[</span><span class="token string">"community"</span><span class="token punctuation">]</span> <span class="token operator">=</span> entity_df<span class="token punctuation">[</span><span class="token string">"community"</span><span class="token punctuation">]</span><span class="token punctuation">.</span>astype<span class="token punctuation">(</span><span class="token builtin">str</span><span class="token punctuation">)</span>
filtered_community_df <span class="token operator">=</span> entity_df<span class="token punctuation">.</span>rename<span class="token punctuation">(</span>columns<span class="token operator">=</span><span class="token punctuation">{</span><span class="token string">"community"</span><span class="token punctuation">:</span> <span class="token string">"community_id"</span><span class="token punctuation">}</span><span class="token punctuation">)</span><span class="token punctuation">[</span>
<span class="token string">"community_id"</span>
<span class="token punctuation">]</span><span class="token punctuation">.</span>drop_duplicates<span class="token punctuation">(</span><span class="token punctuation">)</span>
report_df <span class="token operator">=</span> pd<span class="token punctuation">.</span>read_parquet<span class="token punctuation">(</span><span class="token string-interpolation"><span class="token string">f"</span><span class="token interpolation"><span class="token punctuation">{</span>INPUT_DIR<span class="token punctuation">}</span></span><span class="token string">/</span><span class="token interpolation"><span class="token punctuation">{</span>COMMUNITY_REPORT_TABLE<span class="token punctuation">}</span></span><span class="token string">.parquet"</span></span><span class="token punctuation">)</span>
report_df <span class="token operator">=</span> report_df<span class="token punctuation">[</span>report_df<span class="token punctuation">.</span>level <span class="token operator">&lt;=</span> <span class="token string-interpolation"><span class="token string">f"level_</span><span class="token interpolation"><span class="token punctuation">{</span>COMMUNITY_LEVEL<span class="token punctuation">}</span></span><span class="token string">"</span></span><span class="token punctuation">]</span>
report_df<span class="token punctuation">[</span><span class="token string">"rank"</span><span class="token punctuation">]</span> <span class="token operator">=</span> report_df<span class="token punctuation">[</span><span class="token string">"rank"</span><span class="token punctuation">]</span><span class="token punctuation">.</span>fillna<span class="token punctuation">(</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">)</span>
report_df<span class="token punctuation">[</span><span class="token string">"rank"</span><span class="token punctuation">]</span> <span class="token operator">=</span> report_df<span class="token punctuation">[</span><span class="token string">"rank"</span><span class="token punctuation">]</span><span class="token punctuation">.</span>astype<span class="token punctuation">(</span><span class="token builtin">int</span><span class="token punctuation">)</span>
report_df <span class="token operator">=</span> report_df<span class="token punctuation">.</span>merge<span class="token punctuation">(</span>filtered_community_df<span class="token punctuation">,</span> on<span class="token operator">=</span><span class="token string">"community_id"</span><span class="token punctuation">,</span> how<span class="token operator">=</span><span class="token string">"inner"</span><span class="token punctuation">)</span>
reports <span class="token operator">=</span> read_community_reports<span class="token punctuation">(</span>
df<span class="token operator">=</span>report_df<span class="token punctuation">,</span>
id_col<span class="token operator">=</span><span class="token string">"community_id"</span><span class="token punctuation">,</span>
short_id_col<span class="token operator">=</span><span class="token string">"community_id"</span><span class="token punctuation">,</span>
community_col<span class="token operator">=</span><span class="token string">"community_id"</span><span class="token punctuation">,</span>
title_col<span class="token operator">=</span><span class="token string">"title"</span><span class="token punctuation">,</span>
summary_col<span class="token operator">=</span><span class="token string">"summary"</span><span class="token punctuation">,</span>
content_col<span class="token operator">=</span><span class="token string">"full_content"</span><span class="token punctuation">,</span>
rank_col<span class="token operator">=</span><span class="token string">"rank"</span><span class="token punctuation">,</span>
summary_embedding_col<span class="token operator">=</span><span class="token boolean">None</span><span class="token punctuation">,</span>
content_embedding_col<span class="token operator">=</span><span class="token boolean">None</span><span class="token punctuation">,</span>
<span class="token punctuation">)</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string-interpolation"><span class="token string">f"Report records: </span><span class="token interpolation"><span class="token punctuation">{</span><span class="token builtin">len</span><span class="token punctuation">(</span>report_df<span class="token punctuation">)</span><span class="token punctuation">}</span></span><span class="token string">"</span></span><span class="token punctuation">)</span>
report_df<span class="token punctuation">.</span>head<span class="token punctuation">(</span><span class="token punctuation">)</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-26" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<h4>Build global context based on community reports</h4>
<div style="position: relative">
<pre class="language-python"><code id="code-30" class="language-python">context_builder <span class="token operator">=</span> GlobalCommunityContext<span class="token punctuation">(</span>
community_reports<span class="token operator">=</span>reports<span class="token punctuation">,</span> token_encoder<span class="token operator">=</span>token_encoder
<span class="token punctuation">)</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-30" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<h4>Perform global search</h4>
<div style="position: relative">
<pre class="language-python"><code id="code-34" class="language-python">context_builder_params <span class="token operator">=</span> <span class="token punctuation">{</span>
<span class="token string">"use_community_summary"</span><span class="token punctuation">:</span> <span class="token boolean">False</span><span class="token punctuation">,</span> <span class="token comment"># False means using full community reports. True means using community short summaries.</span>
<span class="token string">"shuffle_data"</span><span class="token punctuation">:</span> <span class="token boolean">True</span><span class="token punctuation">,</span>
<span class="token string">"include_community_rank"</span><span class="token punctuation">:</span> <span class="token boolean">True</span><span class="token punctuation">,</span>
<span class="token string">"min_community_rank"</span><span class="token punctuation">:</span> <span class="token number">0</span><span class="token punctuation">,</span>
<span class="token string">"max_tokens"</span><span class="token punctuation">:</span> <span class="token number">12_000</span><span class="token punctuation">,</span> <span class="token comment"># change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 5000)</span>
<span class="token string">"context_name"</span><span class="token punctuation">:</span> <span class="token string">"Reports"</span><span class="token punctuation">,</span>
<span class="token punctuation">}</span>
map_llm_params <span class="token operator">=</span> <span class="token punctuation">{</span>
<span class="token string">"max_tokens"</span><span class="token punctuation">:</span> <span class="token number">500</span><span class="token punctuation">,</span>
<span class="token string">"temperature"</span><span class="token punctuation">:</span> <span class="token number">0.0</span><span class="token punctuation">,</span>
<span class="token punctuation">}</span>
reduce_llm_params <span class="token operator">=</span> <span class="token punctuation">{</span>
<span class="token string">"max_tokens"</span><span class="token punctuation">:</span> <span class="token number">2000</span><span class="token punctuation">,</span> <span class="token comment"># change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 1000-1500)</span>
<span class="token string">"temperature"</span><span class="token punctuation">:</span> <span class="token number">0.0</span><span class="token punctuation">,</span>
<span class="token punctuation">}</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-34" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<div style="position: relative">
<pre class="language-python"><code id="code-35" class="language-python">search_engine <span class="token operator">=</span> GlobalSearch<span class="token punctuation">(</span>
llm<span class="token operator">=</span>llm<span class="token punctuation">,</span>
context_builder<span class="token operator">=</span>context_builder<span class="token punctuation">,</span>
token_encoder<span class="token operator">=</span>token_encoder<span class="token punctuation">,</span>
max_data_tokens<span class="token operator">=</span><span class="token number">16_000</span><span class="token punctuation">,</span> <span class="token comment"># change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 5000)</span>
map_llm_params<span class="token operator">=</span>map_llm_params<span class="token punctuation">,</span>
reduce_llm_params<span class="token operator">=</span>reduce_llm_params<span class="token punctuation">,</span>
context_builder_params<span class="token operator">=</span>context_builder_params<span class="token punctuation">,</span>
concurrent_coroutines<span class="token operator">=</span><span class="token number">32</span><span class="token punctuation">,</span>
response_type<span class="token operator">=</span><span class="token string">"multiple paragraphs"</span><span class="token punctuation">,</span> <span class="token comment"># free form text describing the response type and format, can be anything, e.g. prioritized list, single paragraph, multiple paragraphs, multiple-page report</span>
<span class="token punctuation">)</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-35" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<div style="position: relative">
<pre class="language-python"><code id="code-36" class="language-python">result <span class="token operator">=</span> <span class="token keyword">await</span> search_engine<span class="token punctuation">.</span>asearch<span class="token punctuation">(</span>
<span class="token string">"What is the major conflict in this story and who are the protagonist and antagonist?"</span>
<span class="token punctuation">)</span>
<span class="token keyword">print</span><span class="token punctuation">(</span>result<span class="token punctuation">.</span>response<span class="token punctuation">)</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-36" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<div style="position: relative">
<pre class="language-python"><code id="code-37" class="language-python"><span class="token comment"># inspect the data used to build the context for the LLM responses</span>
result<span class="token punctuation">.</span>context_data<span class="token punctuation">[</span><span class="token string">"reports"</span><span class="token punctuation">]</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-37" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<div style="position: relative">
<pre class="language-python"><code id="code-38" class="language-python"><span class="token comment"># inspect number of LLM calls and tokens</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string-interpolation"><span class="token string">f"LLM calls: </span><span class="token interpolation"><span class="token punctuation">{</span>result<span class="token punctuation">.</span>llm_calls<span class="token punctuation">}</span></span><span class="token string">. LLM tokens: </span><span class="token interpolation"><span class="token punctuation">{</span>result<span class="token punctuation">.</span>prompt_tokens<span class="token punctuation">}</span></span><span class="token string">"</span></span><span class="token punctuation">)</span></code></pre>
<button class="code-copy " data-clipboard-target="#code-38" style="position: absolute; top: 7.5px; right: 6px; padding-top: 3px; cursor: pointer; outline: none; opacity: 0.8;" title="Copy">
<span style="display:inline-block;background:url(https://api.iconify.design/mdi/content-copy.svg) no-repeat center center / contain;width: 16px; height: 16px;" class=""></span>
</button>
</div>
<p>LLM calls: 13. LLM tokens: 184660</p>
Reference in New Issue View Git Blame Copy Permalink