deploy: 19080f1

_sources/developer_notes/prompt.rst.txt

@@ -19,11 +19,12 @@ Design
 ----------------
 
 `LightRAG` seeks to maximize developers' control over the prompt.
-Thus, in most cases we help developers gather together different sections and form them into one prompt.
+Thus, in most cases, we help developers gather different sections and form them into one prompt.
 This prompt will then be send to the LLM as a single message.
 The default role of the message we use is `system`.
 Though it is not a special token, we use ``<SYS></SYS>`` to represent the system message in the prompt, which works quite well.
 
+
 .. code-block:: python
 
     simple_prompt = r"""<SYS> You are a helpful assistant. </SYS> User: What can you help me with?"""
@@ -53,7 +54,8 @@ Data Flow in LLM applications
 
 Look at the most complicated case: We will have user query, retrieved context, task description, definition of tools, few-shot examples, past conversation history, step history from the agent, and the output format specification.
 All these different parts need to be formatted into a single prompt.
-We have to do all this with flexiblity and also easy for developers to read.
+We have to do all this with flexibility and also make it easy for developers to read.
+
 
 
 Why Jinja2?
@@ -85,10 +87,11 @@ To format the prompt, you can use any of Python's native string formatting.
 
 
 We opted for `Jinja2` [1]_ as the templating engine for the prompt.
-Besides of the placeholders using ``{{}}`` for key-word arguments, Jinja2 also allow users to write code similar to Python syntax.
-This includes conditionals, loops, filters, and even comments that is lacked from Python's native string formatting.
+Besides the placeholders using ``{{}}`` for keyword arguments, Jinja2 also allow users to write code similar to Python syntax.
+This includes conditionals, loops, filters, and even comments, which are lacking in Python's native string formatting.
 Here is one example of using `Jinja2` to format the prompt:
 
+
 .. code-block:: python
 
    def jinja2_template_example(**kwargs):
@@ -144,8 +147,7 @@ Prompt class
 
 
 We created our :class:`Prompt Component<core.prompt_builder.Prompt>` to render the prompt with the string ``template`` and ``prompt_kwargs``.
-It is a rather simple component, but it is rather handy.
-
+It is a simple component, but it is quite handy.
 Let's use the same template as above:
 
 .. code-block:: python
@@ -163,7 +165,7 @@ Let's use the same template as above:
    print(prompt(input_str=input_str)) # takes the rest arguments in keyword arguments
 
 The ``Prompt`` class allow us to preset some of the prompt arguments at initialization, and then we can call the prompt with the rest of the arguments.
-Also, by subclassing ``Component``, we get to easily visualize this component with ``print``.
+Also, by subclassing ``Component``, we can easily visualize this component with ``print``.
 Here is the output:
 
 .. code-block::
@@ -181,12 +183,13 @@ Here is the output:
       User: {{ input_str }}, prompt_kwargs: {'task_desc_str': 'You are a helpful assitant', 'tools': ['google', 'wikipedia', 'wikidata']}, prompt_variables: ['input_str', 'tools', 'task_desc_str']
    )
 
-As all components, you can use ``to_dict`` and ``from_dict`` to serialize and deserialize the component.
+As with all components, you can use ``to_dict`` and ``from_dict`` to serialize and deserialize the component.
 
 Default Prompt Template
 -------------------------
-In default, ``Prompt`` class uses the :const:`DEFAULT_LIGHTRAG_SYSTEM_PROMPT<core.default_prompt_template.DEFAULT_LIGHTRAG_SYSTEM_PROMPT>` as its string template if no template is provided.
-This default template will allow you conditionally passing seven important variables designed from the data flow diagram above.
+
+In default, the ``Prompt`` class uses the :const:`DEFAULT_LIGHTRAG_SYSTEM_PROMPT<core.default_prompt_template.DEFAULT_LIGHTRAG_SYSTEM_PROMPT>` as its string template if no template is provided.
+This default template allows you to conditionally passing seven important variables designed from the data flow diagram above.
 These varaibles are:
 
 .. code-block:: python
@@ -210,7 +213,7 @@ Now, let's see the minimum case where we only have the user query:
    output = prompt(input_str=input_str)
    print(output)
 
-The output will be bare minimum with only the user query and a prefix for assistant to respond:
+The output will be the bare minimum with only the user query and a prefix for assistant to respond:
 
 .. code-block::
 
@@ -221,7 +224,7 @@ The output will be bare minimum with only the user query and a prefix for assist
 
 .. note::
 
-   We barely need to use the raw ``Prompt`` class directly as it is orchestrated by the ``Generator`` component.
+   In reality, we barely need to use the raw ``Prompt`` class directly as it is orchestrated by the ``Generator`` component together with the ``ModelClient`` that we will introduce next.
 
 
 

developer_notes/base_data_class.html

@@ -471,7 +471,7 @@
 In LLM applications, data constantly needs to interact with LLMs in the form of strings via prompt and be parsed back to structured data from LLMs’ text prediction.
 <a class="reference internal" href="../apis/core/core.base_data_class.html#core.base_data_class.DataClass" title="core.base_data_class.DataClass"><code class="xref py py-class docutils literal notranslate"><span class="pre">core.base_data_class.DataClass</span></code></a> is designed to ease the data interaction with LLMs via prompt(input) and text prediction(output).</p>
 <figure class="align-center" id="id1">
-<a class="reference internal image-reference" href="_static/images/dataclass.png"><img alt="DataClass" src="_static/images/dataclass.png" style="width: 680px;" />
+<a class="reference internal image-reference" href="../_images/dataclass.png"><img alt="DataClass" src="../_images/dataclass.png" style="width: 680px;" />
 </a>
 <figcaption>
 <p><span class="caption-text">DataClass is to ease the data interaction with LLMs via prompt(input) and text prediction(output).</span><a class="headerlink" href="#id1" title="Link to this image">#</a></p>

developer_notes/prompt.html

@@ -478,7 +478,7 @@ <h2>Context<a class="headerlink" href="#context" title="Link to this heading">#<
 <section id="design">
 <h2>Design<a class="headerlink" href="#design" title="Link to this heading">#</a></h2>
 <p><cite>LightRAG</cite> seeks to maximize developers’ control over the prompt.
-Thus, in most cases we help developers gather together different sections and form them into one prompt.
+Thus, in most cases, we help developers gather different sections and form them into one prompt.
 This prompt will then be send to the LLM as a single message.
 The default role of the message we use is <cite>system</cite>.
 Though it is not a special token, we use <code class="docutils literal notranslate"><span class="pre">&lt;SYS&gt;&lt;/SYS&gt;</span></code> to represent the system message in the prompt, which works quite well.</p>
@@ -505,7 +505,7 @@ <h3>Data Flow in LLM applications<a class="headerlink" href="#data-flow-in-llm-a
 </figure>
 <p>Look at the most complicated case: We will have user query, retrieved context, task description, definition of tools, few-shot examples, past conversation history, step history from the agent, and the output format specification.
 All these different parts need to be formatted into a single prompt.
-We have to do all this with flexiblity and also easy for developers to read.</p>
+We have to do all this with flexibility and also make it easy for developers to read.</p>
 </section>
 <section id="why-jinja2">
 <h3>Why Jinja2?<a class="headerlink" href="#why-jinja2" title="Link to this heading">#</a></h3>
@@ -531,8 +531,8 @@ <h3>Why Jinja2?<a class="headerlink" href="#why-jinja2" title="Link to this head
 </pre></div>
 </div>
 <p>We opted for <cite>Jinja2</cite> <a class="footnote-reference brackets" href="#id3" id="id2" role="doc-noteref"><span class="fn-bracket">[</span>1<span class="fn-bracket">]</span></a> as the templating engine for the prompt.
-Besides of the placeholders using <code class="docutils literal notranslate"><span class="pre">{{}}</span></code> for key-word arguments, Jinja2 also allow users to write code similar to Python syntax.
-This includes conditionals, loops, filters, and even comments that is lacked from Python’s native string formatting.
+Besides the placeholders using <code class="docutils literal notranslate"><span class="pre">{{}}</span></code> for keyword arguments, Jinja2 also allow users to write code similar to Python syntax.
+This includes conditionals, loops, filters, and even comments, which are lacking in Python’s native string formatting.
 Here is one example of using <cite>Jinja2</cite> to format the prompt:</p>
 <div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">jinja2_template_example</span><span class="p">(</span><span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    <span class="kn">from</span> <span class="nn">jinja2</span> <span class="kn">import</span> <span class="n">Template</span>
@@ -579,8 +579,8 @@ <h3>Why Jinja2?<a class="headerlink" href="#why-jinja2" title="Link to this head
 <section id="prompt-class">
 <h2>Prompt class<a class="headerlink" href="#prompt-class" title="Link to this heading">#</a></h2>
 <p>We created our <a class="reference internal" href="../apis/core/core.prompt_builder.html#core.prompt_builder.Prompt" title="core.prompt_builder.Prompt"><code class="xref py py-class docutils literal notranslate"><span class="pre">Prompt</span> <span class="pre">Component</span></code></a> to render the prompt with the string <code class="docutils literal notranslate"><span class="pre">template</span></code> and <code class="docutils literal notranslate"><span class="pre">prompt_kwargs</span></code>.
-It is a rather simple component, but it is rather handy.</p>
-<p>Let’s use the same template as above:</p>
+It is a simple component, but it is quite handy.
+Let’s use the same template as above:</p>
 <div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">lightrag.core.prompt_builder</span> <span class="kn">import</span> <span class="n">Prompt</span>
 
 <span class="n">prompt</span> <span class="o">=</span> <span class="n">Prompt</span><span class="p">(</span>
@@ -595,7 +595,7 @@ <h2>Prompt class<a class="headerlink" href="#prompt-class" title="Link to this h
 </pre></div>
 </div>
 <p>The <code class="docutils literal notranslate"><span class="pre">Prompt</span></code> class allow us to preset some of the prompt arguments at initialization, and then we can call the prompt with the rest of the arguments.
-Also, by subclassing <code class="docutils literal notranslate"><span class="pre">Component</span></code>, we get to easily visualize this component with <code class="docutils literal notranslate"><span class="pre">print</span></code>.
+Also, by subclassing <code class="docutils literal notranslate"><span class="pre">Component</span></code>, we can easily visualize this component with <code class="docutils literal notranslate"><span class="pre">print</span></code>.
 Here is the output:</p>
 <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">Prompt</span><span class="p">(</span>
    <span class="n">template</span><span class="p">:</span> <span class="o">&lt;</span><span class="n">SYS</span><span class="o">&gt;</span><span class="p">{{</span> <span class="n">task_desc_str</span> <span class="p">}}</span><span class="o">&lt;/</span><span class="n">SYS</span><span class="o">&gt;</span>
@@ -611,12 +611,12 @@ <h2>Prompt class<a class="headerlink" href="#prompt-class" title="Link to this h
 <span class="p">)</span>
 </pre></div>
 </div>
-<p>As all components, you can use <code class="docutils literal notranslate"><span class="pre">to_dict</span></code> and <code class="docutils literal notranslate"><span class="pre">from_dict</span></code> to serialize and deserialize the component.</p>
+<p>As with all components, you can use <code class="docutils literal notranslate"><span class="pre">to_dict</span></code> and <code class="docutils literal notranslate"><span class="pre">from_dict</span></code> to serialize and deserialize the component.</p>
 </section>
 <section id="default-prompt-template">
 <h2>Default Prompt Template<a class="headerlink" href="#default-prompt-template" title="Link to this heading">#</a></h2>
-<p>In default, <code class="docutils literal notranslate"><span class="pre">Prompt</span></code> class uses the <a class="reference internal" href="../apis/core/core.default_prompt_template.html#core.default_prompt_template.DEFAULT_LIGHTRAG_SYSTEM_PROMPT" title="core.default_prompt_template.DEFAULT_LIGHTRAG_SYSTEM_PROMPT"><code class="xref py py-const docutils literal notranslate"><span class="pre">DEFAULT_LIGHTRAG_SYSTEM_PROMPT</span></code></a> as its string template if no template is provided.
-This default template will allow you conditionally passing seven important variables designed from the data flow diagram above.
+<p>In default, the <code class="docutils literal notranslate"><span class="pre">Prompt</span></code> class uses the <a class="reference internal" href="../apis/core/core.default_prompt_template.html#core.default_prompt_template.DEFAULT_LIGHTRAG_SYSTEM_PROMPT" title="core.default_prompt_template.DEFAULT_LIGHTRAG_SYSTEM_PROMPT"><code class="xref py py-const docutils literal notranslate"><span class="pre">DEFAULT_LIGHTRAG_SYSTEM_PROMPT</span></code></a> as its string template if no template is provided.
+This default template allows you to conditionally passing seven important variables designed from the data flow diagram above.
 These varaibles are:</p>
 <div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">LIGHTRAG_DEFAULT_PROMPT_ARGS</span> <span class="o">=</span> <span class="p">[</span>
    <span class="s2">&quot;task_desc_str&quot;</span><span class="p">,</span>  <span class="c1"># task description</span>
@@ -636,7 +636,7 @@ <h2>Default Prompt Template<a class="headerlink" href="#default-prompt-template"
 <span class="nb">print</span><span class="p">(</span><span class="n">output</span><span class="p">)</span>
 </pre></div>
 </div>
-<p>The output will be bare minimum with only the user query and a prefix for assistant to respond:</p>
+<p>The output will be the bare minimum with only the user query and a prefix for assistant to respond:</p>
 <div class="highlight-default notranslate"><div class="highlight"><pre><span></span>&lt;User&gt;
 What is the capital of France?
 &lt;/User&gt;
@@ -645,7 +645,7 @@ <h2>Default Prompt Template<a class="headerlink" href="#default-prompt-template"
 </div>
 <div class="admonition note">
 <p class="admonition-title">Note</p>
-<p>We barely need to use the raw <code class="docutils literal notranslate"><span class="pre">Prompt</span></code> class directly as it is orchestrated by the <code class="docutils literal notranslate"><span class="pre">Generator</span></code> component.</p>
+<p>In reality, we barely need to use the raw <code class="docutils literal notranslate"><span class="pre">Prompt</span></code> class directly as it is orchestrated by the <code class="docutils literal notranslate"><span class="pre">Generator</span></code> component together with the <code class="docutils literal notranslate"><span class="pre">ModelClient</span></code> that we will introduce next.</p>
 </div>
 <div class="highlight admonition">
 <p class="admonition-title">References</p>