deploy: 9d4a1bb

en/_sources/tutorial/206-prompt.md.txt

@@ -241,11 +241,11 @@ messages as input. The message must obey the following rules (updated in
 
 #### Prompt Strategy
 
-Given a list of messages, we will parse each message as follows:
-
-- `Msg`:  Fill the `role` and `content` fields directly. If it has an `url`
-  field, which refers to an image, we will add it to the message.
-- `List`: Parse each element in the list according to the above rules.
+- If the role field of the first input message is `"system"`,
+it will be treated as system prompt and the other messages will consist
+dialogue history in the system message prefixed by "## Dialogue History".
+- If the `url` attribute of messages is not `None`, we will gather all urls in
+the `"images"` field in the returned dictionary.
 
 ```python
 from agentscope.models import OllamaChatWrapper
@@ -268,9 +268,11 @@ print(prompt)
 
 ```bash
 [
-  {"role": "system", "content": "You are a helpful assistant"},
-  {"role": "assistant", "content": "Hi."},
-  {"role": "assistant", "content": "Nice to meet you!", "images": ["https://example.com/image.jpg"]},
+  {
+    "role": "system",
+    "content": "You are a helpful assistant\n\n## Dialogue History\nBob: Hi.\nAlice: Nice to meet you!",
+    "images": ["https://example.com/image.jpg"]
+  },
 ]
 ```
 

en/agentscope.models.html

@@ -675,9 +675,36 @@
 <dl class="py method">
 <dt class="sig sig-object py" id="agentscope.models.OllamaChatWrapper.format">
 <span class="sig-name descname"><span class="pre">format</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">*</span></span><span class="n"><span class="pre">args</span></span><span class="p"><span class="pre">:</span></span><span class="w"> </span><span class="n"><a class="reference internal" href="agentscope.message.html#agentscope.message.MessageBase" title="agentscope.message.MessageBase"><span class="pre">MessageBase</span></a><span class="w"> </span><span class="p"><span class="pre">|</span></span><span class="w"> </span><span class="pre">Sequence</span><span class="p"><span class="pre">[</span></span><a class="reference internal" href="agentscope.message.html#agentscope.message.MessageBase" title="agentscope.message.MessageBase"><span class="pre">MessageBase</span></a><span class="p"><span class="pre">]</span></span></span></em><span class="sig-paren">)</span> <span class="sig-return"><span class="sig-return-icon">&#x2192;</span> <span class="sig-return-typehint"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">dict</span><span class="p"><span class="pre">]</span></span></span></span><a class="reference internal" href="_modules/agentscope/models/ollama_model.html#OllamaChatWrapper.format"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#agentscope.models.OllamaChatWrapper.format" title="Link to this definition"></a></dt>
-<dd><p>A basic strategy to format the input into the required format of
-Ollama Chat API.</p>
-<p>Note for ollama chat api, the content field shouldn’t be empty string.</p>
+<dd><p>Format the messages for ollama Chat API.</p>
+<p>All messages will be formatted into a single system message with
+system prompt and dialogue history.</p>
+<p>Note:
+1. This strategy maybe not suitable for all scenarios,
+and developers are encouraged to implement their own prompt
+engineering strategies.
+2. For ollama chat api, the content field shouldn’t be empty string.</p>
+<p>Example:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">prompt</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">format</span><span class="p">(</span>
+    <span class="n">Msg</span><span class="p">(</span><span class="s2">&quot;system&quot;</span><span class="p">,</span> <span class="s2">&quot;You&#39;re a helpful assistant&quot;</span><span class="p">,</span> <span class="n">role</span><span class="o">=</span><span class="s2">&quot;system&quot;</span><span class="p">),</span>
+    <span class="n">Msg</span><span class="p">(</span><span class="s2">&quot;Bob&quot;</span><span class="p">,</span> <span class="s2">&quot;Hi, how can I help you?&quot;</span><span class="p">,</span> <span class="n">role</span><span class="o">=</span><span class="s2">&quot;assistant&quot;</span><span class="p">),</span>
+    <span class="n">Msg</span><span class="p">(</span><span class="s2">&quot;user&quot;</span><span class="p">,</span> <span class="s2">&quot;What&#39;s the date today?&quot;</span><span class="p">,</span> <span class="n">role</span><span class="o">=</span><span class="s2">&quot;user&quot;</span><span class="p">)</span>
+<span class="p">)</span>
+</pre></div>
+</div>
+<p>The prompt will be as follows:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="p">[</span>
+    <span class="p">{</span>
+        <span class="s2">&quot;role&quot;</span><span class="p">:</span> <span class="s2">&quot;user&quot;</span><span class="p">,</span>
+        <span class="s2">&quot;content&quot;</span><span class="p">:</span> <span class="p">(</span>
+            <span class="s2">&quot;You&#39;re a helpful assistant</span><span class="se">\n\n</span><span class="s2">&quot;</span>
+            <span class="s2">&quot;## Dialogue History</span><span class="se">\n</span><span class="s2">&quot;</span>
+            <span class="s2">&quot;Bob: Hi, how can I help you?</span><span class="se">\n</span><span class="s2">&quot;</span>
+            <span class="s2">&quot;user: What&#39;s the date today?&quot;</span>
+        <span class="p">)</span>
+    <span class="p">}</span>
+<span class="p">]</span>
+</pre></div>
+</div>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><p><strong>args</strong> (<cite>Union[MessageBase, Sequence[MessageBase]]</cite>) – The input arguments to be formatted, where each argument

en/agentscope.models.ollama_model.html

@@ -169,9 +169,36 @@
 <dl class="py method">
 <dt class="sig sig-object py" id="agentscope.models.ollama_model.OllamaChatWrapper.format">
 <span class="sig-name descname"><span class="pre">format</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">*</span></span><span class="n"><span class="pre">args</span></span><span class="p"><span class="pre">:</span></span><span class="w"> </span><span class="n"><a class="reference internal" href="agentscope.message.html#agentscope.message.MessageBase" title="agentscope.message.MessageBase"><span class="pre">MessageBase</span></a><span class="w"> </span><span class="p"><span class="pre">|</span></span><span class="w"> </span><span class="pre">Sequence</span><span class="p"><span class="pre">[</span></span><a class="reference internal" href="agentscope.message.html#agentscope.message.MessageBase" title="agentscope.message.MessageBase"><span class="pre">MessageBase</span></a><span class="p"><span class="pre">]</span></span></span></em><span class="sig-paren">)</span> <span class="sig-return"><span class="sig-return-icon">&#x2192;</span> <span class="sig-return-typehint"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">dict</span><span class="p"><span class="pre">]</span></span></span></span><a class="reference internal" href="_modules/agentscope/models/ollama_model.html#OllamaChatWrapper.format"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#agentscope.models.ollama_model.OllamaChatWrapper.format" title="Link to this definition"></a></dt>
-<dd><p>A basic strategy to format the input into the required format of
-Ollama Chat API.</p>
-<p>Note for ollama chat api, the content field shouldn’t be empty string.</p>
+<dd><p>Format the messages for ollama Chat API.</p>
+<p>All messages will be formatted into a single system message with
+system prompt and dialogue history.</p>
+<p>Note:
+1. This strategy maybe not suitable for all scenarios,
+and developers are encouraged to implement their own prompt
+engineering strategies.
+2. For ollama chat api, the content field shouldn’t be empty string.</p>
+<p>Example:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">prompt</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">format</span><span class="p">(</span>
+    <span class="n">Msg</span><span class="p">(</span><span class="s2">&quot;system&quot;</span><span class="p">,</span> <span class="s2">&quot;You&#39;re a helpful assistant&quot;</span><span class="p">,</span> <span class="n">role</span><span class="o">=</span><span class="s2">&quot;system&quot;</span><span class="p">),</span>
+    <span class="n">Msg</span><span class="p">(</span><span class="s2">&quot;Bob&quot;</span><span class="p">,</span> <span class="s2">&quot;Hi, how can I help you?&quot;</span><span class="p">,</span> <span class="n">role</span><span class="o">=</span><span class="s2">&quot;assistant&quot;</span><span class="p">),</span>
+    <span class="n">Msg</span><span class="p">(</span><span class="s2">&quot;user&quot;</span><span class="p">,</span> <span class="s2">&quot;What&#39;s the date today?&quot;</span><span class="p">,</span> <span class="n">role</span><span class="o">=</span><span class="s2">&quot;user&quot;</span><span class="p">)</span>
+<span class="p">)</span>
+</pre></div>
+</div>
+<p>The prompt will be as follows:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="p">[</span>
+    <span class="p">{</span>
+        <span class="s2">&quot;role&quot;</span><span class="p">:</span> <span class="s2">&quot;user&quot;</span><span class="p">,</span>
+        <span class="s2">&quot;content&quot;</span><span class="p">:</span> <span class="p">(</span>
+            <span class="s2">&quot;You&#39;re a helpful assistant</span><span class="se">\n\n</span><span class="s2">&quot;</span>
+            <span class="s2">&quot;## Dialogue History</span><span class="se">\n</span><span class="s2">&quot;</span>
+            <span class="s2">&quot;Bob: Hi, how can I help you?</span><span class="se">\n</span><span class="s2">&quot;</span>
+            <span class="s2">&quot;user: What&#39;s the date today?&quot;</span>
+        <span class="p">)</span>
+    <span class="p">}</span>
+<span class="p">]</span>
+</pre></div>
+</div>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><p><strong>args</strong> (<cite>Union[MessageBase, Sequence[MessageBase]]</cite>) – The input arguments to be formatted, where each argument

en/tutorial/206-prompt.html

@@ -346,11 +346,12 @@ <h3>OllamaChatWrapper<a class="headerlink" href="#ollamachatwrapper" title="Link
 </ul>
 <section id="id3">
 <h4>Prompt Strategy<a class="headerlink" href="#id3" title="Link to this heading"></a></h4>
-<p>Given a list of messages, we will parse each message as follows:</p>
 <ul class="simple">
-<li><p><code class="docutils literal notranslate"><span class="pre">Msg</span></code>:  Fill the <code class="docutils literal notranslate"><span class="pre">role</span></code> and <code class="docutils literal notranslate"><span class="pre">content</span></code> fields directly. If it has an <code class="docutils literal notranslate"><span class="pre">url</span></code>
-field, which refers to an image, we will add it to the message.</p></li>
-<li><p><code class="docutils literal notranslate"><span class="pre">List</span></code>: Parse each element in the list according to the above rules.</p></li>
+<li><p>If the role field of the first input message is <code class="docutils literal notranslate"><span class="pre">&quot;system&quot;</span></code>,
+it will be treated as system prompt and the other messages will consist
+dialogue history in the system message prefixed by “## Dialogue History”.</p></li>
+<li><p>If the <code class="docutils literal notranslate"><span class="pre">url</span></code> attribute of messages is not <code class="docutils literal notranslate"><span class="pre">None</span></code>, we will gather all urls in
+the <code class="docutils literal notranslate"><span class="pre">&quot;images&quot;</span></code> field in the returned dictionary.</p></li>
 </ul>
 <div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">agentscope.models</span> <span class="kn">import</span> <span class="n">OllamaChatWrapper</span>
 
@@ -371,9 +372,11 @@ <h4>Prompt Strategy<a class="headerlink" href="#id3" title="Link to this heading
 </pre></div>
 </div>
 <div class="highlight-bash notranslate"><div class="highlight"><pre><span></span><span class="o">[</span>
-<span class="w">  </span><span class="o">{</span><span class="s2">&quot;role&quot;</span>:<span class="w"> </span><span class="s2">&quot;system&quot;</span>,<span class="w"> </span><span class="s2">&quot;content&quot;</span>:<span class="w"> </span><span class="s2">&quot;You are a helpful assistant&quot;</span><span class="o">}</span>,
-<span class="w">  </span><span class="o">{</span><span class="s2">&quot;role&quot;</span>:<span class="w"> </span><span class="s2">&quot;assistant&quot;</span>,<span class="w"> </span><span class="s2">&quot;content&quot;</span>:<span class="w"> </span><span class="s2">&quot;Hi.&quot;</span><span class="o">}</span>,
-<span class="w">  </span><span class="o">{</span><span class="s2">&quot;role&quot;</span>:<span class="w"> </span><span class="s2">&quot;assistant&quot;</span>,<span class="w"> </span><span class="s2">&quot;content&quot;</span>:<span class="w"> </span><span class="s2">&quot;Nice to meet you!&quot;</span>,<span class="w"> </span><span class="s2">&quot;images&quot;</span>:<span class="w"> </span><span class="o">[</span><span class="s2">&quot;https://example.com/image.jpg&quot;</span><span class="o">]}</span>,
+<span class="w">  </span><span class="o">{</span>
+<span class="w">    </span><span class="s2">&quot;role&quot;</span>:<span class="w"> </span><span class="s2">&quot;system&quot;</span>,
+<span class="w">    </span><span class="s2">&quot;content&quot;</span>:<span class="w"> </span><span class="s2">&quot;You are a helpful assistant\n\n## Dialogue History\nBob: Hi.\nAlice: Nice to meet you!&quot;</span>,
+<span class="w">    </span><span class="s2">&quot;images&quot;</span>:<span class="w"> </span><span class="o">[</span><span class="s2">&quot;https://example.com/image.jpg&quot;</span><span class="o">]</span>
+<span class="w">  </span><span class="o">}</span>,
 <span class="o">]</span>
 </pre></div>
 </div>