smolagents/pr_2321/en/conceptual_guides/react.html

<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;How do multi-step agents work?&quot;,&quot;local&quot;:&quot;how-do-multi-step-agents-work&quot;,&quot;sections&quot;:[],&quot;depth&quot;:1}">
		<link href="/docs/smolagents/pr_2321/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/entry/start.fe75d00e.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/chunks/scheduler.545a7209.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/chunks/singletons.d667ceb3.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/chunks/index.1d3c4622.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/chunks/paths.cd5ec750.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/entry/app.bd903b71.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/chunks/preload-helper.bbc9581e.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/chunks/index.5e5f0156.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/nodes/0.119a3025.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/chunks/each.e59479a4.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/nodes/3.27494a38.js">
		<link rel="modulepreload" href="/docs/smolagents/pr_2321/en/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.b21a3620.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;How do multi-step agents work?&quot;,&quot;local&quot;:&quot;how-do-multi-step-agents-work&quot;,&quot;sections&quot;:[],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END -->      <p></p>  <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"><div class="inline-flex rounded-md max-sm:rounded-sm"><button class="inline-flex items-center gap-1 h-7 max-sm:h-7 px-2 max-sm:px-1.5 text-sm font-medium text-gray-800 border border-r-0 rounded-l-md max-sm:rounded-l-sm border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-live="polite"><span class="inline-flex items-center justify-center rounded-md p-0.5 max-sm:p-0 hover:text-gray-800 dark:hover:text-gray-200"><svg class="sm:size-3.5 size-3" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg></span> <span>Copy page</span></button> <button class="inline-flex items-center justify-center w-6 max-sm:w-5 h-7 max-sm:h-7 disabled:pointer-events-none text-sm text-gray-500 hover:text-gray-700 dark:hover:text-white rounded-r-md max-sm:rounded-r-sm border border-l transition border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-haspopup="menu" aria-expanded="false" aria-label="Open copy menu"><svg class="transition-transform text-gray-400 overflow-visible sm:size-3.5 size-3 rotate-0" width="1em" height="1em" viewBox="0 0 12 7" fill="none" xmlns="http://www.w3.org/2000/svg"><path d="M1 1L6 6L11 1" stroke="currentColor"></path></svg></button></div> </div>  <h1 class="relative group"><a id="how-do-multi-step-agents-work" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#how-do-multi-step-agents-work"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>How do multi-step agents work?</span></h1> <p data-svelte-h="svelte-nsa9lf">The ReAct framework (<a href="https://huggingface.co/papers/2210.03629" rel="nofollow">Yao et al., 2022</a>) is currently the main approach to building agents.</p> <p data-svelte-h="svelte-2021n9">The name is based on the concatenation of two words, “Reason” and “Act.” Indeed, agents following this architecture will solve their task in as many steps as needed, each step consisting of a Reasoning step, then an Action step where it formulates tool calls that will bring it closer to solving the task at hand.</p> <p data-svelte-h="svelte-zl5ssw">All agents in <code>smolagents</code> are based on singular <code>MultiStepAgent</code> class, which is an abstraction of ReAct framework.</p> <p data-svelte-h="svelte-1chrp8z">On a basic level, this class performs actions on a cycle of following steps, where existing variables and knowledge is incorporated into the agent logs like below:</p> <p data-svelte-h="svelte-3wnwgb">Initialization: the system prompt is stored in a <code>SystemPromptStep</code>, and the user query is logged into a <code>TaskStep</code> .</p> <p data-svelte-h="svelte-1qs057z">While loop (ReAct loop):</p> <ul data-svelte-h="svelte-1xb8sbm"><li>Use <code>agent.write_memory_to_messages()</code> to write the agent logs into a list of LLM-readable <a href="https://huggingface.co/docs/transformers/en/chat_templating" rel="nofollow">chat messages</a>.</li> <li>Send these messages to a <code>Model</code> object to get its completion. Parse the completion to get the action (a JSON blob for <code>ToolCallingAgent</code>, a code snippet for <code>CodeAgent</code>).</li> <li>Execute the action and logs result into memory (an <code>ActionStep</code>).</li> <li>At the end of each step, we run all callback functions defined in <code>agent.step_callbacks</code> .</li></ul> <p data-svelte-h="svelte-2uzzav">Optionally, when planning is activated, a plan can be periodically revised and stored in a <code>PlanningStep</code> . This includes feeding facts about the task at hand to the memory.</p> <p data-svelte-h="svelte-m3d6g5">For a <code>CodeAgent</code>, it looks like the figure below.</p> <div class="flex justify-center" data-svelte-h="svelte-1hb194n"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/smolagents/codeagent_docs.png"></div> <p data-svelte-h="svelte-b9ve8f">Here is a video overview of how that works:</p> <div class="flex justify-center" data-svelte-h="svelte-4ra9xq"><img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/Agent_ManimCE.gif"> <img class="hidden dark:block" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/Agent_ManimCE.gif"></div> <p data-svelte-h="svelte-1r9wlyn">We implement two versions of agents:</p> <ul data-svelte-h="svelte-1fqh22f"><li><a href="/docs/smolagents/pr_2321/en/reference/agents#smolagents.CodeAgent">CodeAgent</a> generates its tool calls as Python code snippets.</li> <li><a href="/docs/smolagents/pr_2321/en/reference/agents#smolagents.ToolCallingAgent">ToolCallingAgent</a> writes its tool calls as JSON, as is common in many frameworks. Depending on your needs, either approach can be used. For instance, web browsing often requires waiting after each page interaction, so JSON tool calls can fit well.</li></ul> <blockquote class="tip" data-svelte-h="svelte-1q4ake2"><p>Read <a href="https://huggingface.co/blog/open-source-llms-as-agents" rel="nofollow">Open-source LLMs as LangChain Agents</a> blog post to learn more about multi-step agents.</p></blockquote> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/smolagents/blob/main/docs/source/en/conceptual_guides/react.md" target="_blank"><svg class="mr-1" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M31,16l-7,7l-1.41-1.41L28.17,16l-5.58-5.59L24,9l7,7z"></path><path d="M1,16l7-7l1.41,1.41L3.83,16l5.58,5.59L8,23l-7-7z"></path><path d="M12.419,25.484L17.639,6.552l1.932,0.518L14.351,26.002z"></path></svg> <span data-svelte-h="svelte-zjs2n5"><span class="underline">Update</span> on GitHub</span></a>  <p></p> 
			
			<script>
				{
					__sveltekit_t2vx7 = {
						assets: "/docs/smolagents/pr_2321/en",
						base: "/docs/smolagents/pr_2321/en",
						env: {}
					};

					const element = document.currentScript.parentElement;

					const data = [null,null];

					Promise.all([
						import("/docs/smolagents/pr_2321/en/_app/immutable/entry/start.fe75d00e.js"),
						import("/docs/smolagents/pr_2321/en/_app/immutable/entry/app.bd903b71.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 3],
							data,
							form: null,
							error: null
						});
					});
				}
			</script>