Patterns

src/content/docs/agents/patterns.mdx

@@ -0,0 +1,88 @@
+---
+pcx_content_type: concept
+title: Patterns
+sidebar:
+  order: 3
+head: []
+---
+
+import { GitHubCode } from "~/components";
+
+Common patterns for implementing AI agents, based on [Anthropic's research on agent patterns](https://www.anthropic.com/research/building-effective-agents).
+
+Code samples use the [AI SDK](https://sdk.vercel.ai/docs/foundations/agents), running in Cloudflare's Durable Objects.
+
+## Prompt Chaining
+
+Decomposes tasks into a sequence of steps, where each LLM call processes the output of the previous one.
+
+![Figure 1: Prompt Chaining](~/assets/images/agents/01-sequential.png "Figure 1: Prompt Chaining")
+
+[**Interactive Demo**](https://patterns.agents.workers.dev)
+
+<GitHubCode
+	repo="cloudflare/agents"
+	file="examples/anthropic-patterns/src/flows/01-sequential.txt"
+	commit="3b8959063c17d53e27a3040670d9cbf8a95644e1"
+	lang="ts"
+/>
+
+## Routing
+
+Classifies input and directs it to specialized followup tasks, allowing for separation of concerns.
+
+![Figure 2: Routing](~/assets/images/agents/02-routing.png "Figure 2: Routing")
+
+[**Interactive Demo**](https://patterns.agents.workers.dev)
+
+<GitHubCode
+	repo="cloudflare/agents"
+	file="examples/anthropic-patterns/src/flows/02-routing.txt"
+	commit="3b8959063c17d53e27a3040670d9cbf8a95644e1"
+	lang="ts"
+/>
+
+## Parallelization
+
+Enables simultaneous task processing through sectioning or voting mechanisms.
+
+![Figure 3: Parallelization](~/assets/images/agents/03-parallel.png "Figure 3: Parallelization")
+
+[**Interactive Demo**](https://patterns.agents.workers.dev)
+
+<GitHubCode
+	repo="cloudflare/agents"
+	file="examples/anthropic-patterns/src/flows/03-parallel.txt"
+	commit="3b8959063c17d53e27a3040670d9cbf8a95644e1"
+	lang="ts"
+/>
+
+## Orchestrator-Workers
+
+A central LLM dynamically breaks down tasks, delegates to worker LLMs, and synthesizes results.
+
+![Figure 4: Orchestrator Workers](~/assets/images/agents/04-orchestrator.png "Figure 4: Orchestrator Workers")
+
+[**Interactive Demo**](https://patterns.agents.workers.dev)
+
+<GitHubCode
+	repo="cloudflare/agents"
+	file="examples/anthropic-patterns/src/flows/04-orchestrator.txt"
+	commit="3b8959063c17d53e27a3040670d9cbf8a95644e1"
+	lang="ts"
+/>
+
+## Evaluator-Optimizer
+
+One LLM generates responses while another provides evaluation and feedback in a loop.
+
+![Figure 5: Evaluator-Optimizer](~/assets/images/agents/05-evaluator.png "Figure 5: Evaluator-Optimizer")
+
+[**Interactive Demo**](https://patterns.agents.workers.dev)
+
+<GitHubCode
+	repo="cloudflare/agents"
+	file="examples/anthropic-patterns/src/flows/05-evaluator.txt"
+	commit="3b8959063c17d53e27a3040670d9cbf8a95644e1"
+	lang="ts"
+/>