Add research example

Author: rlancemartin

examples/research/README.md

@@ -1,81 +1,30 @@
-# Swarm Research
+# Swarm Researcher Example
 
-#TODO: Handoff between planner and researcher
+A two-phase multi-agent system that demonstrates an effective collaborative approach to planning and research tasks. This example showcases a pattern used in many deep research systems:
 
-## Quickstart
-```
-uvx --refresh --from "langgraph-cli[inmem]" --with-editable . --python 3.11 langgraph dev
-```
-
-## Search Tools 
- 
-### Playwright Browser Tool
-
-[Playwright](https://playwright.dev/) offers several advantages over API-based solutions:
-
-  1. Full browser access - can interact with entire websites, not just search results
-  2. Can navigate complex pages, fill forms, click buttons like a human user
-  3. Handles JavaScript-rendered content that APIs can't access
-  4. Maintains session state across multiple interactions
-  5. Better evades anti-scraping measures by mimicking legitimate browser traffic
-  6. Can extract specific data elements from pages with precision
-  7. Works well for research requiring deep site exploration
-
-## Troubleshooting
+1. **Planning Phase**: A dedicated planner agent clarifies requirements, reads documentation, and develops a structured approach
+2. **Research Phase**: A researcher agent implements the solution based on the planner's guidance
 
-### Playwright MCP Server
+## How It Works
 
-Some issues with the Playwright MCP Server are related to the npx cache. 
+- The system starts with the **planner agent** that:
+  - Analyzes the user's request
+  - Reads relevant documentation
+  - Asks clarifying questions to refine scope
+  - Creates a structured plan with clear objectives
+  - Identifies the most relevant resources for implementation
+  - Hands off to the researcher agent
 
-Clear the npx cache:
-```
-rm -rf ~/.npm/_npx/9833c18b2d85bc59
-```
+- The **researcher agent** then:
+  - Follows the structured plan from the planner
+  - Reads the recommended documentation sources
+  - Implements the solution to satisfy all requirements
+  - Can request additional planning if needed
 
-Create a minimal package.json:
-```
-cat > package.json << EOF
-{
-  "name": "zod-fix",
-  "private": true,
-  "dependencies": {
-    "zod": "3.22.4",
-    "@playwright/mcp": "latest"
-  }
-}
-EOF
-```
-Install the dependencies:
-```
-npm install
-```
+This pattern demonstrates how breaking complex tasks into planning and execution phases can lead to more thorough, well-researched outcomes.
 
-Create the MCP server:
-```
-npx @playwright/mcp
-mkdir -p ./node_modules/zod/lib/locales
-echo 'export const errorMap = () => ({ message: "" });' > ./node_modules/zod/lib/locales/en.js
-``` 
-
-Run server in isolation, which will use `stdio` transport:
-```
-npx @playwright/mcp
-```
-
-Test with MCP inspector:
-```
-npx @modelcontextprotocol/inspector
-```
-
-Supply the path to npx directly in your MCP configuration: 
-```
-{
-  "mcpServers": {
-    "playwright": {
-      "command": "npx",
-      "args": ["@playwright/mcp"],
-    }
-  }
-}
-```
+## Quickstart
 
+```bash
+uvx --refresh --from "langgraph-cli[inmem]" --with-editable . --python 3.11 langgraph dev
+```

examples/research/langgraph.json

@@ -1,6 +1,6 @@
 {
     "dependencies": ["."],
     "graphs": {
-      "agent": "./src/agent/agent.py:make_graph"
+      "agent": "./src/agent/agent.py:app"
     }
   }

examples/research/pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "swarm-researcher"
 version = "0.0.1"
-description = "Simple multi-agent system for research."
+description = "Simple multi-agent example for planning and research."
 authors = [
     { name = "Lance Martin" }
 ]
@@ -15,6 +15,7 @@ dependencies = [
     "langchain-mcp-adapters>=0.0.5",
     "langgraph>=0.3.21",
     "langgraph-swarm>=0.0.7",
+    "langchain>=0.3.21",
 ]
 
 [project.optional-dependencies]