chore: import upstream snapshot with attribution
This commit is contained in:
@@ -0,0 +1,103 @@
|
||||
---
|
||||
description:
|
||||
globs:
|
||||
alwaysApply: false
|
||||
---
|
||||
# CopilotKit Agent Development Guide
|
||||
|
||||
## Agent Types
|
||||
|
||||
### Python Agents
|
||||
Python agents are typically in folders named `agent-py/` or `agent/` and use the [sdk-python/](mdc:sdk-python/) package.
|
||||
|
||||
#### Core Python SDK Components
|
||||
- **[copilotkit/agent.py](mdc:sdk-python/copilotkit/agent.py)** - Main agent class
|
||||
- **[copilotkit/action.py](mdc:sdk-python/copilotkit/action.py)** - Action definitions
|
||||
- **[copilotkit/crewai/](mdc:sdk-python/copilotkit/crewai/)** - CrewAI integration
|
||||
- **[copilotkit/integrations/](mdc:sdk-python/copilotkit/integrations/)** - Third-party integrations
|
||||
|
||||
#### Python Agent Examples
|
||||
- **[coagents-starter/agent-py/](mdc:examples/coagents-starter/agent-py/)** - Basic Python agent setup
|
||||
- **[coagents-ai-researcher/agent/](mdc:examples/coagents-ai-researcher/agent/)** - Research agent with web scraping
|
||||
- **[coagents-travel/agent/](mdc:examples/coagents-travel/agent/)** - Travel planning agent
|
||||
- **[langgraph-tutorial-quickstart/agent-py/](mdc:examples/langgraph-tutorial-quickstart/agent-py/)** - LangGraph integration
|
||||
|
||||
### JavaScript Agents
|
||||
JavaScript agents are in folders named `agent-js/` and use the [sdk-js/](mdc:packages/sdk-js/) package.
|
||||
|
||||
#### JavaScript Agent Examples
|
||||
- **[coagents-starter/agent-js/](mdc:examples/coagents-starter/agent-js/)** - Basic JavaScript agent setup
|
||||
- **[coagents-qa-native/agent-js/](mdc:examples/coagents-qa-native/agent-js/)** - Native JavaScript QA agent
|
||||
- **[coagents-routing/agent-js/](mdc:examples/coagents-routing/agent-js/)** - Routing agent
|
||||
|
||||
## Agent Patterns
|
||||
|
||||
### Single Agent Pattern
|
||||
Simple agents that handle specific tasks:
|
||||
- File structure: `agent/main.py` or `agent/index.js`
|
||||
- Direct communication with frontend
|
||||
- Good for focused use cases
|
||||
|
||||
### Multi-Agent (Coagents) Pattern
|
||||
Complex systems with multiple cooperating agents:
|
||||
- Multiple agent files or classes
|
||||
- Shared state management
|
||||
- Agent routing and coordination
|
||||
- Examples: [coagents-routing/](mdc:examples/coagents-routing/), [coagents-shared-state/](mdc:examples/coagents-shared-state/)
|
||||
|
||||
### CrewAI Integration Pattern
|
||||
Using CrewAI for agent orchestration:
|
||||
- **Flows**: [coagents-starter-crewai-flows/](mdc:examples/coagents-starter-crewai-flows/)
|
||||
|
||||
### LangGraph Integration Pattern
|
||||
Using LangGraph for agent state management:
|
||||
- **[langgraph-tutorial-quickstart/](mdc:examples/langgraph-tutorial-quickstart/)** - Basic LangGraph setup
|
||||
- **[langgraph-tutorial-customer-support/](mdc:examples/langgraph-tutorial-customer-support/)** - Customer support bot
|
||||
- Uses [langgraph.json](mdc:sdk-python/langgraph.json) for configuration
|
||||
|
||||
## Agent Setup
|
||||
|
||||
### Python Agent Setup
|
||||
1. Use [requirements-template.txt](mdc:examples/shared/requirements-template.txt) as a starting point
|
||||
2. Configure [pyproject-template.toml](mdc:examples/shared/pyproject-template.toml)
|
||||
3. Import from `copilotkit` package
|
||||
4. Define actions and agent logic
|
||||
|
||||
### JavaScript Agent Setup
|
||||
1. Install `@copilotkit/sdk-js` package
|
||||
2. Configure TypeScript with proper types
|
||||
3. Define actions and agent endpoints
|
||||
4. Set up Express.js server or similar
|
||||
|
||||
## Common Agent Features
|
||||
|
||||
### Actions
|
||||
- Define functions that agents can call
|
||||
- Handle user interactions
|
||||
- Modify application state
|
||||
- Examples in action.py files across examples
|
||||
|
||||
### State Management
|
||||
- Shared state between agents
|
||||
- Persistent state across conversations
|
||||
- Example: [coagents-shared-state/](mdc:examples/coagents-shared-state/)
|
||||
|
||||
### User Input Handling
|
||||
- Wait for user input during agent execution
|
||||
- Interactive workflows
|
||||
- Example: [coagents-wait-user-input/](mdc:examples/coagents-wait-user-input/)
|
||||
|
||||
### Integration Patterns
|
||||
- Vector databases (Pinecone, MongoDB Atlas)
|
||||
- LLM providers (OpenAI, Anthropic)
|
||||
- External APIs and services
|
||||
|
||||
## Development Best Practices
|
||||
1. Start with a simple agent template
|
||||
2. Define clear actions and interfaces
|
||||
3. Handle errors gracefully
|
||||
4. Test with the corresponding UI example
|
||||
5. Use shared utilities from [examples/shared/](mdc:examples/shared/)
|
||||
6. Follow the established patterns in existing examples
|
||||
|
||||
|
||||
Reference in New Issue
Block a user