Skip to main content

What's Next

You now understand how to build AI applications with Saiki! Here's how to keep growing and take your applications to the next level.

Immediate Next Steps​

1. Experiment with Different Providers​

Try different LLM providers to find what works best for your use case:

# Fast and cost-effective
llm:
  provider: groq
  model: llama-3.3-70b-versatile
  apiKey: $GROQ_API_KEY

# Large context for complex tasks
llm:
  provider: anthropic
  model: claude-3-5-sonnet-20240620
  apiKey: $ANTHROPIC_API_KEY

# Multimodal capabilities
llm:
  provider: google
  model: gemini-2.0-flash
  apiKey: $GOOGLE_GENERATIVE_AI_API_KEY

Learn more: LLM Providers Guide

2. Add More Tools​

Expand your agents' capabilities with additional MCP servers:

mcpServers:
  filesystem:
    type: stdio
    command: npx
    args: ["-y", "@modelcontextprotocol/server-filesystem", "."]
  
  puppeteer:
    type: stdio
    command: npx
    args: ["-y", "@truffle-ai/puppeteer-server"]
  
  # Add database access, API calls, and more!

Learn more: MCP Server Configuration

3. Deploy Your Application​

Take your application to production:

  • Development: Run locally with npm start
  • Production: Use Docker containers with proper monitoring

Learn more: Deployment Guide

Going Deeper​

Join Our Community​

Connect with other developers building AI applications:

Explore Examples​

See what others have built and get inspiration:

  • Examples Repository - Working code you can run
  • Community showcase in our Discord
  • Blog posts and tutorials from the community

Read the Advanced Docs​

Dive deeper into Saiki's capabilities:

Pro Tips for Saiki Success​

✅ Use Specific Prompts​

# ❌ Vague
systemPrompt: "You are a helpful assistant."

# ✅ Specific
systemPrompt: |
  You are a code review assistant specializing in security.
  Always check for SQL injection, XSS, and authentication issues.
  Provide specific line numbers and fix suggestions.

✅ Handle Errors Gracefully in Your Application​

When interacting with Saiki agents:

try {
  const response = await agent.run(message); // Assuming 'agent' is your Saiki agent instance
  return response;
} catch (error) {
  // Log the error for debugging
  console.error('Saiki Agent error:', error);
  
  // Return helpful message to user
  return "I'm having trouble processing that request with the AI agent. Could you try rephrasing it?";
}

✅ Optimize for Your Use Case with Saiki Configurations​

For Speed:

llm:
  provider: groq
  model: llama-3.3-70b-versatile
  providerOptions:
    temperature: 0.1  # More deterministic

For Quality:

llm:
  provider: anthropic
  model: claude-3-5-sonnet-20240620
  providerOptions:
    temperature: 0.3  # Balanced creativity

For Cost:

llm:
  provider: openai
  model: gpt-4.1-mini  # Cheaper than full GPT-4

Common Saiki Patterns to Explore​

1. Agent Specialization​

Create different Saiki agents for different tasks:

  • Code Reviewer - Focus on security and best practices
  • Documentation Writer - Generate clear, comprehensive docs
  • Data Analyst - Analyze and visualize data
  • Customer Support - Handle user questions

2. Workflow Automation with Saiki​

Chain multiple Saiki agents or calls together:

  1. Analyzer Agent - Understand the problem
  2. Planner Agent - Create a solution plan based on Analyzer's output
  3. Executor Agent - Implement the solution using tools
  4. Reviewer Agent - Validate the results

Scaling Your Saiki Applications​

Performance Considerations for Saiki​

  • Use faster models configured in Saiki for latency-sensitive tasks.
  • Cache LLM responses where appropriate to reduce redundant API calls, especially for frequently requested information.
  • Optimize tool usage within your agents to ensure they perform efficiently.

Cost Optimization with Saiki​

  • Choose appropriate LLM models in your Saiki configuration for each task's complexity and budget.
  • Optimize prompts for your Saiki agents to reduce token usage.
  • Cache results from Saiki agents or LLM calls if the same information is requested repeatedly.

Security Best Practices for Saiki​

  • Never hardcode API keys in your Saiki configurations. Use environment variables (e.g., $ANTHROPIC_API_KEY).
  • Validate and sanitize user inputs before passing them to Saiki agents to prevent prompt injection or misuse of tools.
  • Restrict tool permissions for MCP servers to only what is necessary for the agent's function.

Stay Updated​

Follow Our Progress​

Community Events​

  • Office hours in Discord
  • Community showcases of projects
  • Workshops on advanced techniques
  • Hackathons with prizes and recognition

Ready to Ship?​

You have everything you need to build amazing AI applications with Saiki:

  • ✅ Understanding of how Saiki works
  • ✅ Practical experience building applications
  • ✅ Advanced patterns for production use with Saiki
  • ✅ Community support when you need help
  • ✅ Resources for continued learning

Final Thoughts​

The AI landscape is evolving rapidly, and Saiki is designed to evolve with it. By focusing on configuration over code, you can:

  • Adapt quickly to new models and providers via Saiki's configuration.
  • Experiment freely with different Saiki setups without rewriting entire applications.
  • Scale efficiently as your needs grow by optimizing Saiki configurations and tool usage.
  • Build confidently with Saiki using proven patterns.

The only limit is your imagination! 🚀

What will you build next with Saiki?

Share your creations in our Discord community - we'd love to see what you create!