Skip to main content

Next Steps

Now that you've built your first flow, here's how to level up:

Immediate Next Steps​

1. Add AI Model Integration​

Enhance your echo flow with AI capabilities:

{
  "init_params": {
    "topic": "artificial intelligence"
  },
  "step_configs": {
    "generate": {
      "activity": "litellm_chat",
      "model": "gpt-4o-mini",
      "system_prompt": "You are a helpful assistant.",
      "user_prompt": "Write a haiku about {{ init_params.topic }}"
    }
  },
  "steps": ["generate"]
}

2. Chain Multiple Steps​

Connect step outputs to create pipelines:

{
  "init_params": {
    "text": "The quick brown fox jumps over the lazy dog"
  },
  "step_configs": {
    "analyze": {
      "activity": "litellm_chat",
      "model": "gpt-4o-mini",
      "user_prompt": "Count the words in this text: {{ init_params.text }}"
    },
    "summarize": {
      "activity": "litellm_chat",
      "model": "gpt-4o-mini",
      "user_prompt": "Based on: {{ analyze.outputs.content }}\n\nProvide a one-sentence summary."
    }
  },
  "steps": ["analyze", "summarize"]
}

3. Generate Images from Text​

Create your first text-to-image workflow:

{
  "init_params": {
    "prompt": "A serene lake with mountains at sunset"
  },
  "step_configs": {
    "generate": {
      "activity": "replicate_text2image",
      "model": "black-forest-labs/flux-schnell",
      "prompt_path": "init_params.prompt",
      "output_format": "jpg"
    }
  },
  "steps": ["generate"]
}

4. Run Workflows via API​

Integrate Jetty into your applications:

# Set your API token
export JETTY_API_TOKEN="your_token_here"

# Run synchronously (wait for result)
curl -X POST "https://flows-api.jetty.io/api/v1/run-sync/your-collection/your-task" \
  -H "Authorization: Bearer $JETTY_API_TOKEN" \
  -F 'init_params={"topic": "machine learning"}'

# Run asynchronously (get workflow_id immediately)
curl -X POST "https://flows-api.jetty.io/api/v1/run/your-collection/your-task" \
  -H "Authorization: Bearer $JETTY_API_TOKEN" \
  -F 'init_params={"topic": "machine learning"}'

Learning Paths​

Path A: AI Application Developer​

  1. LiteLLM Integration - Multi-provider LLM access
  2. Image Generation - Text-to-image workflows
  3. Document Processing - End-to-end pipeline

Path B: Evaluation Engineer​

  1. Simple Judge - LLM-as-Judge
  2. Control Flow - Parallel processing
  3. Evaluation Pipelines - Model comparison

Path C: ML Operations​

  1. API Reference - Integration patterns
  2. Integrations - Secure API keys and model configurations
  3. Troubleshooting - Debug workflows

Path D: Agent & MCP User​

  1. Using Jetty with Agents - Overview and setup options
  2. Agent Recipes - Copy-paste workflows for your agent
  3. Common Gotchas - Parameter mismatches and API quirks