➕ Create

An AI Pipeline is defined in a simple YAML file.

Check out different AI Pipeline examples.

Basic Example

A basic text generation AI Pipeline that uses the default Prompt Template, has a predefined System Prompt (instructs the AI Model on how to generally act) and connects to the default built-in Text AI Model (LLM)

kind: floom/pipeline/1.2

pipeline:
  name: my-first-pipeline

prompt:  
  template: 
    package: floom/prompt/template/default
    system: "You are a nice assistant"

You can learn more about AI Pipeline objects.

Save this pipeline configuration YAML to 'my-first-pipeline.yml'.

Production Example

A robust production AI Pipeline scenario, including Plugins: Caching, Cost Management, Prompt Validation and Response Validation executed on OpenAI GPT-3.5.

The AI Pipeline incorporates the "tutorial.pdf" file as Context (RAG). Consequently, the AI Model will be equipped with the knowledge contained in "tutorial.pdf".

kind: 'floom/pipeline/1.2'

pipeline:
  name: my-first-pipeline
  
  # Use 🧠 OpenAI-GPT 3.5
  model:
    - package: floom/model/connectors/openai
      model: gpt-3.5
      api-key: sk-1234567890abcdefg=
      
  prompt:  
    template: 
      package: floom/prompt/templates/default
      system: "You are helpful assistant" 
    
    # 📄 tutorial.pdf as context
    context:
      - package: floom/prompt/context/pdf
        path: /etc/tutorial.pdf
        
    # ❌ Filter Profanity   
    validation:
      - package: floom/plugins/bad-words-filter
        disallow: ['profanity']
        
  response:
    # ✔️ Response should be 3 sentences max, in English (US) 
    format:
     - package: floom/response/formatter 
       type: text
       language: en-us
       max-sentences: 3
        
    # ❌ Filter Profanity & Negativity
    validation:
     - package: floom/plugins/bad-words-filter 
       disallow: ['profanity', 'negativity']
       language: en-us
  
  global:
    # 💵 Cost limit
    - package: floom/plugins/cost-management 
      limits:
        user:
         - day: 1000 # max tokens per day
        pipeline:
         - month: 600000 # max tokens per month
			
    # ⚡Caching
    - package: floom/plugins/cache 
      cache-type: memory