Rasa - Conversational AI Framework Cheatsheet

Rasa is an open-source framework for building contextual conversational assistants. Unlike prompt-only chatbots, Rasa gives you explicit control over dialogue: you define intents, entities, and flows, and the assistant follows business logic you specify rather than whatever the model improvises. Everything runs on your own infrastructure, which matters for regulated domains. Modern Rasa (CALM) blends LLM-based understanding with deterministic flow execution — the model interprets, your flows decide.

Installation

Method	Command
pip	`pip install rasa`
New project	`rasa init`
Requirements	Python 3.9–3.11 (check current support)
Verify	`rasa --version`

Project Structure

Path	Contains
`domain.yml`	Intents, entities, slots, responses, actions
`data/nlu.yml`	Training examples per intent
`data/stories.yml`	Example conversation paths
`data/rules.yml`	Deterministic rules
`data/flows.yml`	CALM business-logic flows
`config.yml`	NLU/policy pipeline
`actions/actions.py`	Custom Python actions

Core CLI

Command	Description
`rasa init`	Scaffold a new assistant
`rasa train`	Train the model
`rasa shell`	Chat with it in the terminal
`rasa run`	Start the server
`rasa run actions`	Start the custom action server
`rasa test`	Run tests against test stories
`rasa interactive`	Interactive training/correction
`rasa data validate`	Check data consistency

Defining Intents & Responses

# domain.yml
intents:
  - greet
  - check_balance

responses:
  utter_greet:
    - text: "Hi! How can I help?"

slots:
  account_id:
    type: text
    mappings:
      - type: from_entity
        entity: account_id

actions:
  - action_fetch_balance

# data/nlu.yml
nlu:
  - intent: check_balance
    examples: |
      - what's my balance
      - how much money do I have
      - show account [12345](account_id)

Flows (CALM)

# data/flows.yml
flows:
  check_balance:
    description: Look up the user's account balance
    steps:
      - collect: account_id
      - action: action_fetch_balance
      - action: utter_balance

Step type	Does
`collect`	Ask for and fill a slot
`action`	Run a custom or utter action
`link`	Jump to another flow
`set_slots`	Assign slot values
`next` with `if`	Conditional branching

Flows are deterministic: the LLM decides which flow the user wants, but the flow itself executes exactly as written.

Custom Actions

from rasa_sdk import Action, Tracker
from rasa_sdk.executor import CollectingDispatcher

class ActionFetchBalance(Action):
    def name(self) -> str:
        return "action_fetch_balance"

    def run(self, dispatcher: CollectingDispatcher, tracker: Tracker, domain):
        account = tracker.get_slot("account_id")
        balance = lookup_balance(account)
        dispatcher.utter_message(text=f"Your balance is ${balance}")
        return []

Run with rasa run actions (default port 5055).

Testing

rasa test                 # end-to-end tests
rasa test nlu --cross-validation
rasa data validate        # catch domain/data mismatches

Test type	Checks
NLU tests	Intent/entity accuracy
Story tests	Full conversation paths
Cross-validation	Generalization of NLU

Channels & Deployment

Channel	Note
REST	Default HTTP endpoint
Web widget	Embeddable chat
Slack / Teams / Telegram	Built-in connectors
Voice	Via telephony integrations
Deployment	Docker, Kubernetes/Helm, self-hosted

Rasa vs LLM-Only Chatbots

Aspect	Rasa	Prompt-only LLM bot
Control	Explicit flows/business logic	Emergent, hard to constrain
Predictability	High	Variable
Compliance	Auditable paths	Difficult
Hosting	Fully self-hosted	Usually API-dependent
Best for	Regulated, task-oriented assistants	Open-ended conversation

For open-ended agents see LangGraph or Mastra; Rasa’s strength is deterministic, auditable task flows.

Rasa - Conversational AI Framework Cheatsheet

Rasa - Conversational AI Framework Cheatsheet

Installation

Project Structure

Core CLI

Defining Intents & Responses

Flows (CALM)

Custom Actions

Testing

Channels & Deployment

Rasa vs LLM-Only Chatbots

Resources