Skip to content

HOW-TO-RUN-RASA.md

Latest commit

 

History

History
264 lines (180 loc) · 6.16 KB

HOW-TO-RUN-RASA.md

File metadata and controls

264 lines (180 loc) · 6.16 KB

How to Run Rasa – Stress Detection Chatbot

Step-by-step guide to set up and run the Rasa assistant: Python version, training, actions server, and core API.

Python version

Requirement Version
Recommended Python 3.9
Minimum Python 3.8
Rasa 3.6.21
Rasa SDK 3.6.2

Rasa 3.6.x is tested with Python 3.8–3.10. This project uses Python 3.9 (e.g. C:\Program Files\Python39\python.exe on Windows).

Check your version:

python --version
# or
py -3.9 --version

1. Project directory

All Rasa commands are run from the Django project folder where config.yml, domain.yml, and data/ live:

cd "D:\surag_projects\Stress Detection for IT Professionals\STRESS-DETECTION-FOR-IT-PROFESSIONALS\strees_dection"

(Use your actual path if different.)

2. Virtual environment and install

Create and activate a venv (optional but recommended):

# Create
python -m venv venv

# Activate – Windows PowerShell
.\venv\Scripts\Activate.ps1

# Activate – Windows CMD
venv\Scripts\activate.bat

# Activate – Linux/macOS
source venv/bin/activate

Install Rasa and the SDK:

pip install --upgrade pip
pip install rasa==3.6.21 rasa-sdk==3.6.2

Or install from the project’s requirements (includes Django and other deps):

pip install -r requirements.txt

3. Rasa train

Train the NLU and dialogue model from data/, config.yml, and domain.yml:

# From strees_dection directory, with venv activated
rasa train
  • Reads: config.yml, domain.yml, data/nlu.yml, data/stories.yml, data/rules.yml.
  • Writes: a timestamped model under models/ (e.g. models/20250425-123419-neat-salmon.tar.gz).

Use a specific config or domain if you have multiple:

rasa train --config config.yml --domain domain.yml

Train with a fixed model name (optional):

rasa train --fixed-model-name stress_bot

After changing intents, stories, rules, or config, run rasa train again, then restart the Rasa server so it loads the new model.

4. Rasa actions server

Custom actions (e.g. mental health assessment) run in a separate process. The core server calls them via endpoints.yml.

Start the actions server:

rasa run actions --port 5055
  • Default port: 5055
  • Serves: actions/actions.py (and any other modules in actions/).

Windows – batch file (from strees_dection):

start-rasa-actions.bat

Or explicitly with Python 3.9:

& "C:\Program Files\Python39\python.exe" -m rasa run actions --port 5055

Health check:

curl http://localhost:5055/health

endpoints.yml must point the core server to this URL, e.g.:

action_endpoint:
  url: "http://localhost:5055/webhook"

5. Rasa core server (API)

Run the main Rasa server (NLU + dialogue, REST API):

rasa run --enable-api --cors "*" --port 5005
  • Port: 5005
  • --enable-api: enables REST and webhook endpoints.
  • --cors "*": allows the Django frontend to call the API.

Windows – batch file (from strees_dection):

start-rasa-core.bat

Or with Python 3.9:

& "C:\Program Files\Python39\python.exe" -m rasa run --enable-api --cors "*" --port 5005

Use a specific model:

rasa run --enable-api --cors "*" --port 5005 --model models/20250425-123419-neat-salmon.tar.gz

Quick API check:

curl http://localhost:5005/

6. Run order (full stack)

  1. Terminal 1 – Actions (so the core can call custom actions):

    cd strees_dection
rasa run actions --port 5055

  2. Terminal 2 – Rasa core:

    cd strees_dection
rasa run --enable-api --cors "*" --port 5005

  3. Terminal 3 – Django (optional, for the web UI):

    cd strees_dection
python manage.py runserver

The chatbot in the app uses: http://localhost:5005/webhooks/rest/webhook.

7. Useful Rasa commands

Command Purpose
rasa train Train model from data/, config.yml, domain.yml
rasa run Start core server (use --enable-api --cors "*" for API)
rasa run actions Start custom actions server
rasa shell Chat in the terminal (uses latest model)
rasa test Run tests (e.g. tests/test_stories.yml)
rasa data validate Check data/ and domain.yml for errors
rasa interactive Collect new training data interactively

8. Project layout (Rasa)

strees_dection/
├── config.yml          # NLU pipeline & policies
├── domain.yml          # Intents, entities, slots, responses, actions
├── endpoints.yml       # action_endpoint URL (e.g. http://localhost:5055/webhook)
├── credentials.yml     # Optional channel config
├── data/
│   ├── nlu.yml         # Intents & entities
│   ├── stories.yml     # Conversation flows
│   └── rules.yml       # Rules (e.g. one-shot responses)
├── actions/
│   └── actions.py      # Custom action code (e.g. MentalHealthAssessment)
├── models/             # Trained models (*.tar.gz)
├── start-rasa-core.bat
└── start-rasa-actions.bat

9. Troubleshooting

“Module not found” when running Rasa

  • Use the same Python that has rasa and rasa-sdk installed (e.g. python -m rasa ... or full path to python.exe).
  • If you use a global Python 3.9 for Rasa, run:
    "C:\Program Files\Python39\python.exe" -m rasa run ...
    and the same for rasa run actions.

Actions not running / 404 from core

  • Start the actions server first (rasa run actions --port 5055).
  • In endpoints.yml, set action_endpoint.url to http://localhost:5055/webhook (or the host/port you use).

Port 5005 or 5055 in use

  • Windows: netstat -ano | findstr :5005 then taskkill /PID <pid> /F.
  • Or use different ports: rasa run --port 5006, rasa run actions --port 5056, and update endpoints.yml and the frontend URL.

Model not updating

  • Run rasa train after changing data/ or config.yml.
  • Restart the core server so it loads the new model from models/.

Last updated: March 2025