JobQuest AI

A Rasa-based conversational assistant that answers candidate questions and walks them through applying for a job.

JobQuest AI is an intent-based chatbot built on the Rasa framework. It recognizes what a candidate is asking (job openings, application status, company info, salary/benefits, requirements, ...), walks them through submitting an application, and hands structured applicant details off to a custom action for storage. A small Streamlit page provides a chat UI in front of the Rasa server.

How it works

• NLU (data/nlu.yml): training examples for 13 intents, from greet and job_openings through apply_job, inquire_salary, and inquire_requirements.
• Dialogue (data/stories.yml, data/rules.yml, domain.yml): maps recognized intents to scripted responses and conversation flow.
• Custom action (actions/actions.py): ActionJobApplication reads the collected applicant details out of the conversation and appends them to job_applications.csv.
• Frontend (app.py): a one-page Streamlit app that posts user messages to the Rasa server and renders the bot's replies.

flowchart LR
    UI[Streamlit app.py] -->|POST message| RASA[Rasa server\nNLU + dialogue]
    RASA -->|intent match| ACTIONS[Action server\nactions.py]
    ACTIONS -->|append| CSV[(job_applications.csv)]
    RASA -->|reply| UI

Loading

Running it

You'll need Python 3.8+ and the Rasa stack:

pip install rasa rasa-sdk streamlit requests

# train an NLU/dialogue model from data/ + domain.yml
rasa train

# terminal 1: the action server (runs actions.py)
rasa run actions --port 5056

# terminal 2: the Rasa server
rasa shell --port 5006 --endpoints endpoints.yml

# terminal 3: the Streamlit UI
streamlit run app.py

rasa train produces a .tar.gz model artifact under models/; that directory is gitignored since models are build output, not source.

Repository layout

data/         NLU training examples, stories, and rules
actions/      custom action server (writes applications to CSV)
domain.yml    intents, slots, and response templates
app.py        Streamlit chat frontend
tests/        story-based conversation tests (rasa test)

Roadmap

• Wire the chatbot up to a real job-postings API instead of the hard-coded openings list in domain.yml.
• Replace the CSV sink with a real datastore.
• Multilingual intent recognition.