Streamlit Interview Simulator

A minimal, recruiter-friendly Streamlit app that simulates a job interview using the OpenAI API.
It guides the user through a brief setup, runs a short interview with streamed model responses, supports an immediate Stop control (button or ESC), and then generates structured feedback.

---

Features

• Guided setup: Candidate details + target role/company.
• Interview chat: OpenAI Chat Completions with streaming output.
• Stop control: Button on the UI + ESC key to halt generation mid-stream.
• Early-stop rule: If stopped before any user message, feedback will not be offered (by design).
• Post-interview feedback: Score (1–10) + concise feedback using a fixed format.

---

Quick Start (Local)

1. Clone the repository and move into it.

2. Create and activate a Python 3.10+ environment.

3. Set your OpenAI key.

   Windows (PowerShell):

   setx OPENAI_API_KEY "sk-your-key"

   macOS/Linux (bash/zsh):

   export OPENAI_API_KEY="sk-your-key"

4. Install dependencies:

   pip install -r requirements.txt

5. Run the app:

   python -m streamlit run app.py

The app reads OPENAI_API_KEY from env vars locally. In Streamlit Cloud, set it under Advanced settings → Secrets.

---

Deployment (Streamlit Community Cloud)

1. Push this repo to public GitHub (required by Community Cloud).
2. Go to share.streamlit.io → New app.
3. Select your repo, branch (main), and file path (app.py).
4. Under Advanced settings → Secrets, add your OpenAI key in TOML format:

   OPENAI_API_KEY = "sk-your-key"

5. Deploy. The app will build and provide a shareable URL.

---

Live Demo

The app is deployed on Streamlit Community Cloud and available here:
👉 Streamlit Interview Simulator

---

Demo Screenshots

Setup Screen
Candidate provides personal details, experience, skills, and selects target role/company.

Interview in Action
The candidate introduces themselves, the interviewer responds with a contextual question,
and the red Stop button is visible for early-stop control.

---

Usage Notes (UX)

• The flow is: Setup → Interview → Stop/Complete → Feedback.
• Stop:
  • Click Stop (top-right) or press ESC to halt generation.
  • If Stop is pressed before the first user message, the interview ends and no Feedback option appears.
  • If Stop is pressed mid-answer, streaming stops and the session is marked complete.
• Limits: The interview accepts up to 5 user messages in total.

---

Configuration & Secrets

• Local: environment variable OPENAI_API_KEY.
• Streamlit Cloud: set OPENAI_API_KEY under Advanced settings → Secrets.
• The app fails fast with a clear error if the key is missing.

---

Tech Stack

• Frontend: Streamlit
• LLM: OpenAI Chat Completions API
• Language: Python 3.10+
• State: st.session_state (single-file prototype)

---

Project Structure

streamlit-interview-simulator/
├─ .github/
│  └─ workflows/
│     └─ ci.yml                       # GitHub Actions workflow (CI)
├─ .idea/                              # IDE config (PyCharm)
├─ assets/
│  ├─ demo_conversation.png           # 1378×1003 conversation view (README demo)
│  ├─ demo_setup.png                  # 877×1134 setup view (README demo)
│  └─ interview_social.png            # 1280×640 conversation view (GitHub/LinkedIn social preview)
├─ tests/                               # Smoke tests
│  └─ test_import.py
├─ .gitignore                           # Keeps secrets & caches out of version control
├─ LICENSE                              # MIT license
├─ README.md                            # Project documentation
├─ app.py                               # Streamlit app (main entrypoint)
└─ requirements.txt                     # Minimal, pinned dependencies

---

Requirements

Minimal set for fast builds:

streamlit>=1.37
openai>=1.42
streamlit-js-eval>=0.1.7

Add only what you actually import. Keeping this minimal improves reliability and cold-start time.

---

Known Limitations / Future Considerations

• This is a lightweight prototype (no database, no authentication).
• Feedback is model-generated and non-deterministic.
• If you add embeddings, RAG, or vector search later, update requirements.txt accordingly and document the changes.

---

---

Acknowledgments

Originally developed during The AI Engineer Course 2025 (Udemy).
Now actively maintained as a public, open-source project.

---

License

MIT — see LICENSE