InTurn.AI — Smart Internship Assistant

An AI-powered internship tracking assistant for university students.
Includes three tools: InternTrack (hours tracking), MentorBridge (workplace communication), and Report Writer (professional activity reports).

Latest Updates

• InternTrack now supports signed hour updates from chat input:
  • +8, + 8 hours, -2, - 2 hrs
• Hour updates are saved to backend config automatically.
• Days Left is calculated using your configured Daily Hours.
• Added AI disclaimer blocks in all AI sections.
• Added English-only input validation in backend routing.
• Added internship-scope guard response:
  • "I can only help with internship-related topics in InternTrack, MentorBridge, and ReportWriter."
• MentorBridge is now handled as single-turn (no prior chat history is sent).
• The backend now includes a local RAG layer over markdown knowledge files.
• The backend can also call tools during generation for an agentic reasoning loop.
• Added prompt-injection defenses that refuse obvious prompt override attempts.
• Chat UI now uses Messenger-style alignment:
  • User messages on the right (orange avatar)
  • Assistant messages on the left (red avatar)
  • Bubble width auto-adjusts based on content.

Tech Stack

Layer	Technology
Frontend	Streamlit
Backend	FastAPI + Uvicorn
Database	MongoDB (local or Atlas)
AI	OpenAI GPT-4o + local RAG + tools + Langfuse tracing

---

Prerequisites

• Python 3.11+
• MongoDB running locally on localhost:27017 or a MongoDB Atlas cluster
• An OpenAI API key

For Langfuse tracing, add these environment variables to your .env file:

LANGFUSE_PUBLIC_KEY=your_langfuse_public_key
LANGFUSE_SECRET_KEY=your_langfuse_secret_key
LANGFUSE_HOST=https://cloud.langfuse.com
LANGFUSE_TRACING_ENABLED=true
LANGFUSE_TRACING_ENVIRONMENT=development
LANGFUSE_RELEASE=ai-chatbot

The app still runs without them, but Langfuse tracing stays disabled until the keys are set.

Langfuse prompt management

This app now fetches its system prompts from Langfuse by prompt name and label. The prompt names used by the backend are:

• interntrack-system-prompt
• mentorbridge-system-prompt
• report-writer-system-prompt
• chatbot-security-prompt

The app loads the production label by default and records the prompt name, label, and version in each Langfuse trace. If a prompt does not exist yet, the backend can seed it automatically from the current fallback text the first time it runs.

If you update a prompt in Langfuse, the app will pick up the new version the next time it fetches that prompt.

---

Setup

1. Clone the repository

git clone <repo-url>
cd AI-Chatbot

2. Create and activate a virtual environment

python -m venv .venv

# Windows
.venv\Scripts\activate

# macOS / Linux
source .venv/bin/activate

3. Install dependencies

pip install -r backend/requirements.txt
pip install -r frontend/requirements.txt

4. Create the .env file

Create a .env file in the project root (same level as this README):

OPENAI_API_KEY=your_openai_api_key_here
API_BASE_URL=http://your_railway_url

API_BASE_URL is optional and defaults to http://localhost:8001.

Running the App

You need two terminals running at the same time.

Terminal 1 — Start the backend

cd backend
uvicorn main:app --reload --port 8001

Backend runs at: http://localhost:8001
API docs available at: http://localhost:8001/docs

Terminal 2 — Start the frontend

cd frontend
streamlit run app.py

Frontend runs at: http://localhost:8501

---

InternTrack Input Behavior

InternTrack accepts quick chat commands for hours updates.

• Add hours: +8 today
• Subtract hours: -2 correction

Rules:

• Hours are clamped between 0 and Total Required Hours.
• Days Left updates using ceil(remaining / daily_hours).
• Daily Hours comes from the Configuration panel.

---

Language and Scope Rules

• The backend accepts English-only user input.
• Non-English input receives this response:
  • Please write your message in English only.
• Programming/coding queries are out of scope and receive this response:
  • I can only help with internship-related topics in InternTrack, MentorBridge, and ReportWriter.

---

MentorBridge Behavior

• MentorBridge runs in single-turn mode.
• Previous MentorBridge messages are not carried into new prompts.
• This keeps each workplace-message draft focused on the latest situation only.

---

MongoDB Collections

The app uses a database named interntrack with these collections:

Collection	Purpose
hours	Logged internship hours
tasks	Task entries
chat	All chat messages (all three tools)
config	Saved InternTrack configuration
reports	Generated Report Writer reports

---

Project Structure

AI-Chatbot/
├── backend/
│   ├── main.py          # FastAPI app and routes
│   ├── ai_router.py     # AI prompt routing (InternTrack / MentorBridge / Report Writer)
│   ├── database.py      # MongoDB connection and collections
│   └── tracker.py       # Hours logging logic
│   └── knowledge_base/  # Markdown files used by the retriever
├── frontend/
│   ├── app.py           # Streamlit UI
│   └── styles.py        # CSS styles
├── .env                 # API keys (never commit this)
├── .gitignore
└── README.md