HSE Bot

A Streamlit dashboard that helps Bayer's (HSE) teams browse observation data with visualization tools.

Key Features

• Tool registry with statistics, charting, embeddings, and classification helpers.
• Dual-language UI support (Finnish and English).
• Optional persistence layer for saving intermediate analysis states per query.

Repository Layout

• hse_visualizer.py – Streamlit entry point and multi-step workflow.
• mock_database.py – This is for development purposes and was used to simulate the LLM and Database access from RAG pipeline. It handles configuration, CSV loading, LLM client bootstrap, and persistence.
• tools.py – Tool definitions, prompt builders, and visualization helpers.
• config/ – Prompts, translations, and query metadata per language.
• data/ – CSV inputs plus optional pre-computed embeddings (*.npz + *.csv pairs).

Requirements

• Python 3.10+
• Access to an LLM endpoint compatible with the OpenAI API surface (default: Featherless API).
• For development we used the featherless AI API given to us for the hackathon.
• Api key defined in your .env. Default: FEATHERLESS_API_KEY

Quick Start

python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -e .               
streamlit run hse_visualizer.py

The Streamlit sidebar lets you select a language, connect to the LLM, choose an embedding preset, and load/save query runs.

Environment, Data, and Config

• Create a .env file in the project root (or export vars) and set FEATHERLESS_API_KEY=<your-secret>.
• Translation strings live in config/translations.json