2026
Adaptive Tool Log Pipeline
Parses semiconductor tool logs in any format and learns a common schema from feedback
- Log formats
- 4
- Storage
- SQLite
- Optional parsing hook
- LLM
- Python
- FastAPI
- SQLite
- Streamlit
- scikit-learn
Summary
A prototype pipeline that ingests semiconductor tool logs, parses them, normalises them to a common schema, stores them, and explains them in plain language. Structured logs are parsed with code; free-text logs use regular expressions with an optional LLM hook.
Problem
Every tool vendor logs in its own format, so engineers spend time translating logs before they can compare or analyse them.
Approach
A FastAPI backend with hybrid parsing, semantic normalisation, and a schema mapper that learns aliases from every ingestion and from user feedback. Results persist in SQLite, with optional anomaly detection and a Streamlit demo UI.