Glossary

Technical Terms

CI/CD: Continuous Integration / Continuous Deployment. DevOps practice for automating code testing and deployment.
Coverage: Measure of code percentage tested by unit tests. Project objective: ≥90%.
DuckDB: High-performance columnar OLAP (Online Analytical Processing) database. SQLite alternative for analytics.
DNAT: Destination Network Address Translation. iptables technique to redirect network traffic directly to port 3910, bypassing reverse proxy for 10x performance gain.
Flake8: Python tool verifying PEP8 compliance (standard Python code style).
Parquet: Compressed columnar file format optimized for analytics. 5-10x more performant than CSV.
Polars: High-performance Python DataFrame library, Pandas alternative. Rust engine, Arrow format.
Pytest: Most popular Python unit testing framework.
S3 Garage: Self-hosted Amazon S3 protocol implementation. Used for project dataset storage.
Streamlit: Python framework for creating interactive data science web apps without JavaScript.
TTL: Time To Live. Cache duration before expiration (project: 3600s = 1h).
uv: Modern ultra-fast Python package manager. pip replacement, 10-100x faster.

PREPROD: Pre-production environment for testing before PROD deployment. URL: https://mangetamain.lafrance.io/, port 8500.
PRODUCTION: Stable production environment. URL: https://backtothefuturekitchen.lafrance.io/, port 8501.
Recipes Clean: Cleaned dataset of 178,265 Food.com recipes (1999-2018). File: recipes_clean.parquet (250 MB).
Ratings Longterm: Dataset of 1.1M+ user interactions. File: ratings_longterm.parquet (180 MB).
Back to the Kitchen: Project name and visual identity (orange/black palette).
Runner Self-Hosted: GitHub Actions machine running CI/CD directly on dataia VM (avoids VPN).

Trends Analysis: Module analyse_trendlines_v2 studying recipe evolution 1999-2018 (volume, duration, complexity).
Seasonal Analysis: Module analyse_seasonality identifying seasonal patterns (winter/summer, December peaks).
Weekend Analysis: Module analyse_weekend comparing weekday vs weekend publications (Monday +45%, Saturday -49%).
Ratings Analysis: Module analyse_ratings studying user rating distribution (78% positive bias = 5★).
Complexity Score: 0-10 score calculated from number of steps + ingredients + preparation time.
Visual Identity: Unified color palette: ORANGE_PRIMARY (#FF8C00), BACKGROUND_MAIN (#1E1E1E).

Dataia: VM hosting PREPROD/PROD services, Garage S3, GitHub Actions runner. IP: 192.168.80.202.
Docker Compose: Docker container orchestration tool. Files: docker-compose.yml (PREPROD), docker-compose-prod.yml (PROD).
Health Check: HTTP endpoint verifying application health. URL: /_stcore/health. Retry 3 times, timeout 10s.
Loguru: Modern Python logging library. Config: 2 files (debug.log, errors.log), automatic rotation.
Discord Webhook: Real-time CI/CD notifications in #deployments channel.

API: Application Programming Interface. Documented Python module reference.
ASCII: American Standard Code for Information Interchange. Text diagrams used in documentation.
CSV: Comma-Separated Values. Tabular file format (replaced by Parquet for performance).
EDA: Exploratory Data Analysis. Jupyter notebooks for data exploration (00_eda/).
FAQ: Frequently Asked Questions. Common questions page.
JSON: JavaScript Object Notation. Data exchange format.
OLAP: Online Analytical Processing. Database type optimized for analytics (DuckDB).
PEP8: Python Enhancement Proposal 8. Standard Python code style guide.
PR: Pull Request. GitHub code merge request.
REST: Representational State Transfer. API architecture (not used in project, web app).
RST: reStructuredText. Sphinx documentation markup format.
SQL: Structured Query Language. Database query language.
SSH: Secure Shell. Protocol for secure dataia VM connection.
TTL: Time To Live. Cache duration.
UI: User Interface. Streamlit user interface.
URL: Uniform Resource Locator. Web address.
VPN: Virtual Private Network. Virtual private network (dataia accessible via VPN).