I deployed LeadFlow AI on Ubuntu 24.04 LTS following a native‑Linux workflow. The system packages I installed are: bash sudo apt update && sudo apt upgrade -y sudo apt install -y python3.12 python3.12-venv python3.12-dev postgresql postgresql-contrib redis-server nginx git curl libpq-dev
- System packages installed: python3.12, python3.12‑venv, python3.12‑dev, postgresql, postgresql‑contrib, redis‑server, nginx, git, curl, libpq-dev.
I created an unprivileged system user named leadflow to own the application files.
The PostgreSQL role leadflow and database leadflow_production were created with full privileges; the placeholder password is YOUR_SECURE_PASSWORD.
Application code resides in /opt/leadflow-ai, owned by leadflow; a virtual environment at /opt/leadflow-ai/.venv installs dependencies from requirements.txt.
Production environment variables are stored in /etc/leadflow/leadflow.env with mode 600 and symlinked into the project root.
Systemd unit files for leadflow-api, leadflow-campaign-worker, leadflow-enrichment-worker, and leadflow-maintenance-worker were installed, enabled, and started.
Nginx is configured via leadflow.conf, passes the test (nginx -t), and reloads successfully.
The readiness probe returns {"status":"ready","app":"LeadFlow AI","version":"2.0.0","checks":{"database":"connected","redis":"connected"}} and the validation script reports 32 successful checks.
I designed the production architecture for LeadFlow AI around a lightweight, container‑free stack. Nginx 1.24 terminates TLS and enforces security headers and rate limiting, forwarding traffic to a FastAPI application served by four Uvicorn workers under systemd. Redis 7 provides fast locking and caching, while PostgreSQL 16 stores multi‑tenant data with row‑level locking. The stack is benchmarked on a fresh Ubuntu 24.04 LTS VPS and shows zero errors up to 10,000 requests.
The reverse proxy uses Nginx 1.24 for HTTPS termination, modern TLS ciphers, security headers, and rate limiting.
FastAPI is served by Uvicorn with 4 workers managed by systemd, providing fail‑closed authentication and health/metrics endpoints.
Redis 7 runs as a standalone service offering fast locks and cache, enabling graceful degradation under load.
I built LeadFlow AI, an enterprise‑grade outbound intelligence and cold‑outreach platform hosted on @mostlyunstable's public GitHub repository. It is written in Python, created in April 2026, with the most recent commit in September 2026. The system runs as native Linux processes supervised by systemd, terminated by Nginx HTTPS, and relies on PostgreSQL 16 and Redis 7 for storage and graceful degradation.
Repository created: 2026‑04
Last update: 2026‑09
PostgreSQL version: 16
Redis version: 7
Automated VPS tests passing: 79
Real‑world VPS tests passing: 32
Backup script execution time: 249 ms
Restore script RTO: 966 ms
Password hash rounds: 310,000