Build, host and run AI solutions.
MLflow
MLflow — track your ML experiments, version your models and trace your LLM apps on your VPS, over HTTPS and protected.
In validation
Automatic installation for this solution is ready and currently going through our tests on a real server. Ordering will open as soon as validation is complete.
Tech stack
MLflow is the reference open-source platform (Apache-2.0 licensed) for the machine learning lifecycle. A central tracking server receives, from your notebooks and training scripts, the parameters, metrics and artifacts of every run: you compare your attempts side by side instead of copying them into a spreadsheet.
The model registry versions what goes to production, with aliases and history. And for generative AI applications, MLflow traces every call to a language model — inputs, outputs, latency, intermediate steps — so you can understand an answer and debug it. On ServOrbit, the server installs in one click, protected by an administrator account and served over HTTPS.
Key features
When to use this solution?
Compare your training runs
Every notebook or script run records its hyperparameters and metrics: you find your top-scoring attempt on one screen, with its code and files.
A model registry for the team
Register validated models, give them an alias such as "production", and let your inference service always load the right version.
Debug an LLM application
Turn on tracing for your assistant or agent: every request shows its model calls, its steps and its latency, so you can understand a bad answer.
Deploy MLflow on your VPS
Guide optimized for ServOrbit Cloud VPS.
Create the VPS and choose the address
Order an Ubuntu 24.04 VPS with 2 GB of RAM (4 GB and more disk if you store many models). MLflow requires a domain: your scripts, notebooks and team connect to it, so it must answer over HTTPS on a stable name. It opens on a subdomain of one of your domains, or on the VPS's free subdomain (mlflow.{dns_slug}.servorbit-dns.com), HTTPS included. No need to buy a domain name.
Install MLflow from 1-click apps
In your client area, select MLflow, choose the address and click Install. The script deploys the official container, generates the admin account's password, enables authentication on the interface and the API, limits the accepted host names to your address and configures the nginx reverse proxy over HTTPS.
Log in for the first time
Open https://your-domain/: the browser asks for credentials. Enter admin and the password shown in your client area. You can change it later through the users/update-password API; the new password is kept across restarts.
Connect your notebooks and scripts
Set MLFLOW_TRACKING_URI=https://your-domain, MLFLOW_TRACKING_USERNAME=admin and MLFLOW_TRACKING_PASSWORD, then call mlflow.start_run() in your code. Parameters, metrics and artifacts appear in the interface right away.
Open access to your team
Create one account per person with the users/create API, logged in as admin. A new account sees no experiment until you share one with it: permissions are granted experiment by experiment and model by model.
Frequently asked questions
Complementary solutions
Langfuse
Open-source LLM observability: trace every AI request, run evaluations, manage prompt versions and compare model costs — self-hosted, compatible with OpenAI, LiteLLM, LangChain and Ollama.
Developmentmarimo
Reactive Python notebook, open source: notebooks stored as plain .py files, cells that recompute automatically, data libraries included.
Artificial IntelligenceOllama
Run open-source LLMs locally via a dead-simple API. Pull Llama 3, Mistral, Qwen or DeepSeek in one command — OpenAI-compatible, zero per-token cost.
Need help?
Browse our help center and FAQ, or reach our team — callback, WhatsApp or email. Support in French, English and Arabic.
Message us on WhatsAppopens in a new tab