How it works
From the signed job queue to pipelines, projects and the notebook workbench.
From queue to browser
Every job takes the same path, and every result is checked before it reaches you.
- Queue. Submitted jobs wait in a Redis queue until the worker for their model takes them.
- Routing. A routing service hands each job to the worker for its model — one worker per model, each with its own signing key.
- Signed results. Workers encrypt and sign what they return. The routing service verifies the signature, rejects replayed results and discards any result that fails.
- Live updates. Status changes and finished results are pushed to your browser as they happen, with a notification when a job completes.
Google DeepMind / Pexels ↗
A pipeline lists its stages up front: folding, docking, protein design, molecule generation or DNA generation. When every job of a stage has finished, Phonos collects that stage’s outputs and starts the next stage with them. Where structures and ligands meet, every structure is paired with every ligand.
Phonos checks a pipeline before it starts: each stage needs something the stages before it produce. Docking needs a structure, folding needs a sequence; generators need nothing and can sit anywhere in the chain.
Typical chains
- Fold, then dock — a sequence becomes a structure, and ligands are docked into it.
- Generate DNA, fold, then dock — generated sequences go straight on to folding.
- Fold and generate molecules, then dock — both feed the docking stage, structure by molecule.
Projects, roles and live collaboration
Research on Phonos happens in projects. Each member has one of three roles.
- Project manager
- Manages members and their roles, and runs jobs.
- Project contributor
- Runs jobs, uploads files and works in the project’s notebooks.
- Project viewer
- Follows files, results and discussions without changing them.
In a shared session, collaborators see each other’s cursor, selection and camera in the 3D viewer, add annotations, chat, and can start a video call.
Google DeepMind / Pexels ↗
More in the workbench
Notebook IDE
Python notebooks run inside a project, with its files and job outputs available under /project/. A version history keeps checkpoints you can compare, restore and export as .ipynb.
Public databases
Search the Protein Data Bank, the AlphaFold Database, UniProt and PubChem from inside Phonos, preview an entry and import it into your project.
REST API
Models, jobs, projects and files are also available over a REST API with API keys, including batch submission of jobs and the table of which output feeds which model.
Model chaining
Send any output to the models that accept it. Phonos classifies files by extension and content, so a structure is offered for docking and a designed sequence for folding.
Security
Logging in asks for a code sent to your email by default. Worker results are encrypted and signed, and stored job data is encrypted.
Two languages
The interface is available in English and German, and you can switch at any time.