How Our AI Chatbot Automates ML Project Setup and Developer Tooling
Our AI chatbot connects directly to user hosts and automates everything from repo setup to dev-tool orchestration.
Instead of manually cloning repositories, configuring DVC, wiring up W&B, or launching Jupyter servers, users simply describe what they want — and the agent executes it safely, consistently, and with full observability.
Below are examples of common project-setup and developer-tooling requests and how the agent handles them behind the scenes.
1. “Clone my GitHub repo and set up dependencies.”
The agent handles full project onboarding automatically:
- Clones the specified GitHub repository using secure tokens
- Detects dependency managers (pip, Poetry, Conda, requirements.txt, environment.yml)
- Installs packages in the correct environment
- Runs import validation to ensure the project executes cleanly
This gives users a ready-to-run development environment within minutes.
2. “Initialize DVC tracking for my data.”
The chatbot configures DVC (Data Version Control) by:
- Initializing a new DVC repo (or detecting an existing one)
- Connecting remote storage (S3, GCS, Azure, SSH, local)
- Adding datasets and pipeline stages
- Creating initial commits and validating the workflow
Users get reproducible, trackable dataset pipelines without manually touching DVC commands.
3. “Connect my local run to Weights & Biases.”
The agent streamlines experiment tracking setup by:
- Injecting the user’s W&B API key securely
- Configuring environment variables and wandb.init defaults
- Updating training scripts if necessary
- Testing connectivity with a short diagnostic run
This allows users to log metrics and artifacts without manually editing code or managing credentials.
4. “Show me which ports are used by Jupyter or VS Code.”
The chatbot performs a safe scan of active development ports and reports:
- JupyterLab / Notebook ports
- VS Code Server ports
- Conflicting services
- Potential security exposures
It can also free ports, reassign them, or reserve new ones for development tools.
5. “Start a Jupyter notebook server.”
The agent launches a fully configured Jupyter server by:
- Selecting the correct Python environment
- Starting the server with secure tokens
- Managing ports and network configuration
- Exposing the notebook directly inside the Skyportal UI
This allows users to start coding immediately, without typing a single terminal command.
Security and Safety Guarantees
✔ Credential-safe operations
API keys (GitHub, W&B, DVC remotes) are encrypted and never logged in plaintext.
✔ Read-only repo validation
Before installation, the agent inspects repo files to prevent accidental execution of unsafe scripts.
✔ Sandboxed execution
Jupyter and dev servers run in tightly controlled environments with restricted permissions.
✔ Redaction of sensitive paths
The agent automatically filters secure paths, keys, tokens, and config files from all outputs.
Why This Matters
Setting up ML projects is time-consuming and error-prone — from cloning repos to fixing dependencies, wiring experiment trackers, and launching Jupyter servers.
Our chatbot removes all of that friction.
Whether the user needs to:
- Clone a repo and install dependencies
- Initialize DVC tracking
- Connect to Weights & Biases
- Manage ports for dev tools
- Launch Jupyter servers automatically
…they can do it instantly, reliably, and securely — without touching the terminal.