Overview
Navigate the Console workspace, evaluation, reinforcement learning, and services.
The OpenPond Console brings your workspace, evaluation resources, training, and services into one place. These docs follow the same four sections as its sidebar. Open Console, choose your workspace, and use the Project picker to focus on a particular project or view the collections together.
| Console section | What you do here |
|---|---|
| Workspace | Manage your Profile, Marketplace, and Connected agents |
| Evaluation | Prepare Datasets, validate Graders, and run Experiments |
| Reinforcement Learning | Start Training, configure Continual Learning, and review Models |
| Services | Use the LLM Gateway, Ponder, Sandboxes, and Serving |
Workspaces and Projects
A workspace sets the membership and access boundary. A Console Project groups related targets, datasets, graders, experiments, and training resources. The Overview page opens the Project directory or the selected Project's resources and recent experiments. Create a Project there, give it a useful name and objective, and configure its targets and attached resources.
A Project groups existing resources; it does not change their versions or publish them. A Chat & Work repository Project has a different purpose: it provides source and execution context. See Work Projects.
Start with a measured workflow
- Set up your Profile or connect an agent whose work you want to inspect.
- Create a dataset from authored examples, approved captured work, or an available dataset release.
- Attach and validate a grader, then run an experiment and inspect failures.
- Start training only when the tasks, reward, runtime, held-out evaluation, and spending limit are ready.
- Compare the resulting candidate before choosing a version for Serving.
Use Inbox for work waiting on human review. API integrations start with Create an API key. For local work, see Desktop & Open Source and Choose a surface.