Closing out open catalog APIs for authenticated and owned users (#784)
Apache Airavata is an all-in-one platform for executing and managing computational jobs on distributed computing resources, including local clusters, supercomputers, national grids, and academic and commercial clouds.
Airavata is also a foundational layer for building higher-level tools on top of distributed resources, such as:
Airavata is built around concepts to streamline research computing works in practice:
See the introduction
| Interface | Best for |
|---|---|
| Dashboard | Day-to-day use in a browser: registering clusters, answering MFA prompts, launching and following runs. |
| HTTP API | Scripts, portals and integrations. Every feature is a REST endpoint (API reference, OpenAPI spec). |
| Jupyter | Running notebooks on the compute node of a running job, with its files and kernels, from the dashboard (docs). |
| Local VS Code | Working from your own VS Code against the same running job, with the editor and tooling you already use. |
| airavata-mcp | Working through an AI assistant (Claude, Copilot, Cursor and others) that can turn an sbatch script into an application, plan runs and explain failures (docs). |
Jupyter and local VS Code give you one familiar interface to your work, whatever kind of compute resource it runs on and whatever the network between you and it allows. Jobs dial out to Airavata over a single outbound connection, so compute nodes need no inbound ports and nothing extra runs on the login node. Whether the job is on an HPC cluster behind a firewall or on a cloud VM, you open it the same way.
Prerequisites: Go 1.25+, PostgreSQL 14+ (Docker Compose is the easiest way to run it locally), and optionally Node.js 20.9+ to build the dashboard.
# 1. Start PostgreSQL (host port 15432) docker compose -f dev-tools/compose/compose.yml up -d postgres # 2. Build the server go build -o bin/airavata-server ./cmd/airavata-server # 3. Build the dashboard (optional) npm --prefix dashboard ci && npm --prefix dashboard run build # 4. Run ./bin/airavata-server
On first start the server creates its schema and prints a root account token. Copy it, since it is how you make the first authenticated call.
curl -s localhost:9095/health # {"status":"UP"}
Then open http://localhost:9095/dashboard/ and sign in with the token.
The defaults suit local development. For configuration options, cross-compiling and production setup, see INSTALL.md. Before a production deployment, configure CILogon and disable the root account as described in Authentication and authorization.
| Path | Contents |
|---|---|
cmd/airavata-server | The API server (a single Go binary). |
cmd/airavata-mcp | The Model Context Protocol server, a thin client of the REST API. |
api/ | HTTP API handlers, grouped by domain (application, compute, credentials, data, iam, process, sharing, audit). |
internal/ | Core implementation: auth, orchestration, SSH, database, permissions, configuration and more. |
dashboard/ | The web UI (Next.js), served by the API server under /dashboard/. |
docs/ | User and API documentation. |
website/ | The Docusaurus site that renders docs/ (airavata.apache.org). |
dev-tools/ | Docker Compose files and development clients. |
go build ./... # build everything go test ./... # run the tests go vet ./... # static checks gofmt -l . # formatting (no output means clean) go generate ./internal/mcp/client/ # regenerate the MCP client after editing docs/openapi.yaml
CI runs formatting, generated-code, vet, build and race-enabled test checks on every pull request. Please make sure these pass locally before opening one.
Contributions are welcome. To get started:
master.Apache Airavata is an Apache Software Foundation project, so contributors are expected to follow the ASF Code of Conduct. To report a bug or request a feature, open an issue.
Apache Airavata is licensed under the Apache License, Version 2.0.
Apache Airavata, Airavata, Apache, and the Apache feather logo are trademarks of The Apache Software Foundation.