mirror of
https://github.com/different-ai/openwork
synced 2026-04-25 17:15:34 +02:00
800602f4e3
* add pre-baked microsandbox image Bake openwork, openwork-server, and the pinned opencode binary into a single Docker image so micro-sandbox remote-connect smoke tests can boot quickly and be verified with curl and container health checks. * add Rust microsandbox example Add a standalone microsandbox SDK example that boots the OpenWork image, validates remote-connect endpoints, and streams sandbox logs so backend-only sandbox behavior can be exercised without Docker. * exclude Rust example build output Keep the standalone microsandbox example in git, but drop generated Cargo target artifacts so the branch only contains source, docs, and lockfile. * test * add microsandbox feature flag for sandbox creation Made-with: Cursor * refactor sandbox mode isolation Made-with: Cursor
70 lines
3.3 KiB
Markdown
70 lines
3.3 KiB
Markdown
# Microsandbox OpenWork Rust Example
|
|
|
|
Small standalone Rust example that starts the OpenWork micro-sandbox image with the `microsandbox` SDK, publishes the OpenWork server on a host port, persists `/workspace` and `/data` with host bind mounts, verifies `/health`, checks that `/workspaces` is `401` without a token and `200` with the client token, then keeps the sandbox alive until `Ctrl+C` while streaming the sandbox logs to your terminal.
|
|
|
|
## Run
|
|
|
|
```bash
|
|
cargo run --manifest-path examples/microsandbox-openwork-rust/Cargo.toml
|
|
```
|
|
|
|
Useful environment overrides:
|
|
|
|
- `OPENWORK_MICROSANDBOX_IMAGE` - OCI image reference to boot. Defaults to `openwork-microsandbox:dev`.
|
|
- `OPENWORK_MICROSANDBOX_NAME` - sandbox name. Defaults to `openwork-microsandbox-rust`.
|
|
- `OPENWORK_MICROSANDBOX_WORKSPACE_DIR` - host directory bind-mounted at `/workspace`. Defaults to `examples/microsandbox-openwork-rust/.state/<sandbox-name>/workspace`.
|
|
- `OPENWORK_MICROSANDBOX_DATA_DIR` - host directory bind-mounted at `/data`. Defaults to `examples/microsandbox-openwork-rust/.state/<sandbox-name>/data`.
|
|
- `OPENWORK_MICROSANDBOX_REPLACE` - set to `1` or `true` to replace the sandbox instead of reusing persistent state. Defaults to off.
|
|
- `OPENWORK_MICROSANDBOX_PORT` - published host port. Defaults to `8787`.
|
|
- `OPENWORK_CONNECT_HOST` - hostname you want clients to use. Defaults to `127.0.0.1`.
|
|
- `OPENWORK_TOKEN` - remote-connect client token. Defaults to `microsandbox-token`.
|
|
- `OPENWORK_HOST_TOKEN` - host/admin token. Defaults to `microsandbox-host-token`.
|
|
|
|
Example:
|
|
|
|
```bash
|
|
OPENWORK_MICROSANDBOX_IMAGE=ghcr.io/example/openwork-microsandbox:dev \
|
|
OPENWORK_MICROSANDBOX_WORKSPACE_DIR="$PWD/examples/microsandbox-openwork-rust/.state/demo/workspace" \
|
|
OPENWORK_MICROSANDBOX_DATA_DIR="$PWD/examples/microsandbox-openwork-rust/.state/demo/data" \
|
|
OPENWORK_CONNECT_HOST=127.0.0.1 \
|
|
OPENWORK_TOKEN=some-shared-secret \
|
|
OPENWORK_HOST_TOKEN=some-owner-secret \
|
|
cargo run --manifest-path examples/microsandbox-openwork-rust/Cargo.toml
|
|
```
|
|
|
|
## Test
|
|
|
|
The crate includes an ignored end-to-end smoke test that:
|
|
|
|
- boots the microsandbox image
|
|
- waits for `/health`
|
|
- verifies unauthenticated `/workspaces` returns `401`
|
|
- verifies authenticated `/workspaces` returns `200`
|
|
- creates an OpenCode session through `/w/:workspaceId/opencode/session`
|
|
- fetches the created session and its messages
|
|
|
|
Run it explicitly:
|
|
|
|
```bash
|
|
OPENWORK_MICROSANDBOX_IMAGE=ttl.sh/openwork-microsandbox-11559:1d \
|
|
cargo test --manifest-path examples/microsandbox-openwork-rust/Cargo.toml -- --ignored --nocapture
|
|
```
|
|
|
|
## Persistence behavior
|
|
|
|
By default, the example creates and reuses two host directories under `examples/microsandbox-openwork-rust/.state/<sandbox-name>/`:
|
|
|
|
- `/workspace`
|
|
- `/data`
|
|
|
|
That keeps OpenWork and OpenCode state around across sandbox restarts, while using normal host filesystem semantics instead of managed microsandbox named volumes.
|
|
|
|
If you want a clean reset, either:
|
|
|
|
- change the sandbox name or bind mount paths, or
|
|
- set `OPENWORK_MICROSANDBOX_REPLACE=1`
|
|
|
|
## Note on local Docker images
|
|
|
|
`microsandbox` expects an OCI image reference. If `openwork-microsandbox:dev` only exists in your local Docker daemon, the SDK may not be able to resolve it directly. In that case, push the image to a registry or otherwise make it available as a pullable OCI image reference first, then set `OPENWORK_MICROSANDBOX_IMAGE` to that ref.
|