Add workspace summary across the CLI, SDK, and MCP, and include it in the workspace-core profile so chat hosts can review one concise view of the current session.
Persist lightweight review events for syncs, file edits, patch applies, exports, service lifecycle, and snapshot activity, then synthesize them with command history, current services, snapshot state, and current diff data since the last reset.
Update the walkthroughs, use-case docs, public contract, changelog, and roadmap for 4.3.0, and make dist-check invoke the CLI module directly so local package reinstall quirks do not break the packaging gate.
Validation: uv lock; ./.venv/bin/pytest --no-cov tests/test_vm_manager.py tests/test_cli.py tests/test_api.py tests/test_server.py tests/test_public_contract.py tests/test_workspace_use_case_smokes.py; UV_OFFLINE=1 UV_CACHE_DIR=.uv-cache make check; UV_OFFLINE=1 UV_CACHE_DIR=.uv-cache make dist-check; real guest-backed workspace create -> patch apply -> workspace summary --json -> delete smoke.
make test was dominated by teardown-heavy workspace integration tests, not by coverage overhead. Service shutdown was treating zombie processes as live, which forced repeated timeout waits, and one shell test was leaving killpg monkeypatched during cleanup, which made shell close paths burn the full wait budget.\n\nTreat Linux zombie pids as stopped in the workspace manager so service teardown completes promptly. Restore the real killpg implementation before shell test cleanup so the shell close path no longer pays the artificial timeout. Also isolate sys.argv in the runtime-network-check main() test so parallel pytest flags do not leak into argparse-based tests.\n\nAdd pytest-xdist to the dev environment and run make test with pytest -n auto by default so available cores are used automatically during local iteration.\n\nValidation:\n- uv lock\n- targeted hot-spot pytest rerun after the fix dropped the worst tests from roughly 10-21s each to sub-second timings\n- UV_CACHE_DIR=.uv-cache make check\n- UV_CACHE_DIR=.uv-cache make dist-check
Turn the stable workspace surface into five documented, runnable stories with a shared guest-backed smoke runner, new docs/use-cases recipes, and Make targets for cold-start validation, repro/fix loops, parallel workspaces, untrusted inspection, and review/eval workflows.
Bump the package and catalog surface to 3.6.0, update the main docs to point users from the stable workspace walkthrough into the recipe index and smoke packs, and mark the 3.6.0 roadmap milestone done.
Fix a regression uncovered by the real parallel-workspaces smoke: workspace_file_read must not bump last_activity_at. Verified with uv lock, UV_CACHE_DIR=.uv-cache make check, UV_CACHE_DIR=.uv-cache make dist-check, and USE_CASE_ENVIRONMENT=debian:12 UV_CACHE_DIR=.uv-cache make smoke-use-cases.
Make workspace shell reads usable as direct chat-model input without changing the PTY or cursor model. This adds optional plain rendering and idle-window batching across CLI, SDK, and MCP while keeping raw reads backward-compatible.
Implement the rendering and wait-for-idle logic in the manager layer so the existing guest/backend shell transport stays unchanged. The new helper strips ANSI and other terminal control noise, handles carriage-return overwrite and backspace, and preserves raw cursor semantics even when plain output is requested.
Refresh the stable shell docs/examples to recommend --plain --wait-for-idle-ms 300, mark the 3.5.0 roadmap milestone done, and bump the package/catalog version to 3.5.0.
Validation: uv lock; UV_CACHE_DIR=.uv-cache make check; UV_CACHE_DIR=.uv-cache make dist-check; real guest-backed Firecracker smoke covering shell open/write/read with ANSI plus delayed output.
Make concurrent workspaces easier to rediscover and resume without relying on opaque IDs alone.
Add optional workspace names, key/value labels, workspace list, and workspace update across the CLI, Python SDK, and MCP surface, and persist last_activity_at so list ordering reflects real mutating activity.
Update the stable contract, install/first-run docs, roadmap, and Python workspace example to teach the new discovery flow, and validate it with focused manager/CLI/API/server coverage plus uv lock, make check, make dist-check, and a real multi-workspace smoke for create, list, update, exec, reorder, and delete.
Remove shell-escaped file mutation from the stable workspace flow by adding explicit file and patch tools across the CLI, SDK, and MCP surfaces.
This adds workspace file list/read/write plus unified text patch application, backed by new guest and manager file primitives that stay scoped to started workspaces and /workspace only. Patch application is preflighted on the host, file writes stay text-only and bounded, and the existing diff/export/reset semantics remain intact.
The milestone also updates the 3.2.0 roadmap, public contract, docs, examples, and versioning, and includes focused coverage for the new helper module and dispatch paths.
Validation:
- uv lock
- UV_CACHE_DIR=.uv-cache make check
- UV_CACHE_DIR=.uv-cache make dist-check
- real guest-backed smoke for workspace file read, patch apply, exec, export, and delete
Finish the 3.1.0 secondary disk-tools milestone so stable workspaces can be
stopped, inspected offline, exported as raw ext4 images, and started again
without changing the primary workspace-first interaction model.
Add workspace stop/start plus workspace disk export/list/read across the CLI,
SDK, and MCP, backed by a new offline debugfs inspection helper and guest-only
validation. Scrub runtime-only guest state before disk inspection/export, and
fix the real guest reliability gaps by flushing the filesystem on stop and
removing stale Firecracker socket files before restart.
Update the docs, examples, changelog, and roadmap to mark 3.1.0 done, and
cover the new lifecycle/disk paths with API, CLI, manager, contract, and
package-surface tests.
Validation: uv lock; UV_CACHE_DIR=.uv-cache make check; UV_CACHE_DIR=.uv-cache
make dist-check; real guest-backed smoke for create, shell/service activity,
stop, workspace disk list/read/export, start, exec, and delete.
Replace the workspace-level boolean network toggle with explicit network policies and attach localhost TCP publication to workspace services.
Persist network_policy in workspace records, validate --publish requests, and run host-side proxy helpers that follow the service lifecycle so published ports are cleaned up on failure, stop, reset, and delete.
Update the CLI, SDK, MCP contract, docs, roadmap, and examples for the new policy model, add coverage for the proxy and manager edge cases, and validate with uv lock, UV_CACHE_DIR=.uv-cache make check, UV_CACHE_DIR=.uv-cache make dist-check, and a real guest-backed published-port probe smoke.
Add explicit workspace secrets across the CLI, SDK, and MCP, with create-time secret definitions and per-call secret-to-env mapping for exec, shell open, and service start. Persist only safe secret metadata in workspace records, materialize secret files under /run/pyro-secrets, and redact secret values from exec output, shell reads, service logs, and surfaced errors.
Fix the remaining real-guest shell gap by shipping bundled guest init alongside the guest agent and patching both into guest-backed workspace rootfs images before boot. The new init mounts devpts so PTY shells work on Firecracker guests, while reset continues to recreate the sandbox and re-materialize secrets from stored task-local secret material.
Validation: uv lock; UV_CACHE_DIR=.uv-cache make check; UV_CACHE_DIR=.uv-cache make dist-check; and a real guest-backed Firecracker smoke covering workspace create with secrets, secret-backed exec, shell, service, reset, and delete.
Implement the 2.8.0 workspace milestone with named snapshots and full-sandbox reset across the CLI, Python SDK, and MCP server.
Persist the immutable baseline plus named snapshot archives under each workspace, add workspace reset metadata, and make reset recreate the sandbox while clearing command history, shells, and services without changing the workspace identity or diff baseline.
Refresh the 2.8.0 docs, roadmap, and Python example around reset-over-repair, then validate with uv lock, UV_CACHE_DIR=.uv-cache make check, UV_CACHE_DIR=.uv-cache make dist-check, and a real guest-backed create/snapshot/reset/diff smoke test outside the sandbox.
Make persistent workspaces capable of running long-lived background processes instead of forcing everything through one-shot exec calls.
Add workspace service start/list/status/logs/stop across the CLI, Python SDK, and MCP server, with multiple named services per workspace, typed readiness probes (file, tcp, http, and command), and aggregate service counts on workspace status. Keep service state and logs outside /workspace so diff and export semantics stay workspace-scoped, and extend the guest agent plus backends to persist service records and logs across separate calls.
Update the 2.7.0 docs, examples, changelog, and roadmap milestone to reflect the shipped surface.
Validation: uv lock; UV_CACHE_DIR=.uv-cache make check; UV_CACHE_DIR=.uv-cache make dist-check; real guest-backed Firecracker smoke for workspace create, two service starts, list/status/logs, diff unaffected, stop, and delete.
Complete the 2.6.0 workspace milestone by adding explicit host-out export and immutable-baseline diff across the CLI, Python SDK, and MCP server.
Capture a baseline archive at workspace creation, export live /workspace paths through the guest agent, and compute structured whole-workspace diffs on the host without affecting command logs or shell state. The docs, roadmap, bundled guest agent, and workspace example now reflect the new create -> sync -> diff -> export workflow.
Validation: uv lock, UV_CACHE_DIR=.uv-cache make check, UV_CACHE_DIR=.uv-cache make dist-check, and a real guest-backed Firecracker smoke covering workspace create, sync push, diff, export, and delete.
Let agents inhabit a workspace across separate calls instead of only submitting one-shot execs.
Add workspace shell open/read/write/signal/close across the CLI, Python SDK, and MCP server, with persisted shell records, a local PTY-backed mock implementation, and guest-agent support for real Firecracker workspaces.
Mark the 2.5.0 roadmap milestone done, refresh docs/examples and the release metadata, and verify with uv lock, UV_CACHE_DIR=.uv-cache make check, and UV_CACHE_DIR=.uv-cache make dist-check.
Replace the public persistent-sandbox contract with workspace-first naming across CLI, SDK, MCP, payloads, and on-disk state.
Rename the task surface to workspace equivalents, switch create-time seeding to `seed_path`, and store records under `workspaces/<workspace_id>/workspace.json` without carrying legacy task aliases or migrating old local task state.
Keep `pyro run` and `vm_*` unchanged. Validation covered `uv lock`, focused public-contract/API/CLI/manager tests, `UV_CACHE_DIR=.uv-cache make check`, and `UV_CACHE_DIR=.uv-cache make dist-check`.
Tasks could start from host content in 2.2.0, but there was still no post-create path to update a live workspace from the host. This change adds the next host-to-task step so repeated fix or review loops do not require recreating the task for every local change.
Add task sync push across the CLI, Python SDK, and MCP server, reusing the existing safe archive import path from seeded task creation instead of introducing a second transfer stack. The implementation keeps sync separate from workspace_seed metadata, validates destinations under /workspace, and documents the current non-atomic recovery path as delete-and-recreate.
Validation:
- uv lock
- UV_CACHE_DIR=.uv-cache uv run pytest --no-cov tests/test_cli.py tests/test_vm_manager.py tests/test_api.py tests/test_server.py tests/test_public_contract.py
- UV_CACHE_DIR=.uv-cache make check
- UV_CACHE_DIR=.uv-cache make dist-check
- real guest-backed smoke: task create --source-path, task sync push, task exec to verify both files, task delete
Current persistent tasks started with an empty workspace, which blocked the first useful host-to-task workflow in the task roadmap. This change lets task creation start from a host directory or tar archive without changing the one-shot VM surfaces.
Expose source_path on task create across the CLI, SDK, and MCP, add safe archive upload and extraction support for guest and host-compat backends, persist workspace_seed metadata, and patch the per-task rootfs with the bundled guest agent before boot so seeded guest tasks work without republishing environments. Also switch post--- command reconstruction to shlex.join() so documented sh -lc task examples preserve argument boundaries.
Validation:
- uv lock
- UV_CACHE_DIR=.uv-cache uv run pytest --no-cov tests/test_vm_guest.py tests/test_vm_manager.py tests/test_cli.py tests/test_api.py tests/test_server.py tests/test_public_contract.py
- UV_CACHE_DIR=.uv-cache make check
- UV_CACHE_DIR=.uv-cache make dist-check
- real guest-backed smoke: task create --source-path, task exec -- cat note.txt, task delete
Start the first workspace milestone toward the task-oriented product without changing the existing one-shot vm_run/pyro run contract.
Add a disk-backed task registry in the manager, auto-started task workspaces rooted at /workspace, repeated non-cleaning exec, and persisted command journals exposed through task create/exec/status/logs/delete across the CLI, Python SDK, and MCP server.
Update the public contract, docs, examples, and version/catalog metadata for 2.1.0, and cover the new surface with manager, CLI, SDK, and MCP tests. Validation: UV_CACHE_DIR=.uv-cache make check and UV_CACHE_DIR=.uv-cache make dist-check.
