eb7c342fb1
* fix(sandbox): measure warm-pool idle age on a monotonic clock Warm-pool expiry measures a duration, but every provider stamped and reaped entries using the wall clock. A forward correction could therefore destroy fresh sandboxes, while a backward correction could retain expired ones until the clock caught up. Move every warm-pool writer and reader to time.monotonic(), including the Tenki and OpenSandbox reset paths missed by the earlier closed attempt. Keep absolute SandboxInfo.created_at and reconciliation grace timestamps on the wall clock. Convert AIO's sibling active-idle timestamps as one clock domain, and update finite clock mocks for the additional duration read. Fixes #6970 * test(sandbox): keep active-idle race fixtures in the monotonic domain The production AIO activity clock moved to monotonic time, but the shared orphan-reconciliation helper still stamped active fixtures with wall time. CI then computed a negative idle age, so two teardown race tests never reached the ownership gate. Use the same monotonic domain as production. * test(sandbox): read warm-pool idle fixtures on the monotonic clock The idle-age conversion left several fixtures seeding the fields the reaper now reads with values from the clock it no longer reads: - `test_idle_reaping_forces_a_new_remote` froze `warm_pool_lifecycle.time.time`, so the benchmark's reaper compared against real elapsed monotonic seconds — zero over the test's lifetime. Nothing expired, the entry survived, and the second turn came back as a warm hit. - Three orphan-reconciliation fixtures seeded `_last_activity` with `0.0` and two aio fixtures with `time.time()`, both of which are far below a running monotonic clock: the computed idle age went negative, so the reaper never reached the ownership claim these tests are built to exercise. Seed them from `time.monotonic()` instead, and keep `created_at` (a wall-clock field) on `time.time()`.
BoxLite backend
Runs each DeerFlow sandbox as a BoxLite micro-VM — a daemonless, OCI-native VM with its own kernel (libkrun/KVM on Linux, Hypervisor.framework on macOS). Motivated by the resource/cold-start pain with the default AIO Docker sandbox in #3439 and #3213; discussion in #3936.
Configuration
sandbox:
use: deerflow.community.boxlite:BoxliteProvider
image: python:3.12-slim # any OCI image (default: python:3.12-slim)
memory_mib: 1024 # per-box memory cap (optional)
cpus: 2 # per-box vCPUs (optional)
replicas: 3 # active + warm VM cap per gateway process (default: 3)
idle_timeout: 600 # warm VM idle seconds before stop; 0 disables
health_check_skip_seconds: 0.0 # optional low-latency mode: skip reclaim
# health checks for recent releases; 0 keeps
# reliability-first validation (default: 0.0)
environment: # injected into every command
PYTHONUNBUFFERED: "1"
Install the optional runtime before selecting this provider:
pip install "deerflow-harness[boxlite]"
The boxlite package is an optional DeerFlow harness extra, not part of the
default install. It is also limited to the host platforms and architectures
where BoxLite publishes wheels and can boot micro-VMs. Unsupported development
hosts, such as Windows, should use another sandbox provider or run DeerFlow from
a supported Linux/macOS environment.
Host requirement: BoxLite boots micro-VMs, so a Linux host needs KVM — i.e. nested virtualization when DeerFlow runs inside a cloud VM. macOS uses Hypervisor.framework. This is the main deployment constraint to weigh vs. the container-based providers.
Design
DeerFlow's Sandbox contract is synchronous; BoxLite's SDK is async-native and
its box handles are event-loop-affine. The provider owns one private asyncio
loop on a daemon thread and marshals every coroutine onto it via
run_coroutine_threadsafe. BoxLite boxes are named deterministically from
user_id:thread_id, released into an in-process warm pool after each agent turn,
and reclaimed by the same thread on the next acquire.
| File | Role |
|---|---|
provider.py |
SandboxProvider lifecycle + the private-loop bridge |
box.py |
Sandbox adapter; execute_command + file ops |
Contract coverage
The full Sandbox surface is implemented. File operations run as shell commands
inside the box and reuse deerflow.sandbox.search, mirroring e2b_sandbox:
execute_command—sh -lc, with per-call env and timeout.read_file/write_file/update_file—catand chunkedbase64(binary-safe, no arg-size limit).download_file— 100 MB cap, restricted to the/mnt/user-dataprefix.list_dir/glob/grep—find/grepwith busybox-portable flags; results filtered/capped in Python.
Command timeouts pass the same deadline to the SDK and event-loop bridge. A
bridge deadline returns Error: command timed out followed by Exit Code: 124.
BoxLite's SimpleBox.exec converts SDK wait failures to ExecResult(exit_code=-1);
if that result arrives before the bridge deadline, DeerFlow preserves -1.
The bridge handler catches Python's builtin TimeoutError, which is distinct
from the SDK's boxlite.TimeoutError.
The provider creates /mnt/user-data/{workspace,uploads,outputs} and
/mnt/skills on box start so those virtual paths resolve natively.
Warm-pool capacity is governed by sandbox.replicas across active + warm VMs.
sandbox.idle_timeout controls how long released warm VMs stay running; 0
disables idle reaping. Active boxes are never evicted to satisfy the cap.
Status
Verified end-to-end against a live box (provider resolution → execute_command
→ file ops) on macOS/HVF. Linux/KVM validation and benchmarks vs. the AIO
sandbox are tracked in
#3936.