Reimplements the dropped firecracker fix on the refactored code, addressing
didericis-codex's blocking review on #471.
The previous firecracker path let the VM generate its own signing key on the
guest volume and had `_with_signing_key()` unconditionally overwrite the single
host-wide orchestrator-token with it. That breaks a co-running Docker/macOS
control plane: their orchestrators still verify with the old key while new host
clients start signing `cli` tokens with the guest key, so
supervise/teardown/policy calls against those backends begin returning 401
(cross-backend operation is supported).
Keep the host token file the single source of truth. The launcher now pushes the
host-canonical key into the freshly booted infra VM over SSH (atomic write,
mirroring persist_env_var_secret) via `_push_signing_key()`; the VM's init waits
for it and refuses to start the control plane — rather than run OPEN — if it
never arrives. Nothing writes back to the host file, so other backends are
untouched, and the best-effort silent path is gone (both the push and the guest
fail loudly).
Ported onto the reorg'd names (host_orchestrator_token, BOT_BOTTLE_ORCHESTRATOR_
TOKEN, orchestrator-token, bot_bottle.orchestrator_auth, bot_bottle.gateway.
bootstrap). pyright: 0 errors. Full unit suite green (2243).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The codebase used "control plane" both as an architectural role term AND
as an identifier alias for the orchestrator component, producing
duplicate names for one thing (control_plane_url vs orchestrator_url,
CONTROL_PLANE_PORT, host_control_plane_token, …). Going forward the
concrete component is always named for what it is — Gateway or
Orchestrator — and the plane vocabulary is reserved for prose (module
descriptions, the security argument).
Renamed (identifiers + the in-repo env/wire/file string values, all
setters/getters are in this repo so the change is atomic):
ControlPlaneServer -> OrchestratorServer
control_plane_url -> orchestrator_url
probe_control_plane_url -> probe_orchestrator_url
host_control_plane_token -> host_orchestrator_token
CONTROL_PLANE_PORT -> ORCHESTRATOR_PORT
CONTROL_PLANE_TOKEN_ENV/FILE -> ORCHESTRATOR_TOKEN_ENV/FILENAME
BOT_BOTTLE_CONTROL_PLANE_TOKEN-> BOT_BOTTLE_ORCHESTRATOR_TOKEN
control-plane-token (file) -> orchestrator-token
control_auth (module) -> orchestrator_auth (stays top-level;
the gateway imports it and must not
import the orchestrator/ package)
CONTROL_AUTH_HEADER -> ORCHESTRATOR_AUTH_HEADER
x-bot-bottle-control-auth -> x-bot-bottle-orchestrator-auth
CONTROL_AUTH_JWT_ENV -> ORCHESTRATOR_AUTH_JWT_ENV
BOT_BOTTLE_CONTROL_AUTH_JWT -> BOT_BOTTLE_ORCHESTRATOR_AUTH_JWT
_control_auth_headers -> _orchestrator_auth_headers
Prose plane-terms ("control plane", "data plane") are preserved,
including the test name test_data_plane_daemons_get_jwt_not_key (it
names the security invariant). Gateway and orchestrator verified to
agree on the renamed wire header; full unit suite green (2243).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
egress_entrypoint.sh is the mitmdump launcher for the egress data plane,
so it belongs with the gateway egress service, not at the package root.
Moved to bot_bottle/gateway/egress/entrypoint.sh (prefix stripped now
that the package namespaces it).
Only the Dockerfile.gateway COPY source changes; the runtime target
(/app/egress-entrypoint.sh, referenced by bootstrap's egress DaemonSpec)
is unchanged, and the infra images inherit it via FROM gateway. Updated
the test that resolves the script path and the infra-artifact docstring
example. Full unit suite green (2243).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Group the gateway's data-plane modules into three service sub-packages
mirroring the host-side trio (bot_bottle.egress / .supervisor / .git_gate):
gateway/egress/ addon_core, addon, dlp_config, dlp_detectors
gateway/supervisor/ server (was supervise_server)
gateway/git_gate/ render, http_backend
Prefix-stripped filenames now that the package namespaces them; each
sub-package has a thin docstring __init__ (no eager imports, cheap leaf
loads). The two cross-cutting files stay at the gateway root:
policy_resolver (shared per-client lookup) and gateway_init, renamed to
bootstrap now that gateway/ already namespaces it.
Updated all importers (bot_bottle + tests), the in-VM/container `-m`
launch strings, the Dockerfile.gateway addon shim + ENTRYPOINT, and the
five gateway entries in scripts/critical-modules.txt. Full unit suite
green (2243).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Split the flat egress.py into an egress/ package: neutral DTOs
(EgressRoute, EgressPlan) in plan.py, and the concrete Egress service
class plus rendering/env helpers in service.py, behind a thin
__getattr__ facade so leaf imports stay cheap.
Egress drops ABC and becomes a concrete service the backends call:
resolve_token_values / agent_env_entries / prepare are now methods.
Backend launch paths (docker, firecracker, macos_container) call
Egress().<method>(...) instead of the module-level functions.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Turn the git_gate module into a package with GitGate as a concrete service class
(dropping the ABC), mirroring the Supervisor shape. The host-side git-gate
operations are now methods the backend drives:
git_gate/
__init__.py — thin __getattr__ facade (keeps `from bot_bottle.git_gate
import …` working; render/hook names lazily forwarded to
gateway.git_gate_render)
plan.py — GitGatePlan (the launch DTO the backend contract references)
service.py — GitGate: prepare / provision_dynamic_keys /
revoke_provisioned_keys / preflight_host_keys
provision.py — deploy-key lifecycle (was git_gate_provision.py)
host_key.py — host-key preflight (was git_gate_host_key.py)
The backend launch + base.py preflight now call GitGate() methods instead of the
free functions. The runtime rendering / in-gateway hook execution stay in
gateway/git_gate_render.py (data plane) — only the host-side service moved.
Because the facade forwards names lazily, the ~25 `from bot_bottle.git_gate
import GitGatePlan` call-sites are unchanged, and importing git_gate.plan (the
contract's dependency) no longer drags in the provisioning / forge-API code.
Full unit suite green (2243).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The three concrete-backend package inits eagerly imported their
`*BottleBackend` (which pulls the whole framework), so importing any leaf under
them (e.g. `backend.docker.util`) paid ~78 modules. Give each the same thin
`__getattr__` treatment as the top-level backend package: re-export the public
class(es) lazily on first access, with a TYPE_CHECKING block for checkers.
`import bot_bottle.backend.docker.util` drops from 78 -> 6 bot_bottle modules;
the docker/firecracker/macos-container packages themselves are ~3 each. All
`from bot_bottle.backend.docker import DockerBottleBackend` call-sites and the
lazy selection path keep working.
This is what makes backend leaves cheap — and unblocks moving docker_cmd into
backend/docker/util if we want.
Full unit suite green (2243).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Give each service its own store package + manager, and cut the supervise module
along the control/data-plane boundary so nothing in the shared layer reaches up
into the orchestrator.
Stores, by owner:
- bot_bottle/store/ keeps only the shared base (DbStore, migrations) and the
concrete stores that aren't service-owned (audit_store, config_store).
- bot_bottle/orchestrator/store/ now houses the orchestrator-owned stores —
queue_store (supervise queue), secret_store, config_store — plus a new
orchestrator store_manager that migrates them (composing audit/config
downward from the base). The old shared store_manager is gone.
Supervise plane, by tier:
- bot_bottle/supervisor/ (NEUTRAL, importable by every tier including the
gateway): types.py (the Proposal/Response/AuditEntry wire types + the tool/
status/poll constants + the shared daemon constants moved out of
supervise.py) and plan.py (SupervisePlan, a pure DTO).
- bot_bottle/orchestrator/supervisor/ (orchestrator-only): queue.py (the
queue/audit I/O wrappers + render_diff + sha256_hex) and supervise.py (the
Supervise lifecycle that stages the DB via the store manager). Its __init__
re-exports the neutral vocabulary so orchestrator-side callers import from
one place.
The gateway now imports only bot_bottle.supervisor.types (never
bot_bottle.supervise), so the data plane holds no code dependency on the
orchestrator — it reaches the queue over the control-plane RPC. This removes
the circular import that moving queue_store under orchestrator introduced
(supervise -> orchestrator -> service -> supervise).
supervise_types.py -> supervisor/types.py; supervise.py deleted (split). Full
unit suite green (2251).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Separate the gateway (data plane) from the orchestrator (control plane) at the
module level. The gateway runtime files move out of the package root — and the
backend-neutral Gateway lifecycle ABC + GATEWAY_* constants move out of
orchestrator/ — into a new bot_bottle/gateway/ package:
gateway/__init__.py (was orchestrator/gateway.py: Gateway ABC + consts
+ rotate_gateway_ca)
gateway/gateway_init.py (the PID-1 daemon supervisor)
gateway/egress_addon.py, egress_addon_core.py, egress_dlp_config.py,
dlp_detectors.py (the egress mitmproxy daemon)
gateway/git_http_backend.py (the git-http daemon)
gateway/git_gate_render.py (the git-gate pre-receive rendering)
gateway/supervise_server.py (the supervise MCP daemon)
gateway/policy_resolver.py (the data-plane control-plane RPC client)
orchestrator/ now holds only control-plane files. The shared plan/types/auth
layer (egress.py=EgressPlan, git_gate.py=GitGatePlan, supervise.py,
supervise_types.py, control_auth.py) and the launch-time git-gate provisioning
helpers stay at root, so orchestrator/ and backend/ still own them.
Because these daemons are invoked as `python3 -m bot_bottle.<name>`, loaded flat
by mitmproxy, and referenced in Dockerfile.gateway, the move updates more than
Python imports: the `-m` invocations (firecracker/macOS infra scripts), the
Dockerfile.gateway addon shim + ENTRYPOINT, gateway_init's _DAEMONS module
paths, and the git-gate CGI heredocs all now point at bot_bottle.gateway.*.
No behavior change; full unit suite green (2251).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Firecracker infra VM started the control plane with no signing key, so it
ran open and granted every unauthenticated caller the `cli` role. The nft
boundary only fences off the separate agent VM — the egress / supervise /
git-http daemons run in this same VM and reach the orchestrator over 127.0.0.1,
so a compromised data-plane daemon could still drive the operator routes
(approve its own supervise proposals, rewrite policy, read injected tokens).
Generate the signing key on the persistent volume, hand it only to the
orchestrator process, mint a `gateway` JWT for the data-plane daemons, and
mirror the key back to the host token file so the CLI signs `cli` tokens the VM
verifies — the same per-process scoping the docker / macOS launchers use.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
PRD 0070's rule — only the orchestrator opens bot-bottle.db; the data
plane reaches state through the control-plane RPC — was not in force.
Three data-plane daemons held a direct read-write handle on the shared
SQLite file: the supervise MCP server, the egress DLP addon (the most
attack-exposed process, TLS-bumping hostile traffic), and the git-gate
pre-receive hook. An RCE in any of them could read every bottle's
plaintext identity_token and forge attribution fleet-wide (issue #469).
Add the agent half of the supervise flow to the control plane:
POST /supervise/propose -> queue a proposal, 201 {proposal_id}
POST /supervise/poll -> non-blocking decision poll, 200 {status,...}
Both attribute the caller by (source_ip, identity_token) exactly like
/resolve — never a caller-supplied slug — so a bottle can only ever queue
or read its own proposals even if the data plane is compromised. A decided
poll archives server-side, preserving the archive-after-read contract.
Data plane: the supervise server, egress addon, and git-gate hook now
queue/poll through PolicyResolver.propose_supervise / poll_supervise
instead of opening the DB. supervise_server keeps its ~30s grace window
by polling the RPC; egress keeps its safelist keyed by resolved bottle;
the git-gate hook gets (source_ip, identity_token) from the CGI env.
Packaging: drop the DB bind-mount and SUPERVISE_DB_PATH from the
data-plane containers/VMs (docker gateway + infra, macOS infra,
firecracker infra). The orchestrator remains the sole opener of the one
file via BOT_BOTTLE_ROOT / host_db_path().
Update PRD 0070: the rule is now in force; remove the transitional caveat.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Implements the interim secret-provider design (PRD prd-new-secret-provider):
each agent receives a random ENV_VAR_SECRET injected into its container env
at launch. The host uses this key to encrypt each egress auth token value
(HMAC-SHA256 CTR mode, stdlib-only) and store it in a new
bottled_agent_secrets table (one row per env var, key column plaintext for
auditing). The key never touches the DB.
On infra container restart the in-memory token map is lost. launch_consolidated
now calls _reprovision_running_bottles after ensure_running: for each
registered bottle still alive on the gateway network it execs
`printenv ENV_VAR_SECRET` into the agent container and posts the result to the
new POST /bottles/<id>/reprovision_gateway control-plane endpoint, which
decrypts the stored rows and restores _tokens — no manual intervention needed.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
`cleanup()` enumerated *every* firecracker process under the run root and
*every* run dir, so `./cli.py cleanup` would kill running bottles and
delete their rootfs — despite being described as orphan cleanup. The
backend's `enumerate_active` registry is still a stub (#354), so there
was no live/dead signal.
Use the running firecracker processes themselves as that signal: a run
dir with a live VM is protected (never killed, never removed); only run
dirs with no live process (leaked by a hard-killed launch) are reaped,
plus firecracker pids whose run dir is already gone (lingering VMMs).
This makes cleanup safe to run alongside live bottles and lets it back a
periodic GC for the run-dir leak the launch.py teardown fix closes on the
clean-exit path.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Each Firecracker bottle created its run dir (holding an ~1G
rootfs.ext4) but never removed it — launch teardown only terminated
the VM. Every run leaked its rootfs, filling the CI runner's disk
(187 stale sandbox-tester dirs / 141G in one incident).
Register a run-dir rmtree callback on the ExitStack before the
vm.terminate callback so, LIFO, the VM is gone before its rootfs is
removed. ignore_errors keeps teardown resilient.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
pyright strict reportUnusedImport flagged BottleImages in
test_docker_launch_committed_image.py and test_macos_container_launch.py;
neither file references the type by name (they only use the returned
value's attributes).
coverage run's --data-file flag can be overridden or ignored in some
runner environments (Nix Python, older act-based runners). Switching to
the COVERAGE_FILE env var with an absolute ${{ github.workspace }} path
ensures coverage.py writes to a known location in every runner context,
so upload-artifact can find the file.
Also adds --reuse-published to the infra build step: if the artifact for
this content hash already exists in the registry, download it instead of
running the full docker build → mke2fs → gzip pipeline.
Extract the shared CA cert polling loop into `backend/util.poll_ca_cert`
(firecracker and macos backends were duplicating deadline/sleep/raise logic).
Each caller now wraps a fetch lambda and converts TimeoutError to its own
error type. Also corrects the PRD port publication line from {port}:{port}
to {host_port}:8099.
Collapses the two-container Docker model (gateway + orchestrator) into one
bot-bottle-infra container, matching the macOS and Firecracker backends.
- Dockerfile.infra: now a shared gateway+orchestrator base (COPY bot_bottle
from orchestrator build, no CMD override)
- Dockerfile.infra.fc: new Firecracker-specific layer (buildah/crun/netavark)
- gateway_init: adds orchestrator daemon with _OPT_IN_DAEMONS gating so it
only starts when BOT_BOTTLE_GATEWAY_DAEMONS explicitly includes it
- orchestrator/lifecycle: OrchestratorService manages one infra container;
builds orchestrator (intermediate) then infra; live source bind-mounted at
/bot-bottle-src with PYTHONPATH so the subprocess uses the checkout
- backend/consolidated_util: extracts provision_bottle + teardown_consolidated
shared across all three backends; removes duplication in docker/fc/macos
consolidated_launch modules
- firecracker/infra_vm: builds four images (orchestrator→gateway→infra→infra.fc)
- All unit tests updated and passing (1878 tests)
- PRD status: Draft → Active
If resolve_teardown_timeout() raised after launch_consolidated()/register_agent()
returned, the teardown callback was never registered, leaving the bottle
registered with no cleanup path. Resolve the timeout into a local variable
before the registration call so any failure aborts before the bottle exists.
Add tests covering OrchestratorConfigStore, resolve_teardown_timeout priority
ordering (env > db > default), and the source-order invariant that the resolver
runs before registration in all three backends.
Resolves via ENV VAR -> orchestrator DB config -> default (30 s, up from 5 s):
BOT_BOTTLE_ORCHESTRATOR_TEARDOWN_TIMEOUT_SECONDS
teardown_timeout_seconds key in new orchestrator_config table (bot-bottle.db)
New OrchestratorConfigStore (same DbStore/TableMigrations pattern as the
registry) stores the DB-level setting. resolve_teardown_timeout() implements
the priority chain and is called at stack.callback registration time in all
three backends (macos_container, docker, firecracker).
- Remove hardcoded --backend=macos-container flag reference in
firecracker/util.py require_firecracker() error message
- Remove --backend flag from cli.py start; backend selection now
driven exclusively by BOT_BOTTLE_BACKEND env var or auto-selection
- Skip unavailable backends in cli.py cleanup (fixes crash on Linux
when macos-container.prepare_cleanup calls require_container())
- Add two-tier auto-selection: VM backend first (macos-container on
macOS, firecracker on Linux+KVM); fall back to docker with a
security warning and interactive i/d/q prompt; exit if docker
also unavailable and print VM install instructions
Closes#344
Review follow-ups for the infra candidate-artifact flow:
1. publish-infra would fail on the first merge to main. The rootfs version now
hashes the dropbear bytes, and build-infra sets BOT_BOTTLE_FC_DROPBEAR to the
staged dropbear — but publish-infra (ubuntu-latest) set none, so _publish_
bundle re-derived the version with a "<missing>" dropbear and rejected the
candidate as "does not match checkout". Download the same firecracker-inputs
dropbear and export BOT_BOTTLE_FC_DROPBEAR in publish-infra (and add
stage-firecracker-inputs to its needs, since it now consumes that artifact).
2. Guard stage-firecracker-inputs with the same fork-PR check as the other
KVM-runner jobs, so a fork PR can't spin the privileged runner (it only
copies a static binary, but keep the posture consistent; it gates the whole
Firecracker chain via needs).
3. ensure_artifact_gz / _publish_bundle read version.txt before checking it
exists — a bundle missing it raised a raw FileNotFoundError instead of the
intended "bundle is incomplete" die. Guard the read (kept before the gz/sha
completeness check so a wrong-version bundle still reports the version
mismatch, not "incomplete").
4. test_infra_artifact.setUp didn't isolate BOT_BOTTLE_INFRA_ARTIFACT_DIR, so
an ambient candidate dir (the coverage job exports one) would send the
registry-pull tests down the local-bundle path. Pin it off in setUp; the
candidate-path cases set it explicitly.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A9qa3xoavjQScufDfZaXKR
Dockerfile.gateway COPYs pyproject.toml into /src and runs pip install
/src, so it is a real input to the baked rootfs. A dependency-only change
previously reused stale artifact versions, potentially booting a rootfs
whose installed packages differed from the current checkout.
Also adds _fake_repo fixture support and a regression test so this input
can't silently drop out of the hash again.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
build_agent_rootfs_dir cached the built rootfs by Dockerfile content alone,
but util.inject_guest_boot then writes util._GUEST_INIT into it. So a fix to
the init — making /tmp world-writable (1777) so the agent can create scratch
dirs / git worktrees there — did NOT bust the cache: the KVM runner kept
reusing a stale agent-<dockerfilehash> rootfs built with the old init, and the
sandbox-escape README-push test kept failing at `git init /tmp/...` with
"Permission denied".
Key the cache on Dockerfile content AND the injected init (_rootfs_digest), so
an init change rebuilds. Self-busting: the new key yields a fresh cache dir, so
no manual cache clear on the runner.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A9qa3xoavjQScufDfZaXKR
The infra VM is a per-host singleton that outlives short-lived launchers, so
`ensure_running` adopts it when its control plane is healthy. But it adopted
ANY healthy VM regardless of the code that built it — so after an infra-code
change the old VM kept being adopted and the new code never booted. The only
way to dislodge it was an out-of-band `kill`, which then raced whatever
launched next. On CI this meant every infra change needed a manual VM kill.
Make adoption version-aware:
* boot records the infra-artifact version it booted from in a `booted-version`
marker beside the singleton; `stop` clears it.
* `ensure_running` adopts only when the marker matches the current version
(`_adoptable`); a missing/mismatched marker falls through to stop + reboot.
So a stale VM is replaced automatically on the next launch — no manual kill,
and it's concurrency-safe (the reboot happens under the singleton flock).
Also harden teardown: the PID file drifts after crashes / out-of-band kills,
so `stop` now also reaps any orphaned firecracker still bound to the infra
config path (scoped to that path, so interactive-pool VMs are untouched) —
otherwise a survivor holds the orchestrator TAP and the fresh boot dies with
"tap … Resource busy".
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A9qa3xoavjQScufDfZaXKR
The rootless agent rootfs build can land /tmp as 0755/root-owned, so the
agent (uid 1000 node) can't create scratch dirs there. The sandbox-escape
suite's README-push test does `cd /tmp && git init sandbox-escape-repo` and
died with "cannot mkdir sandbox-escape-repo: Permission denied" — before the
git-gate gitleaks hook could run — so the test read it as a missing hook.
On docker the agent inherits node:22-slim's 1777 /tmp, which is why only the
Firecracker path was affected (and only now that the suite runs end-to-end).
Set /tmp to 1777 in the guest PID-1 init, so every agent VM boots with a
usable /tmp regardless of rootfs perm drift.
Also update the infra-init unit test to assert the gateway launches via the
`bot_bottle.gateway_init` module (matching the prior fix), not a file path.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A9qa3xoavjQScufDfZaXKR
The infra VM's init boots the orchestrator control plane AND the gateway data
plane (egress/git-http/supervise). Since 5ad3449 moved the daemons into the
installed `bot_bottle` package, there is no `/app/gateway_init.py` file — the
gateway image's entrypoint is `python3 -m bot_bottle.gateway_init`. But the
firecracker init still spawned the old file path, so the guest logged:
python3: can't open file '/app/gateway_init.py': No such file or directory
The control plane came up (it already used `python3 -m bot_bottle.orchestrator`)
but the gateway never started, so mitmproxy never generated its CA and launch
died with "gateway CA not available after 30s". Mirror the orchestrator line:
run the gateway as `python3 -m bot_bottle.gateway_init` (bot_bottle resolves
from the /app CWD, same as the control plane; egress-entrypoint.sh /
egress_addon.py are present from the gateway base image).
Third build/runtime bug from 5ad3449's package refactor that the Firecracker
integration suite never exercised (the artifact publish was broken, so the
job 404'd before boot). Changes the init, so the infra artifact version moves;
the matching rootfs has been rebuilt and published.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A9qa3xoavjQScufDfZaXKR
Follow-up to the codex review on #398. os.open's mode arg only applies on
creation, so a committed-rootfs.tar.partial left 0644 by an interrupted run
would be opened/truncated (not re-moded) and stay world-readable for the
whole SSH stream. Unlink any leftover and exclusively recreate it
(O_EXCL|O_NOFOLLOW), then fchmod 0600 immediately so umask can't loosen it.
Test pre-creates a 0644 partial and asserts the fd is 0600 mid-stream.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UoEZHDjv84ChoZbozQERhJ
Address the codex review on #398:
- P1: inject_guest_boot no longer follows a symlink at bb-init/bb-dropbear.
A committed snapshot is guest-controlled and could plant those paths as
symlinks aimed at a host file (e.g. bb-init -> ~/.bashrc); write_text /
copy2 would then overwrite the target as the host user during resume.
Replace any pre-existing entry and create the files with
O_EXCL|O_NOFOLLOW so the write stays inside the staging tree.
- P2: write the snapshot tar owner-only (0600). It can contain the bottle's
private workspace; it was being created world-readable (0644).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UoEZHDjv84ChoZbozQERhJ
The last host-Docker dependency in the Firecracker launch path. Freeze
and resume no longer touch the docker daemon, so the backend needs
firecracker + KVM only — completing #348.
Freeze: stream the guest rootfs over SSH straight into a persistent
committed-rootfs.tar (the resumable/migratable artifact) instead of
round-tripping through `docker build` from a scratch image. Written to
a .partial sibling and atomically renamed so a failed freeze leaves no
truncated artifact.
Resume: extract the snapshot tar into a cached base dir and feed it to
the existing rootless `mke2fs -d` pipeline, replacing the
`docker create` + `docker export | tar` path. Recreate the
proc/sys/dev/run mount points the freezer excludes so the guest init
can mount them.
`util.build_base_rootfs_dir` / `docker_image_id` stay — they still back
the opt-in BOT_BOTTLE_INFRA_BUILD=local dev path and off-host
publish_infra, which are out of scope.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UoEZHDjv84ChoZbozQERhJ
Rename the Gitea generic package from bot-bottle-infra to
bot-bottle-firecracker-infra so it's self-evident in the package list which
backend it serves (and leaves room for other artifacts, e.g. a shipped
kernel). The version slot stays the content hash — "firecracker" belongs in
the package name, not the version. Docker image / VM names are unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UoEZHDjv84ChoZbozQERhJ
Generic packages have no description field, so publish_infra now uploads a
short about.txt alongside the rootfs on every publish — it's what identifies
the package as the Firecracker backend's infra rootfs on the package page.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UoEZHDjv84ChoZbozQERhJ
Address PR #395 review (two P1s):
- Version hash covered only `bot_bottle/**.py`, but the image `COPY`s the
whole package — non-Python inputs baked in (egress_entrypoint.sh,
netpool.defaults.env) didn't change the version, so a launch host could
boot a stale rootfs whose code differs from its checkout. Hash every
regular file under bot_bottle/ (excluding __pycache__/.pyc). Regression
tests: a shell-script change bumps the version; .pyc/__pycache__ don't.
- publish_infra `_put` read the whole (hundreds-of-MB) gz into memory via
read_bytes(). Stream it from disk with an explicit Content-Length; the
tiny .sha256 stays in-memory. Test asserts the body is the file object,
not bytes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UoEZHDjv84ChoZbozQERhJ
Drop the fallback to the general-purpose BOT_BOTTLE_CLAUDE_GITEA_TOKEN so
the artifact pull/publish uses a dedicated, package-scoped token that can
be granted (or revoked) independently of the general Gitea token.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UoEZHDjv84ChoZbozQERhJ
Stage 2 of the docker-free Firecracker backend (#348): stop building the
fixed infra image on the launch host. The infra VM's rootfs is host- and
bottle-agnostic (authorized_keys + guest IP ride the kernel cmdline, not the
rootfs), so it's built once off-host and published as a versioned, ready-to-
boot ext4; the launch host downloads + verifies + boots it — no Docker, no
image tooling, just HTTP + gunzip.
- infra_artifact.py: version = content hash of the rootfs inputs (the shipped
bot_bottle package + the three Dockerfiles + the init), so a launch host
pulls the artifact matching its code and a content change can't silently
boot a stale rootfs. Pull + sha256-verify (fail-closed) + gunzip from a
Gitea generic package; base/owner/token configurable, default this Gitea.
- infra_vm.ensure_built/boot default to the pull path; BOT_BOTTLE_INFRA_BUILD=
local keeps the docker build-from-source path for iterating on Dockerfiles.
- publish_infra.py: the off-host half — builds the images with Docker, mke2fs
the rootfs (with buildah slack), gzips, and PUTs it to the generic package.
Rollout note: default=pull means a launch 404s until an artifact is published;
until the Gitea packages endpoint is enabled + an artifact published, use
BOT_BOTTLE_INFRA_BUILD=local. Freeze/migrate's remaining docker use is a
separate PR.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UoEZHDjv84ChoZbozQERhJ
Add BottleBackend.ensure_orchestrator() -> str: the backend-agnostic
entry point that brings up the per-host orchestrator + shared gateway
(idempotent) and returns its host-reachable control-plane URL. Docker
starts the orchestrator + gateway containers (OrchestratorService);
firecracker boots the infra VM; macos-container dies with a pointer
(no orchestrator). Previously bring-up was reachable only through each
backend's consolidated_launch, with no shared handle.
Wire it into `bot-bottle supervise`: supervise is often the first thing
an operator runs, before any bottle has booted the control plane, so
`_resolve_orchestrator_url` now starts the selected backend's
orchestrator on demand when discovery finds nothing, instead of failing
with "launch a bottle first".
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UoEZHDjv84ChoZbozQERhJ
The firecracker supervise MCP daemon 500'd (-32603) on egress-allow/block:
it ran with no BOT_BOTTLE_ROOT/SUPERVISE_DB_PATH, so it targeted a stray,
unmigrated SQLite file and `write_proposal` hit "no such table:
supervise_proposals". Meanwhile the in-VM control plane migrated only the
registry table (orchestrator_bottles) into its own DB, and the host
operator reads a third, disconnected DB — three files, none shared.
Consolidate to one DB per host, owned by the control plane on the
persisted registry volume (/var/lib/bot-bottle/db/bot-bottle.db):
- orchestrator startup now migrates the supervise queue + audit tables
into the same file it migrates the registry into (StoreManager), so the
control-plane DB carries every table.
- the in-VM supervise daemon is pointed at that same file via
SUPERVISE_DB_PATH, so daemon and control plane share one queue.
`list-egress-routes` already worked (no DB); egress-allow now queues +
waits on the single persisted DB instead of erroring. Validated against a
live infra VM: migrate + write_proposal succeeds and the proposal is
queued. The host-operator HTTP bridge (so approvals complete from the
host, unifying docker onto the same path) is the follow-up.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WBMWTEtQdJ4W5UrWuLHCck
The agent-image build ran over ssh with output fully captured, shown
only as a 20-line tail on failure — so a successful (or in-progress)
first build was a long silent wait through the base pull + apt/npm
installs. Stream the `buildah build` step's stdout/stderr straight to
our own (like the docker backend's `docker build`) via a non-capturing
_ssh_streamed helper, so the operator sees `STEP i/n` progress live.
Failure now points at the streamed output above instead of a captured
tail. The smoke test and rootfs export stay captured (short / piped).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WBMWTEtQdJ4W5UrWuLHCck
The interactive agent command is sent to the guest by spreading the
remote argv as separate ssh arguments; ssh space-joins everything after
the host into one line that the guest login shell re-parses. That only
works while every token is a "simple word" — which held for claude
(`--append-system-prompt-file <path>`) but not for codex's
`read_prompt_file` mode, whose positional is a whole sentence:
"Read and follow the instructions in <path>.". The guest shell re-split
it on spaces, so codex received `Read` as the prompt and `and`, `follow`,
… as extra args — failing with `unrecognized subcommand 'and'` the moment
an interactive codex session attached.
Pre-quote each remote token with shlex.quote before ssh joins them (the
same ssh→guest-shell discipline infra_vm/cp_in already use). Simple words
are unchanged, so existing behaviour and the parity/structure tests are
untouched; an arg with spaces now survives as a single argument.
Regression test round-trips the joined remote command back through
shlex.split and asserts codex's prompt comes out as exactly one arg.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WBMWTEtQdJ4W5UrWuLHCck
Codex review: the /31 TAP doesn't make source IP unspoofable, and the
app-layer token was returned by launch but never delivered or enforced, so
a spoofed source could select a victim bottle's policy/tokens. Make the
token mandatory and deliver it on each attributed plane (anti-spoof landed
separately as the network boundary).
Enforcement (control plane):
- `Orchestrator.resolve` now requires a matching (source_ip, identity_token)
pair (constant-time) — no source-IP-only fallback. `/resolve` fail-closes
(403) on a missing/empty/mismatched token.
Delivery, per plane (the token is `token_urlsafe`, safe in a URL):
- egress: proxy credentials (`HTTPS_PROXY=http://bottle:<token>@gw`). The
addon reads `Proxy-Authorization` — from the request (HTTP) or captured at
the CONNECT for HTTPS tunnels (keyed by client conn, cleared on disconnect)
— validates, and strips it (+ the legacy header) before upstream.
- git-http: a URL-scoped `http.<gate>/.extraHeader: x-bot-bottle-identity`
in the agent's git config (only over the http transport).
- supervise: `mcp add --header x-bot-bottle-identity: <token>` (claude +
codex); the server reads the header and passes it to resolve.
Wiring: thread `ctx.identity_token` onto the firecracker + docker plans and
into the agent env/config at launch.
Verified on a KVM host: egress with the correct proxy-cred token returns
200 (HTTP and HTTPS/CONNECT), and no-token / wrong-token return 403; a real
`cli.py start --backend=firecracker` launch provisions git config + the
supervise MCP header and reaches the agent session, all under mandatory
enforcement. Fixed a `claude mcp add` arg-order bug (--header must follow
the positional name/url) found by that launch.
Transparent proxy for tools that ignore proxy env is deferred to a
follow-up (see thread); anti-spoof + host firewall remain the fail-closed
boundary.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WBMWTEtQdJ4W5UrWuLHCck
Codex flagged that parallel `start`s race on shared state: two cold
launches could both stop/build/boot the singleton on the same rootfs/PID,
concurrent builds share the infra VM's buildah store, the cleanup did
`buildah rm -a` (nuking a peer build's container), and the rootfs cache
was populated non-atomically.
- `infra_vm.ensure_running`: a host flock (`singleton.lock`) around the
cold stop/build/boot path, with a double-checked health re-test under
the lock so a second launcher adopts rather than re-boots. The healthy
fast-path stays lock-free.
- `image_builder.build_agent_rootfs_dir`: a host flock (`.build.lock`)
around cache-lookup + build + publish; build into `.building-<digest>`
and publish by atomic `os.rename`, so a partial build never appears as
`agent-<digest>`.
- scoped cleanup: per-build named working containers
(`<tag>-smoke`/`-export`) removed by name instead of `buildah rm -a`;
also cleared before the build to recover from a crashed prior run.
Verified: agent image builds via the locked/atomic path, infra VM stays
healthy, agent boots, `claude --version` = 2.1.172. Full unit suite +
pyright green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WBMWTEtQdJ4W5UrWuLHCck
Addresses the review finding that buildah lived only in the orchestrator
image, so the persistent infra VM wasn't the builder — a separate throwaway
builder VM contended with it for the orchestrator TAP. Consolidate:
- **Rebase the gateway (and thus infra) on `python:3.12-slim` = Debian
trixie**, pip-installing mitmproxy instead of `FROM mitmproxy/mitmproxy`
(Debian bookworm). trixie ships buildah 1.39, which can build agent
Dockerfiles that use heredocs; bookworm's 1.28 can't (`Unknown
instruction: "{"`). CA path is unchanged (set via `--set confdir=`).
- **buildah lives only in `Dockerfile.infra`** now (removed from the
orchestrator image, which is lean/stdlib-only again).
- **Shared orchestrator content**: `Dockerfile.orchestrator` is the single
definition of the control-plane payload; the infra image `COPY --from`s
it (same trixie base → clean copy, and future deps like iroh are added
once). The docker backend runs the orchestrator image directly.
- **`image_builder` builds inside the infra VM** (which now has buildah)
over SSH — no throwaway builder VM, so the `bborch0` contention is gone.
`ensure_built` builds orchestrator + gateway before infra (FROM gateway,
COPY --from orchestrator).
Verified on a KVM host: images build (buildah 1.39 in infra), the agent
image builds *inside* the infra VM (heredoc Dockerfile and all), the infra
VM stays healthy, the agent boots and `claude --version` = 2.1.172. The
rebased gateway still starts as a docker container and generates its CA
(docker backend unaffected).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WBMWTEtQdJ4W5UrWuLHCck
git-gate now works end-to-end for git-upstream bottles on the infra VM.
Three fixes surfaced by driving a real clone through git-http:
- `SshGatewayTransport.cp_into` preserves the source file mode (docker cp
does). The access-hook is staged 0700 and git-http execs it directly; a
plain `cat >` landed it 0644 -> EACCES. Keys stay 0600.
- the infra init runs `BOT_BOTTLE_GATEWAY_DAEMONS=egress,git-http,supervise`:
the VM backend reaches git over git-http (9420), so the git:// daemon
(git-gate, whose /git-gate-entrypoint.sh the consolidated model doesn't
stage) is left out instead of crash-looping.
- `build_infra_rootfs_dir` folds the init's content-hash into the rootfs
cache key, so an init change actually rebuilds the rootfs (the base image
digest alone wouldn't catch it).
Verified on a KVM host: launch_consolidated provisions a git-upstream
bottle's repo + creds into the gateway VM over SSH; an agent VM clones via
git-http (source-IP attributed) and the access-hook resolves the
provisioned key + known_hosts and attempts the upstream SSH fetch —
failing closed only because the test used a dummy key + fake upstream
("refusing to serve stale data"). With real creds the fetch serves.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WBMWTEtQdJ4W5UrWuLHCck
The infra VM's rootfs is ephemeral (rebuilt each boot), so the bottle
registry DB needs durable storage across restarts. Give the infra VM a
firecracker analogue of a docker volume: a host-side ext4 file attached as
a second virtio-block device (guest /dev/vdb), mounted at the control
plane's DB dir (/var/lib/bot-bottle, where host_db_path lives at
db/bot-bottle.db).
- firecracker_vm.boot/_config take an optional `data_drive` (a non-root,
RW second drive).
- infra_vm creates the volume on first use (`mke2fs` an empty ext4 at
<fc-cache>/infra/registry.ext4) and mounts /dev/vdb in the PID-1 init
before the control plane starts. It's a plain ext4 file, so
`sudo mount -o loop <path>` (VM stopped) inspects bot-bottle.db directly.
Verified on a KVM host: register a bottle, restart the infra VM (fresh
rootfs, same volume) — /dev/vdb re-mounts and the registry `db/` dir + a
marker file survive the restart.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WBMWTEtQdJ4W5UrWuLHCck
didericis-claude
didericis
didericis-codex