unsloth/studio/backend/run.py
Daniel Han c608649552
feat(studio): run chats in parallel in the Chat tab (#7455)
* feat(studio): run chats in parallel in the Chat tab

New Chat used to cancel whatever the current conversation was generating.
It now leaves it running, like switching to the Train or Export tab: the
sidebar shows which chats are still going, and Stop is per conversation.

Plain `unsloth studio` launched llama-server with one decode slot, so the
admission queue serialised every chat regardless of what the UI did. Both
entry points now default to the same slot count as `unsloth studio run`.

A model swap still ends every running chat, since they all decode on one
llama-server. /load and /unload now refuse with 409 and name those chats
unless the caller passes force_cancel_active, and the UI asks first.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio): scope the composer tool badge to its own conversation

The green "Running Python: ..." badge above the composer read a single
global store value, so one chat's tool call showed above every other
chat's composer, including a brand-new empty one. Its elapsed counter
also restarted at 0 on every thread switch, and a run ending anywhere
cleared the badge everywhere.

Key the status by thread and store the moment it started, so each
conversation shows only its own tool call and the counter resumes rather
than restarts. Also adds a test that every conversation gets its own
tool sandbox directory, which parallel tool calls depend on.

* Fix stalled tool calls while awaiting approval for PR #7455

Three problems, all from the approval prompt behaving as though only one
chat could ever run.

Arguments were not streamed for a gated call, so the chat stayed blank for
as long as the model took to write the payload, which for a large file is
minutes. Nothing runs before the decision either way, and the code is what
is being approved, so python and terminal now stream their card while
gated. render_html stays suppressed: its card renders the payload.

The status read "Running ..." with a climbing timer while the call had not
started. It now reports that it is waiting for approval, then switches to
running once allowed.

The admission lease was held across the wait, so four unanswered prompts
held all four decode slots and no other chat could start while llama-server
sat idle. A parked run keeps its lease but no longer counts against
capacity.

Measured with four prompts left open: every gated call streamed its code,
none reported running, and a fresh chat answered in 0.4s where it
previously waited 290s and never did.

* Fix duplicated and truncated tool cards for PR #7455

A gated tool call rendered two cards: the provisional one that streams the
arguments, plus a second one keyed by the approval id. Only the second ever
got its tool_end, so the first spun "Running" for the rest of the chat.
Reuse the open part when the approval prompt arrives.

The terminal card also showed nothing but a 60-char trigger label, so a long
heredoc read as no progress at all. It now renders the command the same way
the Python card renders its script, and neither is capped at 10k chars.

Both cells moved inside the collapsible, so one chevron hides the code with
the output and Copy / Download exist only while the card is open. A card
parked on the prompt says so instead of counting up "Running".

* Fix review findings on the parallel-chat gate for PR #7455

Backend:
- /unload rechecks active generations under the lifecycle gate, like /load,
  and lets its 409 through the catch-all instead of rewriting it as a 500.
- /load gates only once _load_model_impl has decided this is a real reload,
  so an Apply on the already-loaded model no longer refuses, and the retry it
  asks for no longer cancels every chat before returning already_loaded.
- The direct /v1/responses stream registers in the cancel registry, so a
  non-forced unload can no longer tear llama-server down under it.
- run_server defaults to the same slot count as the CLI. colab.py calls it
  without the argument, so Colab was still serialising every chat.

Frontend:
- Cancelling a backgrounded chat aborts its own request rather than only
  posting a cancel id, which is the only thing that ends an external-provider
  or audio run.
- The model-swap dialog counts local runs only, and falls back to the backend
  when this tab's map is empty, so a reload or a second tab still gets asked.
- Context usage and the diffusion canvas are scoped to the chat that produced
  them; a compare row reads activity from its member threads.

Tests:
- The extracted-source cancel harnesses supply the active-generations module,
  which the tracked-cancel class now depends on.

* Fix the swap confirmation scope and cancel timing for PR #7455

A forced load cancelled every chat before the model identifier, GPU selection,
training coexistence and download checks had run, so a load that then failed
those checks stopped the chats and replaced nothing. The refusal still happens
early, but the destructive cancel now sits immediately before the teardown it
is paying for, and rechecks under the gate like /unload does.

The swap dialog only reconciled with the backend when this tab looked idle, so
one local chat was enough to hide a second tab's runs. Confirming then sent
force_cancel_active, which cancels every backend run, including the ones the
dialog never mentioned. The backend snapshot is now merged in every time, so
the dialog names what will actually stop. External-provider runs are never
registered there, so the union stays local-only.

Also drops the active-generations docstring claim about restoring sidebar
spinners, which nothing consumes.

* Defer destructive cancels and track every local stream for PR #7455

/unload cancelled the running chats before it had resolved that it unloads
anything. A stale model_path, which a second tab produces routinely, killed
every chat and then no-opped, leaving the resident model up. It now refuses
early and cancels only at each teardown, matching /load.

The swap dialog also stopped every chat locally the moment the user confirmed,
which threw away the two-phase backend behaviour: a load that then failed
identifier resolution, GPU validation or the training guard had already
truncated the replies. The backend now owns the cancel.

Three local streams decoded on llama-server without registering, so a
non-forced unload counted zero generations and tore the server down mid
response: /v1/completions streaming, and the plain and server-tool Anthropic
streams, the first of which is the default /v1/messages path. Note this makes
a non-forced load return 409 during those runs rather than draining quietly,
the same trade the /v1/responses fix made.

The safetensors tool loop still announced a gated call as running while it
waited on a human; only the GGUF loop had been fixed. A source-level parity
test now pins both.

Also drops stopAllChatThreads, which has no callers left.

* Studio: close three load/unload gate races found in review

Re-check the in-flight load guard after the stop-running-chats confirm.
The confirm always GETs active-generations before its zero-running
early-out, so the guard no longer sits atomically ahead of the
reservation and two picks in that window both reached performLoad over
the same refs. ejectModel had the same shape and gets the same re-check.

Reject a sidecar swap immediately before the forced cancel in both load
branches. The previous check was back at the top of preflight, so an
install reserving during identifier resolution, the tier probe, the
training guard or the download check made the post-drain recheck 409 a
load whose chats had already been stopped.

Enter the Anthropic passthrough's cancel tracker inside its body
generator. It was entered eagerly and returned through
_sse_streaming_response, which sets no unstarted_cleanup, so a response
whose body never started left the run registered forever and 409'd every
later non-forced load and unload.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Trim comments across the files this PR touches

Tightens the comments and doc blocks in the backend, CLI, tests and frontend
files changed by this PR: collapses multi-line explanations to a single line
where they still read clearly, and drops the ones the code already says.

No code changes, verified by an AST comparison against the previous commit.

* Studio: defer the destructive cancel and close two gate gaps

Move the forced cancel behind every check that can still reject a swap.
The drain now runs first with the runs it is about to cancel discounted,
so it waits only for inference the cancel cannot end, then the sidecar
check decides, then the cancel fires, then a second drain lets those runs
unwind before teardown. A sidecar install reserving during the drain no
longer 409s a load whose chats have already been stopped.

Track the non-streaming /v1/completions proxy. It was the last local
decode path missing from active_generations, so an unload, which runs no
drain, tore llama-server down under it and force_cancel_active could not
signal it. It now uses the same tracked cancel event and dedicated client
as the OpenAI pass-through.

Skip the client's preliminary unload while chats are generating and let
/load evict at its own post-preflight point instead. Forwarding
force_cancel_active there truncated replies before identifier
resolution, the GPU and training guards and the download check had run.

Keep per-thread context usage so returning to a chat whose background run
finished restores its bar instead of leaving it blank until the next turn.

Make the running-flag clear run-specific. Every run without a resolved
thread id shares the "__default" key, so concurrent compare panes could
clear each other's flag and strand a live stop handle.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: register the embeddings proxy with the swap gate

/v1/embeddings proxied straight through the pooled client with no tracked
cancel event, so it never appeared in active_generations. /unload runs no
idle drain, so a concurrent non-forced unload counted zero generations and
killed llama-server mid-request, and force_cancel_active had no event to
signal. Mirrors the completions proxy: tracked event, dedicated unpooled
client closed by a cancel/disconnect watcher, unregister in a nested
finally so a close failure cannot leave a phantom generation behind.

* Trim comments on the newest changes in this PR

Comments only, no code changes: shorten the ones added by the load-gate
ordering, embeddings and per-thread usage work down to the same density as
the rest of the diff.

* Studio: register the legacy generate stream with the swap gate

/generate/stream built a cancel event but never entered the tracker, so it
was invisible to active_generations. Being in the keep-warm middleware's
inference suffixes only covers /load, which drains; /unload does not, so a
non-forced unload passed the 409 gate and then blocked on the standard
backend's generation lock, and a forced swap had no event to signal.
Registered inside the body generator under a nested finally so a teardown
failure cannot skip the unregister.

The AST contract test asserted the cleanup finally by overwriting its flag
per Try node, so a nested try made the last one win. Accumulate instead,
which is what the existence claim meant.

* Studio: three more swap-gate gaps found in review

Register /audio/generate with the gate. TTS holds the model for the whole
request and /unload runs no drain, so unregistered a non-forced swap counted
zero generations and tore the model down mid-generation; the orchestrator
path only waits 15s for the generation lock, which real TTS exceeds. No
cancel keys: no backend takes a cancel_event for audio, so the event has no
observer and a forced swap still cannot interrupt audio already in flight.

Thread the tracked cancel event into the /v1/responses admission wait. It
was the only admission caller passing None, so a queued run could not be
reached by cancel_all() and a plain /inference/cancel could not stop it at
all. Same omission fixed at the upstream send there and on /v1/completions.

Let an unforced unload of a stale model path reach the no-op check. Before
this PR that request returned 200 and did nothing; the new gate refused it
with 409 for a request that reaches no teardown branch. Gate both refusal
passes on the disjunction of the route's own teardown conditions, including
not is_loaded, so a mid-load GGUF still refuses.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: register the remaining non-streaming decode paths

stream defaults to false on all three of these, so they are the ordinary
shape of their routes, and each holds a local backend for the whole
request. /unload runs no idle drain, so with no registry entry a non-forced
swap counted zero generations and tore the backend down mid-request instead
of returning 409, and a forced one had no event to signal.

Non-streaming /v1/messages: all three helpers ran with an empty registry,
since only the streaming siblings were tracked. Registered at the call site
because the pass-through takes no cancel_event of its own, and with no
cancel keys, matching those siblings.

Non-streaming standard chat and audio-input chat: the trackers in this route
sit inside their `if payload.stream:` arms, so neither else branch was
covered. The GGUF sibling already registers its own non-streaming branch.

Each exit is in a finally on the branch's existing try, so the except arms
are covered too: a leaked entry 409s every later swap until restart.

* Studio: tighten the swap-gate comments

Comment-only pass over the newest swap-gate registrations: collapse the multi-line rationales in /unload, the legacy generate stream, audio generation and the non-streaming chat branches, and the matching test preambles, to the shortest form that still carries the reason. No code changes.

* Studio: stop the reselect dialog promising a stop that never happens

Picking an external provider leaves the local model resident and stops the
status poll mirroring it, so reselecting that model showed the stop-chats
dialog, and /load then answered already_loaded ahead of its cancel hook.
Confirmed with the live backend: the same pick with force_cancel_active set
still returned already_loaded and the chat kept streaming. Not stopping
those chats is right, since the load never interrupts them, so remove the
prompt rather than honour it. Blanket-skipping is unsafe, because the same
id and variant with one sampling setting changed is a real reload and 409s,
so the branch only fires when a status fetch confirms the resident
checkpoint and variant match, and then adopts it without calling /load.

Redact native model paths from the active-generations response. Registering
/generate/stream recorded backend.active_model_name verbatim, which is an
absolute path for a native local model, and this route is the only place
that serialises it. Redacting at the response covers every tracker rather
than the one that surfaced it.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: keep hydrated context usage in the per-thread map

The history loader restores a saved conversation's usage through
setContextUsage only, and it runs once per mount, so switching away and
back left the bar blank for a hydrated chat even after the per-thread map
landed. setContextUsage now writes the value through to the visible
thread's own entry and clears that entry when passed null, which covers
both hydration call sites and any future writer.

* Studio: unblock load cancellation and share unresolved thread keys

Run the two stop-loading fast paths ahead of the unload route's pre-gate
refusal. _unload_may_evict returns True for exactly the model being
cancelled, so the refusal was blocking the branch that cancels a load which
has replaced nothing and can interrupt no chat. The client made that
unrecoverable: cancelLoading sends the unload without force, drops the
result, and its abort never reaches /load, which takes no signal, so the
load ran on and could later cancel those chats and swap the model. Nothing
else is exempted; an unload that would tear down a serving model matches
neither fast path and still 409s. The comment claiming the client lets that
409 surface is corrected, since it discards it.

Hold every owner behind a shared thread key. Runs with no resolved thread id
share "__default" (concurrent compare panes, since startCompare clears
activeThreadId), so a single owner slot let a second run replace the first's
token and then delete the shared entry while it was still generating, and
the server-cancel map lost the older handle the same way. Both now hold a
list, the running and local flags survive until the last owner clears, and
stopChatThread stops every handle under the key.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: carry a confirmed swap into the sidecar install, key restored usage by thread

Picking a model that needs a newer transformers while chats generate raised the
"stop N chats" prompt, but the answer never reached the install that runs before
the load: /install-latest-transformers refused on those same chats and took no
force flag, so Retry hit the same 409 and nothing in the flow stopped them.

Carry force_cancel_active through the consent dialog into the installer. Only
the pre-gate fast path is skipped: the recheck under the lifecycle gate still
has to pass, so an unconfirmed caller is refused as before. The cancel runs last
inside the gate, after every check that can still reject the install, and the
drain behind it is bounded since it holds the gate and the sidecar reservation.

Also key restored context usage by the thread the loader read. history.load()
captures remoteId before two awaited round trips, so a switch inside that window
filed one thread's usage under another and setActiveThreadId kept re-applying it.

Preserve sibling owners when a run key is cleared without an owner: the image
rejection gate now uses its own token, and the reducer leaves owned runs alone.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: bound the post-cancel drains, and make cancellation reach the paths that ignored it

A forced swap cancels the chats it interrupts, then waits for them to unwind.
That wait had no deadline while holding the lifecycle gate, and TTS on the
subprocess backend observes no cancel event at all, so one audio generation
could pin every load, unload and new request for its whole duration. Bound both
post-cancel drains. Pre-cancel drains stay unbounded: the swap can still be
refused there, so shortening them would weaken what they protect.

/unload had the opposite problem and no drain at all, cancelling and tearing
down on the next line, which turned a clean stream end into a dropped
connection. Give it the same bounded wait, gated on the cancel having cancelled
something so an idle Eject pays nothing.

Make the cancel actually land where it can. GGUF TTS now takes a cancel_event
and a watcher closes its client to break the blocking POST. The Anthropic
non-streaming pass-through did the same thing the completions and embeddings
paths used to: register with the gate, then run both POSTs on the pooled client
that cannot be closed. It now uses a per-request client like they do.

Also: park and unpark the admission queue the reservation actually holds, since
queues are keyed by base_url and a reload mints a new port; key tool output by
remoteId on both sides, so the first turn of a New Chat stops writing under one
key and reading another; and give tool status a run owner, so a finishing run
cannot blank the badge a concurrent one is still showing.

Clamp --parallel to 1 on a llama-server without --kv-unified. The new default of
4 would otherwise split -c four ways on such a build, quartering the context
window for a feature it cannot serve.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: scope a chat's Stop to its own generation, and clear the way before a confirmed install

Safetensors generation is serialized on _gen_lock and the worker has a single
cancel event, so a chat still queued on that lock owns no generation. Its Stop
handler called reset_generation_state() anyway, which set the shared event and
ended whichever conversation was actually running. Parallel chats is what makes
that reachable.

_generate_inner now records its cancel_event as the current holder once it takes
the lock, and reset_generation_state drops a reset from anyone else. Every route
call site passes its own request event. A reset with no event stays global, so
unload and model switch cannot leave a generation alive, and a reset while
nothing runs still resets, so an error path before generation is not a no-op.
The other two backends take the argument too, or the standard one raises
TypeError on every cancel.

The sidecar install had the mirror of the /load ordering problem: it cancelled
the chats first and drained second, so an unrelated counted request the cancel
cannot reach (a count_tokens, say) was still there for the recheck, which then
refused an install that had already stopped every chat for nothing. Drain the
unreachable remainder first, discounting the registered chats, then cancel.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: close the windows the previous round's fixes left open

Three follow-ups, two of them holes in the fixes just before them.

The worker claim went in after _send_cmd, so the command was already running
unclaimed and a queued chat's Stop in that window still reset it. Claim first,
with the send inside the same try, so a failed send releases it too.

Tool status kept one entry per key with an owner. That stops a foreign clear but
not an overwrite: under the shared unresolved-thread key the second run replaced
the first's entry, and its own clear then removed the only one while the first
tool was still running. Keep per-run entries and render the newest.

/unload gated its drain on having cancelled something, so a request that passed
the keep-warm middleware but had not reached its tracker yet was invisible to it
and the teardown landed on an already-admitted request. Drain on the middleware
count instead, which covers that window as well as the cancelled runs, then
re-cancel whatever registered while waiting. Bounded, not a refusal: an unload is
deliberate, and on expiry it proceeds exactly as before.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Trim the parallel-chats comments to their reasons

Compress the multi-line rationales added by this branch into shorter forms and drop
restatements of the code below them. The reasons behind the drain bounds, the deferred
cancel, the per-request generation ownership and the thread-scoped tool and usage keys
are kept, just said in fewer lines.

* Studio: own the worker per generation, and make a resumed chat requeue for its slot

Ownership was a single lock holder, so dispatched runs (compare mode bypasses
_gen_lock by design) never claimed it and the guard fell straight through to the
global reset: a Stop on one of them ended its siblings. Track the generations
actually running instead, claimed before the send and released in the same
finally on both paths. A reset still proceeds when nothing is running, so an
error path ahead of generation is not swallowed.

park() hands the freed slot to a waiter, so a chat resuming from a tool approval
could take it back while that waiter was still decoding, putting two holders on
a one-slot server and sending the resumed tool loop past the admission limit.
unpark_async waits for room; the plain unpark stays for a holder tearing down,
which will not decode again.

Audio only observed its cancel event on a forced swap. An explicit Stop just
aborts the fetch, and this route has no cancel id, so llama-server ran on to the
request timeout after the chat reported it stopped. Watch the disconnect.

Also read tool status by remoteId, matching the key the adapter writes and the
fix already made for tool output, and stop an unresolved run from writing its
usage into whichever conversation the user moved to.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: let only the generation the worker is running speak for it, and hold a slot for a resuming chat

The ownership list recorded admission, but the subprocess runs generations one
at a time, so a dispatched request queued behind another counted as an owner and
its Stop signalled the shared cancel event, ending the request that was actually
running. Keep admission for release bookkeeping and gate ownership on execution
instead, promoted when the worker first answers that request. Nothing executing
still permits a reset, so an error path ahead of generation is not swallowed.

The worker has one cancel event and no per-request cancellation, so this decides
who may pull the lever rather than making the lever per-request.

A resuming chat also polled for a slot it could never see: release() grants to
the next waiter under the same lock, so later arrivals overtook an approved chat
indefinitely. A pending unpark now reserves the next slot and they queue behind
it.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: cover the prefill window, and keep a first turn's tool output readable

Gating worker ownership on execution left the interval between the send and the
first response uncovered: nothing is executing then, and the empty case admitted
anyone, so a queued chat's Stop still ended the one in prefill. Split the empty
case. Nothing claimed at all still permits a reset, so an error path ahead of
generation is not swallowed; claimed but unanswered resolves to the oldest
claim, which is what a FIFO command queue is working on.

Putting both sides of the tool-output scope on remoteId left the first turn of a
New Chat writing under the unresolved scope for its whole life while the readers
recomputed the moment the autosave assigned an id, so the card blanked mid-run.
The readers now fall back to the unresolved scope, which only an unpersisted
first turn can occupy.

* Studio: order the parked approvals, and tie a worker claim to its enqueue

The reservation added for admission fairness was a bare count, so every approved
holder counted against every other: park two chats, approve both, and once the
last decoder released, nothing could ever satisfy the check again. That is a
deadlock where the problem it fixed was only unfairness. Make it a FIFO ticket
so a pending unpark blocks the ones behind it and no others.

_owns_worker reads claim order to decide which request the worker is prefilling,
which only holds if claiming and enqueuing cannot interleave. Hold one lock
across both on the dispatched and the locked path.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: follow a first turn's run when its thread gets an id, and key the denoising canvas per chat

A run started before its thread existed filed every handle under "__default". Nothing
moved them once autosave assigned the real id, so the sidebar row showed no spinner and
Stop could not reach the generation, which kept holding a slot.

adoptDefaultThreadRun re-keys the run maps onto the real id from the thread adapter's
initialize(), where the id first exists; anything already filed under that id wins, since
that is a later run. The adapter captures its key once at run start, so it now resolves
the live key per use through runKeyForOwner, looking its own serverCancel up in the owner
map. Without that the migrated entries are stranded and the spinner never clears.

The denoising canvas was one global slot, so two diffusion chats overwrote each other and
the ownership tag then hid the visible preview until that thread emitted again. It is now
activeDiffusionCanvasByThreadId, written and cleared per thread, and the frame no longer
carries a threadId of its own. The bubble reads threadListItem.remoteId, dropping the dead
threadListItem.id arm: the writer tags unstable_threadId, which is exactly remoteId.

Two existing backend tests needed the same treatment. _bare_orchestrator skips __init__, so
it now sets the claim bookkeeping the worker ownership check reads. The Anthropic
passthrough gate test anchored on comment prose that a rewrap had broken; it anchors on the
code instead.

* Studio: hand the worker over cleanly between generations, and stop unresolved runs sharing each other's state

Worker ownership moved off the consumer and onto the dispatcher. Consumers read their
mailbox whenever they get around to it, so a request whose gen_done had been routed still
owned the worker while the next one ran, and a late Stop for it cancelled that one. The
dispatcher is the only place responses arrive in the order the worker produced them: it
now retires a request at its terminal response and promotes the next one, and answering a
request makes it the sole executor, since the subprocess runs one generation at a time.

reserve()'s immediate path ignored the unpark tickets that _grant_waiters_locked already
honours, so a request arriving between a slot freeing and an approved chat's next poll
took it, repeatedly. It applies the same reservation now.

Three places let concurrent first turns share state through the "__default" key. Nothing
links a run filed there to the id its thread later receives, so rather than guess, each
now declines when the key is ambiguous: adoption only re-keys a lone run, the composer
badge only claims a lone status, and the tool-output fallback only applies to a thread
that is still running. That leaves two concurrent first turns where they were before
adoption existed instead of handing one thread the other's handles.

A first turn's usage was never filed, because its key stayed null for the whole run while
autosave moved activeThreadId to the real id, so the context bar went blank after the
first reply. It resolves the adopted key like the cleanup handles do.

Cancelling a forced load left the UI with no model: the previous one stays resident until
/load's teardown, and the cancel path cleared the checkpoint without rolling back. It now
resyncs from the backend, which is right whether or not the load got that far.

The sidecar install drain is weighted 1:4 rather than halved, total unchanged. Only the
second half benefits from patience, and cutting it short refused installs whose chats had
already been stopped for nothing.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: give a first turn its real thread id before the run starts

A first turn filed every run handle under a shared unresolved key because
assistant-ui binds unstable_threadId before the thread is persisted. Two of them
overlapping there is unresolvable afterwards, and the last round's migration could
only decline rather than guess, which left neither sidebar row showing its run.

The id is available earlier than I claimed. append() already tracks
threadListItem.initialize() by the user message id, and createPersistedRunAdapter
already awaits that promise before invoking the adapter, so the thread is persisted
by the time the run begins. It was only being discarded: the tracked promise resolved
to void. It now resolves to the assigned id, and the wrapper hands it to the adapter
when assistant-ui had none. An id that is already set is never replaced, since that
would move a running chat's handles out from under the row watching them. The
existing unresolved-key guards stay as a safety net but should no longer carry weight.

The sidebar counted running thread ids rather than rows, so one compare conversation
read as two chats. It folds ids into rows through the same threadIds the row spinner
uses, and still counts a running id that matches no row.

_TrackedCancel always registered kind="chat", so an embeddings or raw completions
request appeared in the model-swap prompt as an unnamed conversation and confirming
cancelled it while calling it a chat. The non-conversation routes now pass their own
kind, and the prompt says "requests" whenever the snapshot is not all chats.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: withhold the shared worker cancel from a request the worker has left

Moving ownership to the dispatcher fixed reset_generation_state, but the token loop
signals the shared worker event directly and did not carry the same rule. A dispatched
consumer runs with mark_started off and can still be draining tokens buffered before
its gen_done was routed, so stopping it there ended whichever request the worker had
started next.

It now signals only when _owns_worker agrees, the same predicate reset_generation_state
uses. The local drain and return are unconditional, since those touch nothing but this
stream. The remaining _cancel_generation callers are deliberately global: subprocess
shutdown, the pre-load kill and unload_model.

* Studio: add the AGPL-3.0 header to the first-turn identity test

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: stop the dispatcher and a _gen_lock stream fighting over the response queue

Nothing stopped the dispatcher starting under a _gen_lock generation, so once compare
was opened while an ordinary chat was still streaming, both consumed _resp_queue and
whichever response the dispatcher took without a mailbox was dropped, gen_done included.
That chat truncated or hung. This PR is what makes it reachable, since navigating into
compare no longer ends the chat behind it.

Delaying the dispatcher would serialise compare behind whatever chat happens to be
streaming, so the direct readers get a mailbox instead. _direct_reader returns a reader,
a cancel drain and a release, and files the mailbox under _direct_mailboxes rather than
_mailboxes, which means "compare requests are in flight" to the unload and distributed
paths and must not count an ordinary chat.

Both directions close. The dispatcher finds the direct reader's mailbox instead of
dropping. And this reader can already be blocked on the queue when a compare request's
dispatcher starts, so a response that is not ours goes to its own mailbox rather than
being consumed, which would have corrupted the chat and hung the pane. All three
_gen_lock readers use it, and the cancel drain goes through it too.

The sidebar's return target still picked a raw pane id while the count grouped by row,
and /chat addresses compare with `compare`, not `thread`. It resolves through the same
items now, so a running compare row returns to its pair.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: keep worker ownership honest across audio, API traffic and a replaced worker

The audio-input send got a mailbox last round but stayed unclaimed, so a compare request
queued behind it looked like the oldest owner and stopping that queued request signalled
the shared event into the audio chat. It claims under the send lock and releases in the
finally, like _generate_inner.

Ownership is keyed on cancel-event identity with nothing tying it to a worker generation,
so a consumer still blocked on its mailbox when the process was replaced stayed recorded
as the executor, and a generation on the fresh worker could not be stopped.
_shutdown_subprocess clears that state once the process is confirmed dead, mailboxes
included: nothing routes to them again, and a stale one reads as compare activity to the
unload path. Not on the survived-SIGKILL path, which keeps its handle on purpose.

The four public /v1/messages trackers were registering as chats. The distinction is a
Studio thread, not the protocol, and those branches already say "No thread_id: public API
surface" while the Studio path passes payload.thread_id separately. They carry their own
kind now, so the swap prompt stops calling an external request a chat.

The swap confirmation still counted raw pane ids, so a compare conversation asked to stop
two chats and listed its title twice. It folds panes onto pairId and lowers the count by
what it collapsed, leaving a first turn the backend can count but not name.

Deep Research set runningByThreadId but registered no server-cancel handle, and that map
is how Stop, archive and delete reach a thread that is no longer active. Leaving the
outgoing thread running is this PR's doing, so the run was left unreachable while its
supervisor kept working against a conversation the user could delete.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: tighten the parallel-chats comments

* Studio: replay a Deep Research stop that arrived before the run existed

The handle is registered before createResearchRun resolves because the thread can be
stopped while that request is in flight, but it had no id to act on and dropped the stop.
The supervisor then followed a run the user had already stopped, archived or deleted.

It latches instead: a stop with no id yet sets a flag, and the adapter replays it against
the id the moment creation returns rather than starting to follow.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: fix worker ownership on a raced reroute, and the stop-chats prompt

Four review findings on the parallel-chats work, all reproduced first.

- _direct_reader hands a foreign response to its own mailbox, but skipped the
  ownership move the dispatcher makes. A _gen_lock reader already blocked on
  resp_queue can beat the compare dispatcher to that request's first response,
  and the compare consumer opts out of marking, so nothing promoted it: the
  direct request stayed the recorded executor, its late reset cancelled the
  compare generation, and the compare chat's own Stop was ignored.
- A chat stopped while queued on _gen_lock was still claimed and sent once the
  lock freed. Cancellation is only checked on a token, so a long prefill, or a
  generation reaching gen_done without one, occupied the worker after Stop.
  Same hole in the audio-input path, which shares the lock.
- The stop-chats prompt counted generation handles, not conversations. One chat
  holds several while a tool continuation registers its next leg before the
  previous unwinds, so it offered to stop two chats and listed one title.
- Ejecting a model confirms through that dialog, which told the user
  "Unloading the model reloads the model" and offered "Stop and reload".
  Confirming calls /unload and leaves nothing loaded.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: name the TTS run's thread so the stop prompt counts it once

The audio branch registers its run locally under the thread key but sent no
thread_id, so the backend tracker filed the same generation under no thread.
The stop-chats prompt then had a named local run and an unnamed backend one and,
since e8e7594 started adding unnamed entries to the named ones, counted a single
TTS chat as two requests. The backend already reads payload.thread_id, so
sending it lines both registries up on the same run.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2026-07-28 04:40:38 -07:00

2009 lines
78 KiB
Python

# SPDX-License-Identifier: AGPL-3.0-only
# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
"""Run script for Unsloth UI Backend.
Self-contained; can be moved to any directory.
"""
import os
import sys
import time
from pathlib import Path
from typing import Optional, Tuple
def _fix_torch_cuda_ld_path():
"""Prepend torch's bundled CUDA libs to LD_LIBRARY_PATH.
PyTorch wheels ship their own CUDA runtime (libcudart, libcublas, ...) in
``site-packages/nvidia/*/lib``. On Linux the dynamic linker reads
LD_LIBRARY_PATH before the RUNPATH baked into torch's .so files, so a
pre-existing LD_LIBRARY_PATH pointing at a different system CUDA (e.g.
/usr/local/cuda-13/lib64 from conda or a Docker base image) shadows torch's
libs and triggers "undefined symbol" errors when torch is imported. Detect
torch's lib dirs (without importing torch) and prepend them. Returns True if
LD_LIBRARY_PATH was changed.
"""
if sys.platform != "linux":
return False
ld_path = os.environ.get("LD_LIBRARY_PATH", "")
if not ld_path:
return False
try:
import importlib.util
spec = importlib.util.find_spec("torch")
if not spec or not spec.origin:
return False
torch_dir = os.path.dirname(spec.origin)
site_pkgs = os.path.dirname(torch_dir)
nvidia_dir = os.path.join(site_pkgs, "nvidia")
lib_dirs = []
torch_lib = os.path.join(torch_dir, "lib")
if os.path.isdir(torch_lib):
lib_dirs.append(torch_lib)
if os.path.isdir(nvidia_dir):
for sub in sorted(os.listdir(nvidia_dir)):
lib = os.path.join(nvidia_dir, sub, "lib")
if os.path.isdir(lib):
lib_dirs.append(lib)
if not lib_dirs:
return False
existing = ld_path.split(":")
if existing[: len(lib_dirs)] == lib_dirs:
return False # already at the front, nothing to do
torch_set = set(lib_dirs)
cleaned = [p for p in existing if p not in torch_set]
os.environ["LD_LIBRARY_PATH"] = ":".join(lib_dirs + cleaned)
return True
except Exception:
return False
_LD_FIXED_SENTINEL = "_UNSLOTH_STUDIO_LD_FIXED"
def _maybe_reexec_for_cuda_ld_path():
"""Re-exec once so the dynamic linker sees the corrected LD_LIBRARY_PATH.
LD_LIBRARY_PATH is read at process start, so editing os.environ in-process
cannot fix the running interpreter; a single re-exec is required. Call only
from a true entry point (the ``if __name__ == "__main__"`` block), never at
import time, because os.execv replaces the whole process (an embedder such
as Colab that does ``from run import run_server`` must not be re-exec'd).
"""
if _LD_FIXED_SENTINEL in os.environ:
return
if not _fix_torch_cuda_ld_path():
return
os.environ[_LD_FIXED_SENTINEL] = "1"
argv = getattr(sys, "orig_argv", None) or [sys.executable, *sys.argv]
os.execv(sys.executable, argv)
# Suppress C-level dependency warnings globally (e.g. SwigPyPacked).
os.environ["PYTHONWARNINGS"] = "ignore"
# Add the backend dir to sys.path early so local modules import.
backend_dir = Path(__file__).parent
if str(backend_dir) not in sys.path:
sys.path.insert(0, str(backend_dir))
from utils.cpu_threads import configure_cpu_threads
try:
configure_cpu_threads()
except ValueError as exc:
configured = os.environ.get("UNSLOTH_CPU_THREADS")
raise SystemExit(f"Error: Invalid UNSLOTH_CPU_THREADS value {configured!r}: {exc}") from None
# Anaconda/conda-forge Python: seed platform._sys_version_cache before imports
# that trigger attrs -> rich -> structlog -> platform crash.
# See: https://github.com/python/cpython/issues/102396
import _platform_compat # noqa: F401
from loggers import get_logger
from startup_banner import print_studio_access_banner, print_studio_stop_hint
logger = get_logger(__name__)
DISABLE_PUBLIC_CHECK_ENV = "UNSLOTH_STUDIO_DISABLE_PUBLIC_CHECK"
def public_check_disabled() -> bool:
"""True when the operator has turned off the third-party startup lookups.
On a wildcard bind Unsloth asks ifconfig.me for the public IP and check-host.net
whether the port is reachable. Both are useful for sharing a Studio but both tell
an outside service this machine is running one, which lab and privacy-sensitive
deployments do not want (#7307 Problem 8). Set the var to opt out.
"""
return os.environ.get(DISABLE_PUBLIC_CHECK_ENV, "").strip().lower() in {"1", "true", "yes"}
def _resolve_external_ip() -> str:
"""Resolve the machine's external IP address.
Tries, in order:
1. GCE metadata server (instant on Google Cloud VMs)
2. ifconfig.me (anywhere with internet, skipped by UNSLOTH_STUDIO_DISABLE_PUBLIC_CHECK)
3. LAN IP via UDP socket trick (fallback)
"""
import urllib.request
import socket
# 1. GCE metadata server (<10ms on GCE, times out fast elsewhere).
try:
req = urllib.request.Request(
"http://metadata.google.internal/computeMetadata/v1/instance/network-interfaces/0/access-configs/0/external-ip",
headers = {"Metadata-Flavor": "Google"},
)
with urllib.request.urlopen(req, timeout = 1) as resp:
ip = resp.read().decode().strip()
if ip:
return ip
except Exception:
pass
# 2. Public IP service. Third-party, so skippable; the LAN address below still works.
if not public_check_disabled():
try:
with urllib.request.urlopen("https://ifconfig.me", timeout = 3) as resp:
ip = resp.read().decode().strip()
if ip:
return ip
except Exception:
pass
# 3. Fallback: LAN IP via UDP socket trick
try:
s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
s.connect(("8.8.8.8", 80))
ip = s.getsockname()[0]
s.close()
return ip
except Exception:
return "0.0.0.0"
def _install_uvicorn_startup_log_rewrite(bind_host: str, display_host: str) -> None:
"""Rewrite Uvicorn's startup log line: swap wildcard bind for the
externally-reachable address, use our Mac-aware stop hint, and rename the
prefix to "Unsloth Studio running on"."""
import logging
import re
rewrite_host = (
bind_host in ("0.0.0.0", "::") and bool(display_host) and display_host != bind_host
)
new_suffix = "(To stop: press Ctrl+C -- on macOS, Control+C not Command+C)"
old_suffix_re = re.compile(r"\(Press CTRL\+C to quit\)")
old_prefix = "Uvicorn running on "
new_prefix = "Unsloth Studio running on "
def _rewrite(text: str) -> str:
if text.startswith(old_prefix):
text = new_prefix + text[len(old_prefix) :]
return old_suffix_re.sub(new_suffix, text)
class _UvicornStartupRewrite(logging.Filter):
def filter(self, record: logging.LogRecord) -> bool:
try:
msg = record.msg if isinstance(record.msg, str) else ""
if (
msg.startswith(old_prefix)
and isinstance(record.args, tuple)
and len(record.args) >= 3
):
if rewrite_host and record.args[1] == bind_host:
record.args = (
record.args[0],
display_host,
record.args[2],
*record.args[3:],
)
record.msg = _rewrite(msg)
cmsg = getattr(record, "color_message", None)
if isinstance(cmsg, str):
record.color_message = _rewrite(cmsg)
except Exception:
pass
return True
f = _UvicornStartupRewrite()
for name in ("uvicorn", "uvicorn.error"):
logging.getLogger(name).addFilter(f)
def _local_port_open(
host: str,
port: int,
timeout: float = 1.0,
) -> bool:
"""True iff a TCP connection to (host, port) succeeds within timeout."""
import socket
try:
with socket.create_connection((host, port), timeout = timeout):
return True
except OSError:
return False
def _working_local_url(port: int) -> "str | None":
"""A working loopback URL on this machine, or None if neither 127.0.0.1 nor
::1 responds. Fallback when external reachability fails."""
if _local_port_open("127.0.0.1", port):
return f"http://127.0.0.1:{port}"
if _local_port_open("::1", port):
return f"http://[::1]:{port}"
return None
def _localhost_ipv6_mismatch_url(bind_host: str, port: int) -> "str | None":
"""Return the IPv4 loopback URL when localhost won't reach 127.0.0.1.
Local Unsloth binds to 127.0.0.1. Where localhost resolves to IPv6 only (::1),
http://localhost:<port> fails (or hits a different process on ::1) even though
http://127.0.0.1:<port> works. Return the IPv4 URL for the caller to surface.
"""
import socket
if bind_host != "127.0.0.1" or not port or port <= 0:
return None
ipv4_url = f"http://127.0.0.1:{port}"
# Only warn once Unsloth is confirmed answering on IPv4 loopback.
if _working_local_url(port) != ipv4_url:
return None
try:
addr_info = socket.getaddrinfo("localhost", port, socket.AF_UNSPEC, socket.SOCK_STREAM)
except Exception:
return None
if not addr_info:
return None
has_ipv4_loopback = False
has_ipv6_loopback = False
for family, _, _, _, sockaddr in addr_info:
if family == socket.AF_INET and sockaddr and sockaddr[0] == "127.0.0.1":
has_ipv4_loopback = True
elif family == socket.AF_INET6 and sockaddr:
host = sockaddr[0].split("%", 1)[0]
if host == "::1":
has_ipv6_loopback = True
# A connection to ::1 is NOT evidence Unsloth is reachable there: Unsloth binds
# 127.0.0.1 only, so anything on ::1 is a different process. Dual-stack
# localhost is fine (browsers fall back to 127.0.0.1), so only the IPv6-only
# case strands the user.
if has_ipv6_loopback and not has_ipv4_loopback:
return ipv4_url
return None
def _stdout_color_ok() -> bool:
"""Whether to emit ANSI color codes on stdout. Mirrors startup_banner."""
if os.environ.get("NO_COLOR", "").strip():
return False
if os.environ.get("FORCE_COLOR", "").strip():
return True
try:
return sys.stdout.isatty()
except (AttributeError, OSError, ValueError):
return False
def _print_localhost_ipv6_mismatch_warning(local_url: str, port: int) -> None:
"""Warn that localhost points at ::1 while Unsloth is bound to 127.0.0.1."""
use_color = _stdout_color_ok()
warn_c = "\033[38;5;215;1m" if use_color else ""
reset = "\033[0m" if use_color else ""
print(
f"{warn_c} Warning: localhost resolves to IPv6 (::1), but Unsloth "
f"Studio is listening on 127.0.0.1 only. Open {local_url} instead of "
f"http://localhost:{port}.{reset}",
flush = True,
)
def _verify_global_reachability(display_host: str, port: int) -> None:
"""Probe check-host.net to confirm display_host:port is reachable from the
public internet. Synchronous so output lands between the banner URLs and the
stop hint. Bounded at ~15s; failures swallowed (verifier failing != Unsloth
failing). Only meaningful for a wildcard bind, and skipped entirely by
UNSLOTH_STUDIO_DISABLE_PUBLIC_CHECK."""
global _public_reachable
# Reset to "unknown" each run; set True/False only when the probe decides.
_public_reachable = None
import ipaddress
import json
import time
import urllib.error
import urllib.parse
import urllib.request
if not display_host or display_host in ("0.0.0.0", "::"):
return
use_color = _stdout_color_ok()
dim = "\033[38;5;245m" if use_color else ""
ok_c = "\033[38;5;120;1m" if use_color else ""
err_c = "\033[38;5;203;1m" if use_color else ""
warn_c = "\033[38;5;215;1m" if use_color else ""
local_url_c = "\033[38;5;108;1m" if use_color else "" # matches banner's URL color
reset = "\033[0m" if use_color else ""
url = f"http://{_url_host(display_host)}:{port}"
# Private/loopback/link-local addresses aren't globally routable.
try:
addr = ipaddress.ip_address(display_host)
if addr.is_loopback or addr.is_private or addr.is_link_local:
_public_reachable = False
print(
f"{dim} Note: {display_host} is a private/LAN address -- "
f"reachable on this network only, not from the public internet."
f"{reset}",
flush = True,
)
return
except ValueError:
# Not an IP literal; probe by hostname.
pass
# The probe hands display_host:port to a third party and asks it to connect.
if public_check_disabled():
logger.debug("Skipping the check-host.net probe (%s).", DISABLE_PUBLIC_CHECK_ENV)
return
try:
qs = urllib.parse.urlencode({"host": f"{display_host}:{port}", "max_nodes": 3})
req = urllib.request.Request(
f"https://check-host.net/check-tcp?{qs}",
headers = {
"Accept": "application/json",
"User-Agent": "unsloth-studio-reachability/1",
},
)
with urllib.request.urlopen(req, timeout = 5) as resp:
init = json.loads(resp.read().decode("utf-8", errors = "replace"))
req_id = init.get("request_id")
if not req_id:
return
results = {}
deadline = time.monotonic() + 15.0
poll_req = urllib.request.Request(
f"https://check-host.net/check-result/{req_id}",
headers = {
"Accept": "application/json",
"User-Agent": "unsloth-studio-reachability/1",
},
)
while time.monotonic() < deadline:
time.sleep(1.5)
try:
with urllib.request.urlopen(poll_req, timeout = 5) as resp:
results = json.loads(resp.read().decode("utf-8", errors = "replace"))
except Exception:
continue
if results and all(v is not None for v in results.values()):
break
# Two decisive nodes is enough; stop early.
decisive = [
v
for v in results.values()
if isinstance(v, list)
and v
and isinstance(v[0], dict)
and ("time" in v[0] or "error" in v[0])
]
if len(decisive) >= 2:
break
ok_nodes = err_nodes = 0
for v in results.values():
if not isinstance(v, list) or not v or not isinstance(v[0], dict):
continue
if "time" in v[0]:
ok_nodes += 1
elif "error" in v[0]:
err_nodes += 1
total = ok_nodes + err_nodes
print("", flush = True)
if ok_nodes:
_public_reachable = True
print(
f"{ok_c} Reachability check: {url}/ is reachable from the "
f"public internet ({ok_nodes}/{total} probe nodes connected).{reset}",
flush = True,
)
elif err_nodes:
_public_reachable = False
print(
f"{err_c} Reachability check: {url}/ is NOT reachable from "
f"the public internet ({err_nodes}/{total} probe nodes failed).{reset}",
flush = True,
)
print(
f"{dim} Usually a cloud firewall (AWS security group, "
f"GCP firewall / Azure NSG rule) or home router isn't "
f"allowing inbound TCP {port}.{reset}",
flush = True,
)
print(
f"{dim} No firewall change needed -- SSH local-forward "
f"from your own computer:{reset}",
flush = True,
)
print(
f"{dim} ssh -L {port}:localhost:{port} <user>@{display_host}{reset}",
flush = True,
)
print(
f"{dim} then open http://localhost:{port}/ in your browser.{reset}",
flush = True,
)
# Only offer the local URL if loopback answers.
local_url = _working_local_url(port)
if local_url:
print(
f"{local_url_c} You can access Unsloth Studio locally "
f"in the meantime: {local_url}{reset}",
flush = True,
)
else:
print(
f"{warn_c} Reachability check: probe nodes did not respond "
f"in time -- could not verify {url}/.{reset}",
flush = True,
)
except urllib.error.URLError:
# Outbound HTTPS blocked; skip.
pass
except Exception:
pass
def _display_host_for_bind(host: str) -> str:
return _resolve_external_ip() if host in ("0.0.0.0", "::") else host
def _loopback_bind_host_for(host: str) -> str:
return "::1" if host == "::" else "127.0.0.1"
def _url_host(host: str) -> str:
return (
f"[{host}]" if ":" in host and not (host.startswith("[") and host.endswith("]")) else host
)
def _tool_policy_notice(host: str, secure: bool, enable_tools: "Optional[bool]") -> str:
"""One-line tool-policy summary for the plain-server startup banner, so a
network-reachable launch is never silent about code execution."""
if enable_tools is False:
return "Server-side tools are DISABLED (--disable-tools)."
state = (
"ENABLED (--enable-tools)"
if enable_tools
else "ENABLED by default (per-request setting honored)"
)
if secure:
return (
f"Server-side tools are {state}, reachable via the authenticated "
"Cloudflare HTTPS tunnel. Anyone with the API key can run code on "
"this machine. Do not share the API key. Pass --disable-tools to turn off."
)
from utils.host_policy import is_external_host
if host in ("0.0.0.0", "::") or is_external_host(host):
return (
f"Server-side tools are {state} and this port is network-reachable. "
"Anyone who can reach it with the API key can run code on this "
"machine. Do not share the API key. Pass --disable-tools to turn off."
)
return f"Server-side tools are {state} for loopback. Pass --disable-tools to turn off."
def _emit_tool_policy_notice(host: str, secure: bool, enable_tools: "Optional[bool]") -> None:
print(_tool_policy_notice(host, secure, enable_tools), flush = True)
def _emit_secure_startup_output(port: int, enable_tools: "Optional[bool]" = None) -> None:
"""Secure-mode banner: only the Cloudflare link (loopback has no public raw URL)."""
print("")
print("🦥 Unsloth Studio is running (secure)")
print("" * 52)
_print_cloudflare_line(secure = True)
print(f" On this machine only: http://127.0.0.1:{port}/")
print("" * 52)
_emit_tool_policy_notice("127.0.0.1", True, enable_tools)
print_studio_stop_hint()
def _emit_startup_output(
host: str,
port: int,
display_host: str,
secure: bool = False,
enable_tools: "Optional[bool]" = None,
) -> None:
"""Print the access banner, post-startup warnings, the tool-policy notice,
then a single stop hint. Extracted from ``_run`` so the wiring is testable."""
if secure:
_emit_secure_startup_output(port, enable_tools)
return
wildcard_bind = host in ("0.0.0.0", "::")
localhost_mismatch_url = _localhost_ipv6_mismatch_url(host, port)
print_studio_access_banner(
port = port,
bind_host = host,
display_host = display_host,
include_stop_hint = False,
)
if localhost_mismatch_url:
_print_localhost_ipv6_mismatch_warning(localhost_mismatch_url, port)
elif wildcard_bind:
_verify_global_reachability(display_host, port)
_print_cloudflare_line(loopback_host = _loopback_bind_host_for(host))
_emit_tool_policy_notice(host, False, enable_tools)
print_studio_stop_hint()
def _print_cloudflare_line(secure: bool = False, loopback_host: str = "127.0.0.1") -> None:
"""Print Cloudflare tunnel state for startup banners."""
from startup_banner import stdout_supports_color
accent = "\033[38;5;150;1m"
warn = "\033[38;5;215;1m"
reset = "\033[0m"
color = stdout_supports_color()
def _emit(text: str, style: str = "") -> None:
print(f"{style}{text}{reset}" if (color and style) else text)
if _cloudflare_url:
if _public_reachable is False:
_emit(f" Use the secure link access via Cloudflare instead: {_cloudflare_url}", accent)
else:
_emit(f" Secure link access via Cloudflare: {_cloudflare_url}", accent)
if not secure:
if _public_reachable is True:
_emit(
" Cloudflare tunnel: ON. This Cloudflare URL is PUBLIC, and the "
"raw port is also publicly reachable. --no-cloudflare disables "
f"only the Cloudflare URL; bind {loopback_host} or close firewall "
"access to keep Unsloth private.",
warn,
)
else:
_emit(
" Cloudflare tunnel: ON. This is a PUBLIC internet URL: anyone "
"who has it can reach this Unsloth. Relaunch with --no-cloudflare "
f"to disable the Cloudflare URL; bind {loopback_host} or close "
"firewall access to keep Unsloth private.",
warn,
)
return
if _cloudflare_requested:
if _public_reachable is True:
_emit(
" Cloudflare tunnel: requested but failed to start. The raw port is "
"still reachable from the public internet (see the reachability check "
"above): anyone who can reach it can access this Unsloth.",
warn,
)
elif _public_reachable is False:
_emit(
" Cloudflare tunnel: requested but failed to start. Unsloth is reachable "
"on your local network only (no public link).",
warn,
)
else:
_emit(
" Cloudflare tunnel: requested but failed to start. There is no "
"Cloudflare public link. Raw port reachability was not verified; "
f"bind {loopback_host} or close firewall access to keep Unsloth private.",
warn,
)
elif _cloudflare_flag:
if _public_reachable is True:
_emit(
" Cloudflare tunnel: OFF for this mode. The raw port is still "
"reachable from the public internet (see the reachability check above): "
"anyone who can reach it can access this Unsloth.",
warn,
)
elif _public_reachable is False:
_emit(
" Cloudflare tunnel: OFF for this mode. Unsloth is reachable on your "
"local network only (no public link)."
)
else:
_emit(
" Cloudflare tunnel: OFF for this mode. There is no Cloudflare public "
"link. Raw port reachability was not verified; "
f"bind {loopback_host} or close firewall access to keep Unsloth private.",
warn,
)
elif _cloudflare_flag is False or _cloudflare_flag is None:
# None = off by default (no flag); False = explicit --no-cloudflare.
_reason = "default" if _cloudflare_flag is None else "--no-cloudflare"
if _public_reachable is True:
_emit(
f" Cloudflare tunnel: OFF ({_reason}). The raw port is still "
"reachable from the public internet (see the reachability check above): "
"pass --cloudflare to also expose a public Cloudflare HTTPS link, or "
f"bind {loopback_host} to keep Unsloth private.",
warn,
)
elif _public_reachable is False:
_emit(
f" Cloudflare tunnel: OFF ({_reason}). Unsloth is reachable on your "
"local network only. Pass --cloudflare to expose a public "
"Cloudflare HTTPS link."
)
else:
_emit(
f" Cloudflare tunnel: OFF ({_reason}). There is no Cloudflare "
"public link. Raw port reachability was not verified; pass --cloudflare "
"to expose a public Cloudflare HTTPS link, or "
f"bind {loopback_host} or close firewall access to keep Unsloth private.",
warn,
)
def _get_pid_on_port(port: int) -> "tuple[int, str] | None":
"""Return (pid, process_name) listening on *port*, or None.
Uses psutil when available, else None so callers can still report the conflict
without process details.
"""
try:
import psutil
except ImportError:
return None
try:
for conn in psutil.net_connections(kind = "tcp"):
if conn.status == "LISTEN" and conn.laddr.port == port:
if conn.pid is None:
return None
try:
proc = psutil.Process(conn.pid)
return (conn.pid, proc.name())
except (psutil.NoSuchProcess, psutil.AccessDenied):
return (conn.pid, "<unknown>")
except (psutil.AccessDenied, OSError) as e:
# net_connections() needs elevated privileges on some platforms.
logger.debug("Failed to scan network connections for port %s: %s", port, e)
return None
def _is_port_free(host: str, port: int) -> bool:
"""Check if a port is available for binding.
For a ``0.0.0.0`` wildcard host, also check whether anything is listening on
``127.0.0.1`` (and ``::1`` when IPv6 exists): an SSH tunnel may hold loopback
while the wildcard bind succeeds, making Unsloth unreachable via ``localhost``.
"""
import socket
# 1. Can we bind to the requested address? getaddrinfo resolves both
# IPv4 and IPv6 to the right address family.
try:
addr_info = socket.getaddrinfo(host, port, socket.AF_UNSPEC, socket.SOCK_STREAM)
family, socktype, proto, _, sockaddr = addr_info[0]
with socket.socket(family, socktype, proto) as s:
s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
s.bind(sockaddr)
except OSError:
return False
# 2. On a wildcard bind, verify localhost isn't already claimed by another
# process (e.g. an SSH -L tunnel); a successful connect means it is.
if host in ("0.0.0.0", "::"):
for loopback, family in [
("127.0.0.1", socket.AF_INET),
("::1", socket.AF_INET6),
]:
try:
with socket.socket(family, socket.SOCK_STREAM) as s:
s.settimeout(1)
if s.connect_ex((loopback, port)) == 0:
# Port is taken on loopback.
return False
except OSError:
# IPv6 disabled or other OS-level restriction -- skip.
continue
return True
def _find_free_port(
host: str,
start: int,
max_attempts: int = 20,
) -> int:
"""Find a free port from `start`, trying up to max_attempts ports."""
for offset in range(max_attempts):
candidate = start + offset
if _is_port_free(host, candidate):
return candidate
raise RuntimeError(f"Could not find a free port in range {start}-{start + max_attempts - 1}")
from utils.paths.storage_roots import studio_root as _studio_root
_PID_FILE = _studio_root() / "studio.pid"
# Direct backend launches bypass the CLI's env re-export; do it here for
# real custom roots so unsloth-zoo's import-time LLAMA_CPP_DEFAULT_DIR
# picks up the custom build. Skip legacy-default to avoid flipping
# default-mode installs into env-override.
try:
_LEGACY_STUDIO_ROOT = (Path.home() / ".unsloth" / "studio").resolve()
except (OSError, ValueError):
_LEGACY_STUDIO_ROOT = Path.home() / ".unsloth" / "studio"
try:
_STUDIO_ROOT_RESOLVED = _studio_root().resolve()
except (OSError, ValueError):
_STUDIO_ROOT_RESOLVED = _studio_root()
if _STUDIO_ROOT_RESOLVED != _LEGACY_STUDIO_ROOT:
if not os.environ.get("UNSLOTH_STUDIO_HOME"):
os.environ["UNSLOTH_STUDIO_HOME"] = str(_STUDIO_ROOT_RESOLVED)
if not os.environ.get("UNSLOTH_LLAMA_CPP_PATH"):
os.environ["UNSLOTH_LLAMA_CPP_PATH"] = str(_STUDIO_ROOT_RESOLVED / "llama.cpp")
# The studio bundles unsloth_zoo; declare unsloth present (as `import unsloth`
# does) so its lazy submodule imports (export, hardware, mlx) and the
# DiffusionGemma runner never trip the install guard on a clean install.
os.environ.setdefault("UNSLOTH_IS_PRESENT", "1")
def _write_pid_file():
"""Write the current process PID to the studio PID file."""
try:
_PID_FILE.parent.mkdir(parents = True, exist_ok = True)
_PID_FILE.write_text(str(os.getpid()), encoding = "utf-8")
except OSError:
pass
def _remove_pid_file():
"""Remove the PID file if it belongs to this process."""
try:
if _PID_FILE.is_file():
stored = _PID_FILE.read_text(encoding = "utf-8").strip()
if stored == str(os.getpid()):
_PID_FILE.unlink(missing_ok = True)
except (OSError, UnicodeDecodeError):
pass
def _graceful_shutdown(server = None):
"""Shut down all subprocess backends and the uvicorn server.
Called from signal handlers to clean up children before exit. Critical on
Windows where atexit handlers are unreliable after Ctrl+C.
"""
_remove_pid_file()
logger.info("Graceful shutdown initiated -- cleaning up subprocesses...")
# 1. Shut down uvicorn (releases the listening socket).
if server is not None:
server.should_exit = True
# 2. Clean up inference subprocess (if instantiated).
try:
from core.inference.orchestrator import _inference_backend
if _inference_backend is not None:
_inference_backend._shutdown_subprocess(timeout = 5.0)
except Exception as e:
logger.warning("Error shutting down inference subprocess: %s", e)
# 3. Clean up export subprocess (if instantiated).
try:
from core.export.orchestrator import _export_backend
if _export_backend is not None:
_export_backend._shutdown_subprocess(timeout = 5.0)
except Exception as e:
logger.warning("Error shutting down export subprocess: %s", e)
# 4. Clean up training subprocess (if active).
try:
from core.training.training import _training_backend
if _training_backend is not None:
_training_backend.force_terminate()
except Exception as e:
logger.warning("Error shutting down training subprocess: %s", e)
# 5. Kill llama-server subprocess (if loaded).
try:
from routes.inference import _llama_cpp_backend
if _llama_cpp_backend is not None:
_llama_cpp_backend._kill_process()
except Exception as e:
logger.warning("Error shutting down llama-server: %s", e)
# 6. Stop the Cloudflare tunnel (if started).
try:
from cloudflare_tunnel import stop_studio_tunnel
stop_studio_tunnel()
except Exception as e:
logger.warning("Error stopping Cloudflare tunnel: %s", e)
# 7. Backstop sweep for any adopted child the steps above missed.
try:
from utils.process_lifetime import terminate_all
terminate_all()
except Exception as e:
logger.warning("Error in process-lifetime sweep: %s", e)
logger.info("All subprocesses cleaned up")
# Bound the join so a stuck uvicorn shutdown cannot hang the terminal.
_SERVER_SHUTDOWN_JOIN_TIMEOUT = 5.0
def _flush_standard_streams() -> None:
for stream in (sys.stdout, sys.stderr):
try:
stream.flush()
except Exception:
pass
def _wait_for_server_shutdown(timeout: Optional[float] = _SERVER_SHUTDOWN_JOIN_TIMEOUT) -> None:
"""Join the uvicorn thread so the prompt returns only after its shutdown logs
flush. Skip the self-join when called from the server thread."""
import threading
thread = _server_thread
if thread is None or thread is threading.current_thread():
_flush_standard_streams()
return
thread.join(timeout = timeout)
if thread.is_alive():
logger.warning("Timed out waiting for uvicorn server thread to stop")
_flush_standard_streams()
# The uvicorn server instance -- set by run_server(), used by callers
# that tell the server to exit (e.g. signal handlers).
_server = None
_server_thread = None
# Shutdown event -- wakes the main loop on signal.
_shutdown_event = None
# trycloudflare.com URL for wildcard binds (set by run_server, read by the banner);
# None when there is no tunnel (loopback, disabled, or a silently-ignored failure).
_cloudflare_url = None
# Public reachability from the last _verify_global_reachability run, read by the
# Cloudflare banner line. True when the public ip:port probe confirmed reachable,
# False when it confirmed NOT reachable, None when the probe did not run or could
# not decide (timeout, blocked, private address).
_public_reachable = None
_cloudflare_requested = False
# Opt-in tri-state (mirrors the CLI): None = off by default, True = on,
# False = explicit --no-cloudflare. run_server overwrites it before the banner.
_cloudflare_flag = None
_DEFAULT_FRONTEND_PATH = Path(__file__).resolve().parent.parent / "frontend" / "dist"
def _iter_frontend_fallback_candidates() -> "list[Path]":
"""Yield `studio/frontend/dist` paths to try when the default is missing.
Covers PATH-shadowed binaries whose __file__ resolves into a site-packages
tree with no vite build (e.g. plain `pip install unsloth`).
"""
import ast
import re
out: list[Path] = []
home_str = (
os.environ.get("UNSLOTH_STUDIO_HOME")
or os.environ.get("STUDIO_HOME")
or str(Path.home() / ".unsloth" / "studio")
)
venv_dir = Path(home_str).expanduser() / "unsloth_studio"
# Installer venv site-packages.
for pattern in (
"lib/python*/site-packages/studio/frontend/dist",
"Lib/site-packages/studio/frontend/dist",
):
out.extend(venv_dir.glob(pattern))
# Editable source roots referenced from the installer venv.
for sp_pattern in ("lib/python*/site-packages", "Lib/site-packages"):
for sp in venv_dir.glob(sp_pattern):
for finder in sp.glob("__editable___*_finder.py"):
try:
src = finder.read_text(encoding = "utf-8")
except (OSError, UnicodeDecodeError):
continue
# Tolerate single/multi-line dict literals; [^}]* rejects nested
# dicts, which the setuptools editable template never emits.
m = re.search(r"^MAPPING\s*(?::[^=]*)?=\s*(\{[^}]*\})", src, re.M | re.S)
if not m:
continue
try:
mapping = ast.literal_eval(m.group(1))
except (SyntaxError, ValueError):
continue
# literal_eval can return a set/list/None if `{...}` isn't a dict.
if not isinstance(mapping, dict):
continue
studio_pkg = mapping.get("studio")
if studio_pkg:
out.append(Path(studio_pkg) / "frontend" / "dist")
return out
def _resolve_frontend_path(frontend_path: Path) -> tuple[Optional[Path], list[Path]]:
"""Pick a frontend dir that contains `index.html`.
Returns (chosen, attempted). `chosen` is None if nothing servable was found;
`attempted` is the ordered list for diagnostics.
"""
attempted: list[Path] = []
seen: set[Path] = set()
def _try(p: Path) -> bool:
try:
key = p.resolve()
except OSError:
key = p
if key in seen:
return False
seen.add(key)
attempted.append(p)
return (p / "index.html").is_file()
if _try(Path(frontend_path)):
return attempted[-1], attempted
for alt in _iter_frontend_fallback_candidates():
if _try(alt):
return attempted[-1], attempted
return None, attempted
class _TeeStream:
"""Mirror writes to the original stream and a session log file.
Console behavior is unchanged (writes/returns delegate to the original
stream; Tauri's structured-stdout protocol and isatty probes see exactly
what they saw before). The file copy is best-effort: a full disk or a
closed handle must never break the console."""
def __init__(self, stream, log_fh):
self._stream = stream
self._log_fh = log_fh
def write(self, data):
try:
self._log_fh.write(data)
except Exception:
pass
return self._stream.write(data)
def flush(self):
try:
self._log_fh.flush()
except Exception:
pass
try:
self._stream.flush()
except Exception:
pass
def close(self):
# We do NOT own the console stream (it is the terminal / Jupyter kernel
# stream we wrapped), so closing the tee must never take the server down.
# Flush the log copy, then forward close() to the wrapped stream
# best-effort: on Colab that stream is an ipykernel OutStream whose
# close() can raise (see _harden_console_close / ipython/ipykernel#867).
try:
self._log_fh.flush()
except Exception:
pass
try:
self._stream.close()
except Exception:
pass
def __getattr__(self, name):
return getattr(self._stream, name)
_WATCH_FD_THREAD_ATTR = "watch_fd_thread"
def _is_missing_watch_fd_thread(exc):
"""True only for ipython/ipykernel#867's missing-``watch_fd_thread`` error.
``AttributeError.name`` exists from Python 3.10; the message carries the
attribute name on every version (possibly with a "Did you mean" tail), so
check both and let every other AttributeError through.
"""
if getattr(exc, "name", None) == _WATCH_FD_THREAD_ATTR:
return True
return _WATCH_FD_THREAD_ATTR in str(exc)
def _harden_console_close(stream):
"""Stop a displaced console stream's close() from aborting Studio startup.
``_setup_server_disk_logging`` replaces ``sys.stdout``/``sys.stderr`` with a
tee. That changes the object identity of the console stream, so a third-party
logging handler that captured the ORIGINAL stream (notably Colab's ``absl``
logging handler, whose ``close()`` skips ``sys.stdout``/``sys.stderr`` but not
a stream that is no longer either) treats it as an ordinary stream and calls
``close()`` on it during logging teardown -- ``uvicorn.Config()`` ->
``logging.config.dictConfig()`` -> ``logging.shutdown()``.
A Jupyter/Colab ``ipykernel`` ``OutStream`` created with ``watchfd=False``
(the Colab default, and every in-process kernel) never gains a
``watch_fd_thread``, yet the ``OutStream.close()`` shipped in the affected
ipykernel versions joins that thread unconditionally and raises
``AttributeError: 'OutStream' object has no attribute 'watch_fd_thread'``
(ipython/ipykernel#867). That AttributeError propagates out of
``uvicorn.Config(...)`` and aborts startup ("Unsloth Studio failed to start").
Wrap the stream's ``close()`` in a transparent pass-through that swallows
ONLY that specific teardown AttributeError. A healthy close() (a real console
stream, or an OutStream with fd-watching on) runs to completion exactly as
before and any other error still propagates, so nothing changes off Colab. A
stream whose ``close`` cannot be reassigned keeps its original close().
"""
try:
_orig_close = stream.close
except Exception:
return
def _safe_close(*args, **kwargs):
try:
return _orig_close(*args, **kwargs)
except AttributeError as exc:
if not _is_missing_watch_fd_thread(exc):
# A real teardown failure; never hide it.
raise
# ipython/ipykernel#867: watchfd=False OutStream.close() joins a
# thread that was never created. Nothing to clean up; keep going.
return None
try:
stream.close = _safe_close
except (AttributeError, TypeError):
# A stream that forbids setting instance attributes; leave it as-is.
pass
def _setup_server_disk_logging():
"""Tee stdout/stderr to ~/.unsloth/studio/logs/server/ and aim
faulthandler at the same file so hard crashes (access violations /
SIGSEGV in the GPU runtime) leave a stack trace on disk.
Also exports PYTHONFAULTHANDLER=1 so child Python processes (training
workers) dump native-crash stacks to their captured stderr. Keeps the
newest 20 session logs. Opt out with UNSLOTH_STUDIO_NO_FILE_LOG=1.
Returns the log path, or None when disabled/unavailable.
"""
if os.environ.get("UNSLOTH_STUDIO_NO_FILE_LOG") == "1":
return None
try:
from utils.paths import studio_root
log_dir = Path(studio_root()) / "logs" / "server"
except Exception:
home = (
os.environ.get("UNSLOTH_STUDIO_HOME")
or os.environ.get("STUDIO_HOME")
or os.path.join(os.path.expanduser("~"), ".unsloth", "studio")
)
log_dir = Path(home) / "logs" / "server"
try:
log_dir.mkdir(parents = True, exist_ok = True)
stamp = time.strftime("%Y%m%d-%H%M%S")
log_path = log_dir / f"server-{stamp}-pid{os.getpid()}.log"
# Line-buffered so the tail survives a hard kill; errors="replace"
# so a console encoding quirk can never take the server down.
log_fh = open(log_path, "w", encoding = "utf-8", errors = "replace", buffering = 1)
except Exception:
return None
import faulthandler
try:
faulthandler.enable(file = log_fh, all_threads = True)
except Exception:
pass
# Children (training workers) inherit: their native-crash stacks land on
# the stderr the server already captures.
os.environ.setdefault("PYTHONFAULTHANDLER", "1")
# Replacing the console streams orphans them from third-party "is this the
# live console?" checks, so guard their close() first (ipython/ipykernel#867).
_harden_console_close(sys.stdout)
_harden_console_close(sys.stderr)
sys.stdout = _TeeStream(sys.stdout, log_fh)
sys.stderr = _TeeStream(sys.stderr, log_fh)
# Best-effort retention: keep the newest 20 session logs.
try:
logs = sorted(log_dir.glob("server-*.log"), key = lambda p: p.stat().st_mtime)
for old in logs[:-20]:
old.unlink(missing_ok = True)
except Exception:
pass
return log_path
def _cloudflare_tunnel_should_start(
*, cloudflare: bool, host: str, secure: bool, api_only: bool, is_colab: bool
) -> bool:
"""Whether to start the Cloudflare tunnel. --secure exposes only the tunnel
(loopback bind), so it tunnels even api-only (headless secure API serving);
otherwise tunnel wildcard binds, never api-only (Tauri) or Colab."""
if is_colab or not cloudflare:
return False
if secure:
return True
return host in ("0.0.0.0", "::") and not api_only
def _stream_isatty(stream) -> bool:
"""isatty() that treats broken streams as non-interactive.
isatty() can raise under service wrappers (closed stdin -> ValueError;
sys.stdin None in Windows GUI -> AttributeError); such a stream can't host a
prompt, which is a fallback, not an error.
"""
try:
return stream.isatty()
except (AttributeError, ValueError):
return False
def _terminal_password_gate(
*,
tunnel_will_start: bool,
host: str,
secure: bool,
api_only: bool,
frontend_served: bool,
is_colab: bool = False,
) -> Tuple[bool, bool]:
"""Force a terminal password change before the public tunnel goes up.
When the tunnel is about to publish Unsloth and the seeded admin password was
never changed, ask for a new one (masked, confirmed) before any public URL
exists. The CLI normally does this before re-exec'ing the backend; this is
the backstop for direct `python run.py` launches and older-CLI installs.
Must run BEFORE the uvicorn socket binds: on a wildcard bind the served HTML
injects the bootstrap credential, so a pre-gate listener would hand the
default password to anyone reaching the raw port while the operator types.
Returns (proceed, drop_bootstrap_injection):
proceed False -> abort the launch (interactive refusal, or a headless
public launch nothing would protect); fail closed.
drop_bootstrap_injection True -> caller must null
app.state.bootstrap_password: the password just changed (stale), or a
public URL is about to serve the default credential and must not leak it.
Without a usable terminal the prompt is skipped: proceed if the bootstrap
deadline (armed later) will protect the launch; if even that is disabled
(api-only, timeout 0) nothing protects it, so refuse. NOT wrapped in a broad
try/except: an auth storage failure must abort rather than expose the default.
"""
if not tunnel_will_start:
return True, False
from auth import hashing as _auth_hashing
from auth import storage as _auth_storage
from auth.bootstrap_timeout import (
bootstrap_timeout_seconds,
should_arm_bootstrap_timeout,
)
from auth.terminal_prompt import (
prompt_for_password_change,
should_prompt_password_change,
)
_admin = _auth_storage.DEFAULT_ADMIN_USERNAME
# Gate can run before lifespan: seed the admin row here (idempotent).
_auth_storage.ensure_default_admin()
requires_change = _auth_storage.requires_password_change(_admin)
if not requires_change:
return True, False
if not should_prompt_password_change(
tunnel_will_start = tunnel_will_start,
requires_change = requires_change,
stdin_isatty = _stream_isatty(sys.stdin),
stderr_isatty = _stream_isatty(sys.stderr),
):
# No terminal: only proceed if the bootstrap deadline will arm; api-only
# and TIMEOUT=0 never arm it, leaving the default credential public.
deadline_arms = should_arm_bootstrap_timeout(
host = host,
secure = secure,
api_only = api_only,
frontend_served = frontend_served,
is_colab = is_colab,
requires_change = True,
timeout_seconds = bootstrap_timeout_seconds(),
)
if not deadline_arms:
print(
"Refusing to publish Unsloth on a public Cloudflare URL: the "
"default admin password was never changed, no terminal is "
"attached to change it here, and the bootstrap shutdown "
"deadline does not apply to this launch (api-only, or "
"UNSLOTH_STUDIO_BOOTSTRAP_TIMEOUT=0). Change the password "
"first (run `unsloth studio` locally and log in, or re-run "
"with a terminal attached), then retry.",
file = sys.stderr,
flush = True,
)
return False, False
# The public page won't auto-fill the bootstrap credential (suppressed
# below) and the seeded file may already be gone, so point recovery at a
# terminal-attached run / reset-password instead of reading it from disk.
print(
" WARNING: the default admin password is still active while "
"Unsloth is about to be published on a public Cloudflare URL, and "
"no terminal is attached to change it here. The public page will "
"NOT auto-fill the bootstrap credential. Set a new password by "
"running `unsloth studio` locally with a terminal attached, or "
"`unsloth studio reset-password`. Unsloth shuts down after the "
"bootstrap deadline (UNSLOTH_STUDIO_BOOTSTRAP_TIMEOUT, default 1h) "
"unless the password is changed.",
file = sys.stderr,
flush = True,
)
# Never serve the default credential in HTML over a public URL.
return True, True
def _is_current_password(candidate: str) -> bool:
record = _auth_storage.get_user_and_secret(_admin)
if record is None:
return False
salt, pwd_hash, _jwt_secret, _must_change = record
return _auth_hashing.verify_password(candidate, salt, pwd_hash)
def _apply_change(new_password: str) -> None:
# Same effects as routes/auth.py change_password: rehash, rotate the JWT
# secret, revoke refresh tokens in the SAME transaction.
_auth_storage.update_password(_admin, new_password, revoke_refresh_tokens = True)
changed = prompt_for_password_change(
min_length = _auth_storage.MIN_PASSWORD_LENGTH,
is_current_password = _is_current_password,
apply_change = _apply_change,
out = sys.stderr,
)
return (True, True) if changed else (False, False)
def _apply_supplied_password(password_value: "Optional[str]") -> None:
"""Non-interactively set the INITIAL admin password before the socket binds,
for a direct ``python run.py`` launch (the CLI does this in its own parent).
Value comes from --password / UNSLOTH_STUDIO_PASSWORD / stdin.
Only ever sets the FIRST password: an already-set one is a hard error, an
invalid value fails closed. NOT wrapped in a broad try/except: an auth
storage failure must abort rather than expose the default credential.
"""
from auth import hashing as _auth_hashing
from auth import storage as _auth_storage
from auth.terminal_prompt import SUPPLIED_PASSWORD_ENV, resolve_supplied_password
supplied = resolve_supplied_password(password_value)
# Strip the env var once read so child subprocesses (cloudflared, llama-server,
# code-exec tools) can't inherit the plaintext via /proc/PID/environ. Mirrors
# the CLI. Unconditional: strips a leftover value even when a literal --password won.
os.environ.pop(SUPPLIED_PASSWORD_ENV, None)
if not supplied:
return
_admin = _auth_storage.DEFAULT_ADMIN_USERNAME
_auth_storage.ensure_default_admin()
if not _auth_storage.requires_password_change(_admin):
print(
"Error: an Unsloth admin password is already set; --password only sets "
"the initial password. Run `unsloth studio reset-password` first.",
file = sys.stderr,
flush = True,
)
sys.exit(1)
def _is_current_password(candidate: str) -> bool:
record = _auth_storage.get_user_and_secret(_admin)
if record is None:
return False
salt, pwd_hash, _jwt_secret, _must_change = record
return _auth_hashing.verify_password(candidate, salt, pwd_hash)
if len(supplied) < _auth_storage.MIN_PASSWORD_LENGTH:
print(
f"Error: password must be at least {_auth_storage.MIN_PASSWORD_LENGTH} "
"characters; not starting.",
file = sys.stderr,
flush = True,
)
sys.exit(1)
if any(ch.isspace() for ch in supplied):
print(
"Error: password cannot contain spaces; not starting.",
file = sys.stderr,
flush = True,
)
sys.exit(1)
if _is_current_password(supplied):
print(
"Error: the new password must differ from the current bootstrap "
"password; not starting.",
file = sys.stderr,
flush = True,
)
sys.exit(1)
_auth_storage.update_password(_admin, supplied, revoke_refresh_tokens = True)
print(f"Password updated for '{_admin}'.", file = sys.stderr, flush = True)
def _apply_cli_tool_policy(enable_tools: "Optional[bool]") -> None:
"""Honor an explicit --enable-tools/--disable-tools; None leaves the policy
unset (tools default on, per-request enable_tools honored). Host is never
inspected here."""
if enable_tools is None:
return
from state.tool_policy import set_tool_policy
set_tool_policy(enable_tools)
# Mirror unsloth_cli/commands/studio.py's _PARALLEL_*: the admission queue caps concurrent
# chats at the slot count, so a direct launch matches the CLI (VRAM fit may still cut it
# back). Defined above run_server() so embedders that omit it do not serialise every chat.
_PARALLEL_MIN = 1
_PARALLEL_MAX = 64
_PARALLEL_DEFAULT_PLAIN = 4
def run_server(
host: str = "127.0.0.1",
port: int = 8888,
frontend_path: Path = _DEFAULT_FRONTEND_PATH,
silent: bool = False,
api_only: bool = False,
llama_parallel_slots: int = _PARALLEL_DEFAULT_PLAIN,
cloudflare: "Optional[bool]" = None,
secure: bool = False,
enable_tools: "Optional[bool]" = None,
password: "Optional[str]" = None,
emit_tauri_port: bool = True,
):
"""
Start the FastAPI server.
Args:
host: Host to bind to
port: Port to bind to (auto-increments if in use)
frontend_path: Path to frontend build directory (optional)
silent: Suppress startup messages
api_only: API server only, no frontend (for Tauri desktop app)
llama_parallel_slots: parallel slots for llama-server (default
_PARALLEL_DEFAULT_PLAIN, matching the CLI entry points)
cloudflare: opt in to the public Cloudflare HTTPS tunnel for a wildcard
bind. Tri-state: None (unset) and False both mean off; True enables it.
--secure implies it (True) and rejects an explicit False.
enable_tools: explicit --enable-tools/--disable-tools policy; None leaves
the default (tools on, per-request enable_tools honored)
emit_tauri_port: print the machine-readable TAURI_PORT line the desktop
app parses from stdout; the headless `run --api-only` path turns it
off so it does not pollute the documented URL/API-key banner
Note:
Signal handlers are NOT registered here so embedders (e.g. Colab) keep
their own interrupt semantics; standalone callers register them after.
"""
global _server, _server_thread, _shutdown_event
boot_started = time.perf_counter()
logger.info("run_server startup begin api_only=%s host=%s port=%s", api_only, host, port)
# Reap every child if the parent dies abnormally (terminal close, Task
# Manager kill, SIGKILL); must run before any child can spawn.
from utils.process_lifetime import initialize_parent_lifetime
initialize_parent_lifetime()
# --secure exposes ONLY the Cloudflare link: reject --secure --no-cloudflare,
# then force a loopback bind so the raw port is never public (even -H 0.0.0.0).
# Otherwise keep the tri-state so the banner distinguishes "off by default"
# from an explicit --no-cloudflare.
if secure:
if cloudflare is False:
raise SystemExit(
"--secure requires the Cloudflare tunnel; do not combine it with --no-cloudflare."
)
cloudflare = True
host = "127.0.0.1"
# `unsloth studio run` installs its own resolved policy and passes None here.
_apply_cli_tool_policy(enable_tools)
# Windows cp1252 can't encode emoji; reconfigure stdout to UTF-8.
if sys.platform == "win32" and hasattr(sys.stdout, "reconfigure"):
try:
sys.stdout.reconfigure(encoding = "utf-8", errors = "replace")
except Exception:
pass
# Persist a session log + native-crash stacks BEFORE importing main, so
# even import-time failures leave evidence on disk. Field report: Unsloth
# "terminates without a warning" -- a native crash in the GPU runtime
# kills the process with no Python traceback, and a desktop-shortcut
# console closes before anything can be read. Console-only logging made
# that undiagnosable.
_session_log = _setup_server_disk_logging()
if _session_log is not None and not silent:
print(f"Session log: {_session_log}")
# Set env vars BEFORE importing main so CORS middleware picks them up.
# secure api-only is a remote server behind Cloudflare, so it keeps the
# any-origin CORS profile; plain api-only stays locked to the Tauri app.
if api_only:
os.environ["UNSLOTH_API_ONLY"] = "1"
if secure:
os.environ["UNSLOTH_SECURE"] = "1"
import asyncio
# nest_asyncio is for Colab/IPython, where the main thread already runs a loop
# the blocking waits below would collide with. Apply it only with a loop running
# (a plain CLI start has nothing to nest) and only on Python <= 3.13: on 3.14+
# its global Task patch leaves asyncio.current_task() None (tracking moved into
# C), which also breaks the background uvicorn loop and 500s every request. It
# is archived upstream, so no 3.14 fix is coming; skip it there.
if sys.version_info < (3, 14):
try:
asyncio.get_running_loop()
except RuntimeError:
pass
else:
import nest_asyncio
nest_asyncio.apply()
from threading import Thread, Event
import uvicorn
# `from main import app` below loads torch/unsloth/transformers (~2 min cold,
# silent), so print a flushed heads-up (piped stdout is block-buffered).
if not silent:
print(
"Loading Unsloth Studio, please wait... (this can take a few minutes)",
flush = True,
)
print(" - loading PyTorch, Unsloth and Transformers...", flush = True)
import_started = time.perf_counter()
from main import app, setup_frontend, _IS_COLAB
logger.info(
"Imported FastAPI app in %.1fms",
(time.perf_counter() - import_started) * 1000,
)
if not silent:
print(" - Starting server...", flush = True)
from utils.paths import ensure_studio_directories
# Allow local stdio MCP servers on a loopback bind (the user's own machine),
# but never on Colab, which is a hosted VM reachable through its proxy. The
# gate reads the env var at request time, so this need not precede the import.
from utils.host_policy import apply_stdio_mcp_loopback_default
apply_stdio_mcp_loopback_default(host, is_colab = _IS_COLAB)
# Create all standard directories on startup.
ensure_studio_directories()
logger.info(
"Ensured Unsloth directories in %.1fms",
(time.perf_counter() - boot_started) * 1000,
)
# Auto-find a free port if the requested one is in use.
if not _is_port_free(host, port):
original_port = port
blocker = _get_pid_on_port(port)
port = _find_free_port(host, port + 1)
if not silent:
print("")
print("=" * 50)
if blocker:
pid, name = blocker
print(f"Port {original_port} is already in use by {name} (PID {pid}).")
else:
print(f"Port {original_port} is already in use.")
print(f"Unsloth Studio will use port {port} instead.")
print(f"Open http://localhost:{port} in your browser.")
print("=" * 50)
print("")
# Setup frontend (skip in api-only). Falls back through alternate locations if
# the default lacks a built dist; errors loudly rather than 404 on `/`.
if frontend_path and not api_only:
chosen, attempted = _resolve_frontend_path(Path(frontend_path))
if chosen is not None and setup_frontend(app, chosen):
if not silent:
# Resolve so logs show an absolute path for support.
try:
display = chosen.resolve()
except OSError:
display = chosen
print(f"[OK] Frontend loaded from {display}")
else:
home_str = (
os.environ.get("UNSLOTH_STUDIO_HOME")
or os.environ.get("STUDIO_HOME")
or str(Path.home() / ".unsloth" / "studio")
)
# Windows shim: $STUDIO_HOME/bin/unsloth.exe; Linux/macOS venv binary:
# $STUDIO_HOME/unsloth_studio/bin/unsloth.
home = Path(home_str).expanduser()
if sys.platform == "win32":
installer_bin = home / "bin" / "unsloth.exe"
else:
installer_bin = home / "unsloth_studio" / "bin" / "unsloth"
tried_lines = "\n".join(f" - {p}" for p in attempted) or " (none)"
raise SystemExit(
"[ERROR] Unsloth frontend build not found.\n"
f"Tried:\n{tried_lines}\n"
"\n"
"Likely cause: another 'unsloth' on PATH is shadowing the "
"installer's binary and points at a site-packages tree with "
"no built dist.\n"
"\n"
"Fix one of:\n"
f" - run the installer's binary directly: {installer_bin} studio\n"
" - pass --frontend <path/to/studio/frontend/dist>\n"
" - pass --api-only to skip serving the web UI\n"
" - reinstall: curl -fsSL https://unsloth.ai/install.sh | sh"
)
# Resolve once; shared by the log rewrite and banner.
display_host = _display_host_for_bind(host)
_install_uvicorn_startup_log_rewrite(host, display_host)
logger.info(
"run_server pre-uvicorn setup completed in %.1fms",
(time.perf_counter() - boot_started) * 1000,
)
ready_event = Event()
startup_failed = Event()
startup_errors = []
class _ReadyServer(uvicorn.Server):
async def startup(self, *args, **kwargs):
await super().startup(*args, **kwargs)
if getattr(self, "started", False) and not self.should_exit:
logger.info(
"Uvicorn startup hook completed in %.1fms",
(time.perf_counter() - boot_started) * 1000,
)
ready_event.set()
# server_header=False suppresses uvicorn's "Server: uvicorn"; SecurityHeadersMiddleware sets its own.
config_kwargs = dict(
host = host,
port = port,
log_level = "info",
access_log = False,
server_header = False,
)
# Colab only: trust X-Forwarded-* from Colab's reverse proxy so the app sees
# the real https origin. forwarded_allow_ips="*" is safe in Colab's
# single-user sandbox but too lax for local/standalone, so leave uvicorn's
# loopback-only default elsewhere.
if _IS_COLAB:
config_kwargs["proxy_headers"] = True
config_kwargs["forwarded_allow_ips"] = "*"
config = uvicorn.Config(app, **config_kwargs)
_server = _ReadyServer(config)
_shutdown_event = Event()
# Expose the actual bound port so handlers build loopback URLs at the real
# backend, not whatever a proxy/tunnel exposed. For ephemeral binds (port==0)
# leave it unset so handlers fall back to the request scope / base_url.
app.state.server_port = port if port and port > 0 else None
# Direct (non-tunnel) base for the API panel; resolve wildcard binds to the LAN IP.
if port and port > 0:
_direct_host = _display_host_for_bind(host)
app.state.server_url = f"http://{_url_host(_direct_host)}:{port}"
else:
app.state.server_url = None
app.state.secure = secure
app.state.llama_parallel_slots = llama_parallel_slots
# Expose a shutdown callable before the server accepts requests so
# /api/shutdown is ready as soon as readiness publishes.
def _trigger_shutdown():
_graceful_shutdown(_server)
if _shutdown_event is not None:
_shutdown_event.set()
app.state.trigger_shutdown = _trigger_shutdown
# A supplied --password / UNSLOTH_STUDIO_PASSWORD / stdin sets the initial
# admin password before the gate and socket bind (direct `python run.py`;
# the CLI applies it in its own parent).
_apply_supplied_password(password)
# Never publish with the seeded default password active: prompt first (or
# warn / fail closed headless; see _terminal_password_gate). Runs BEFORE the
# socket binds so a pre-gate listener can't hand out the injected credential.
_pw_proceed, _pw_drop_bootstrap = _terminal_password_gate(
tunnel_will_start = _cloudflare_tunnel_should_start(
cloudflare = cloudflare,
host = host,
secure = secure,
api_only = api_only,
is_colab = _IS_COLAB,
),
host = host,
secure = secure,
api_only = api_only,
frontend_served = bool(frontend_path) and not api_only,
is_colab = _IS_COLAB,
)
if not _pw_proceed:
print(
"Not starting Unsloth; set a new admin password first, or launch "
"without --secure/--cloudflare.",
file = sys.stderr,
flush = True,
)
sys.exit(1)
if _pw_drop_bootstrap:
# Password just changed (stale) or a public URL is about to serve the
# default credential: don't leak it in the HTML. Lifespan runs AFTER this
# and re-reads the bootstrap password, so the flag (not a plain None)
# makes it skip that re-read.
app.state.suppress_bootstrap_injection = True
app.state.bootstrap_password = None
# Run server in a daemon thread with explicit new_event_loop() +
# run_until_complete() (not asyncio.run) so nest_asyncio's patches don't
# interfere when Colab/IPython already runs a loop on the main thread.
def _run():
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
try:
loop.run_until_complete(_server.serve())
except BaseException as exc:
startup_errors.append(exc)
startup_failed.set()
finally:
loop.close()
if not ready_event.is_set():
startup_failed.set()
thread = Thread(target = _run, daemon = True)
_server_thread = thread
thread.start()
# Wait until uvicorn finishes lifespan startup and binds sockets, or until it
# exits/fails first. No deadline: a slow but live startup stays in progress.
try:
while not ready_event.is_set():
if startup_failed.is_set() or not thread.is_alive():
if startup_errors:
raise RuntimeError(
"Uvicorn server failed before startup completed"
) from startup_errors[0]
raise RuntimeError("Uvicorn server exited before startup completed")
ready_event.wait(timeout = 0.1)
except KeyboardInterrupt:
_graceful_shutdown(_server)
_shutdown_event.set()
raise
logger.info(
"run_server uvicorn ready after %.1fms",
(time.perf_counter() - boot_started) * 1000,
)
_write_pid_file()
import atexit
atexit.register(_remove_pid_file)
from utils.process_lifetime import terminate_all
atexit.register(terminate_all)
# Output port for Tauri (api-only), only after sockets bind and startup done.
# The headless `run --api-only` path opts out so it does not leak this line.
if api_only and emit_tauri_port:
print(f"TAURI_PORT={port}", flush = True)
# Free trycloudflare.com tunnel for wildcard binds (the raw ip:port is often
# unreachable). Started pre-banner and even when silent so the CLI banner can
# read app.state.cloudflare_url; torn down by _graceful_shutdown.
global _cloudflare_url, _cloudflare_requested, _cloudflare_flag
_cloudflare_url = None
_cloudflare_flag = cloudflare
app.state.cloudflare_url = None
_cloudflare_enabled = _cloudflare_tunnel_should_start(
cloudflare = cloudflare,
host = host,
secure = secure,
api_only = api_only,
is_colab = _IS_COLAB,
)
_cloudflare_requested = _cloudflare_enabled
if _cloudflare_enabled:
try: # best-effort: any failure must not block startup
from cloudflare_tunnel import start_studio_tunnel, stop_studio_tunnel
_cloudflare_url = start_studio_tunnel(port)
app.state.cloudflare_url = _cloudflare_url
# Backstop: tear the tunnel down even on an abnormal exit that bypasses
# _graceful_shutdown (e.g. an exception after startup -> sys.exit). Idempotent.
atexit.register(stop_studio_tunnel)
except Exception as e:
logger.debug("Cloudflare tunnel skipped: %s", e)
# --secure fails closed: no tunnel means no public link, so exit rather than
# silently fall back to a raw port.
if secure and not _cloudflare_url:
print(
"A secure Cloudflare link is not allowed, use --no-secure which provides a 0.0.0.0 link",
file = sys.stderr,
flush = True,
)
_graceful_shutdown(_server)
sys.exit(1)
# Time-box a freshly-exposed web UI: if nobody changes the seeded admin
# password within the deadline (default 1h), shut down rather than leave an
# unsecured public instance running. No-op for loopback, --api-only, Colab,
# an already-changed password, or UNSLOTH_STUDIO_BOOTSTRAP_TIMEOUT=0.
try:
from auth import storage as _auth_storage
from auth.bootstrap_timeout import (
arm_bootstrap_timeout,
bootstrap_timeout_seconds,
should_arm_bootstrap_timeout,
)
_bootstrap_timeout = bootstrap_timeout_seconds()
if should_arm_bootstrap_timeout(
host = host,
secure = secure,
api_only = api_only,
frontend_served = bool(frontend_path) and not api_only,
is_colab = _IS_COLAB,
requires_change = _auth_storage.requires_password_change(
_auth_storage.DEFAULT_ADMIN_USERNAME
),
timeout_seconds = _bootstrap_timeout,
):
arm_bootstrap_timeout(
_auth_storage,
_trigger_shutdown,
timeout_seconds = _bootstrap_timeout,
logger = logger,
)
logger.info(
"Unsloth will shut down in %ds unless the default admin password is changed.",
_bootstrap_timeout,
)
except Exception as e: # best-effort: never block startup on the timeout
logger.warning("Bootstrap timeout not armed: %s", e)
if not silent:
_emit_startup_output(host, port, display_host, secure = secure, enable_tools = enable_tools)
return app
def _build_arg_parser():
"""Build the backend CLI argument parser.
Extracted from the __main__ block so the flag wiring (notably the
--secure/--no-secure polarity and its --not-secure alias) stays unit-testable.
"""
import argparse
parser = argparse.ArgumentParser(description = "Run Unsloth UI Backend server")
parser.add_argument(
"--host",
default = "127.0.0.1",
help = "Host to bind to (default: 127.0.0.1; use 0.0.0.0 for network/cloud access)",
)
parser.add_argument(
"--password",
default = None,
help = "Set the INITIAL admin password non-interactively (headless), only when "
"none is set yet. Also reads UNSLOTH_STUDIO_PASSWORD, or --password - for stdin. "
"A literal value is visible in the process list. Rotate later via "
"`unsloth studio reset-password`.",
)
parser.add_argument("--port", type = int, default = 8888, help = "Port to bind to")
parser.add_argument(
"--frontend",
type = str,
default = _DEFAULT_FRONTEND_PATH,
help = "Path to frontend build",
)
parser.add_argument("--silent", action = "store_true", help = "Suppress output")
parser.add_argument(
"--api-only",
action = "store_true",
help = "API server only, no frontend (for Tauri)",
)
parser.add_argument(
"--cloudflare",
action = argparse.BooleanOptionalAction,
default = None,
help = "Expose Unsloth on a PUBLIC internet URL via a free Cloudflare HTTPS "
"tunnel, for non-api-only wildcard binds (0.0.0.0 or ::). Off by default; "
"pass --cloudflare to enable it (--secure implies it), --no-cloudflare to "
"force it off. It does not change a raw wildcard bind. If the admin "
"password was never changed, Unsloth asks for a new one in the terminal "
"before publishing the URL.",
)
parser.add_argument(
"--secure",
action = argparse.BooleanOptionalAction,
default = False,
help = "Expose ONLY a Cloudflare HTTPS link: bind localhost and fail closed "
"if the tunnel can't start. Without it, --no-secure also serves the raw "
"0.0.0.0 port, which is reachable from anywhere on the network. If the "
"admin password was never changed, Unsloth asks for a new one in the "
"terminal before publishing the URL.",
)
# Back-compat: accept --not-secure as a hidden alias for --no-secure.
parser.add_argument(
"--not-secure",
dest = "secure",
action = "store_false",
default = argparse.SUPPRESS,
help = argparse.SUPPRESS,
)
# Tri-state tool policy: no flag -> None (tools on, per-request honored);
# --enable-tools/--disable-tools force on/off.
parser.add_argument(
"--enable-tools",
dest = "enable_tools",
action = "store_true",
default = None,
help = "Force server-side tools (web search, code execution) on for "
"every request. Default: on for every bind, per-request setting honored.",
)
parser.add_argument(
"--disable-tools",
dest = "enable_tools",
action = "store_false",
default = None,
help = "Force server-side tools off for every request.",
)
parser.add_argument(
"--disable-dns-pinning",
action = "store_true",
help = "Allow hostname-based web fetches for enterprise proxies. WARNING: weakens "
"DNS-rebinding protection; hostname and redirect validation remain enabled.",
)
parser.add_argument(
"--parallel",
"--n-parallel",
type = int,
default = _PARALLEL_DEFAULT_PLAIN,
help = (
f"llama-server parallel decode slots ({_PARALLEL_MIN}..{_PARALLEL_MAX}). "
f"Default {_PARALLEL_DEFAULT_PLAIN}."
),
)
return parser
# For direct execution (also invoked by CLI via os.execvp / subprocess).
if __name__ == "__main__":
# Correct a conflicting system CUDA on LD_LIBRARY_PATH before torch is
# imported (below, via run_server). Re-execs once on Linux so the dynamic
# linker uses torch's bundled CUDA libs; no-op on other platforms, when
# LD_LIBRARY_PATH is unset or already correct, or after the single re-exec.
_maybe_reexec_for_cuda_ld_path()
import signal
import traceback
# Ensure stderr handles Unicode on Windows (non-ASCII path tracebacks).
if sys.platform == "win32" and hasattr(sys.stderr, "reconfigure"):
try:
sys.stderr.reconfigure(encoding = "utf-8", errors = "replace")
except Exception:
pass
parser = _build_arg_parser()
args = parser.parse_args()
if not _PARALLEL_MIN <= args.parallel <= _PARALLEL_MAX:
parser.error(f"--parallel must be between {_PARALLEL_MIN} and {_PARALLEL_MAX}")
if args.secure and args.cloudflare is False:
parser.error(
"--secure requires the Cloudflare tunnel; do not combine it with --no-cloudflare"
)
if args.disable_dns_pinning:
os.environ["UNSLOTH_STUDIO_DISABLE_DNS_PINNING"] = "1"
else:
os.environ.setdefault("UNSLOTH_STUDIO_DISABLE_DNS_PINNING", "0")
kwargs = dict(
host = args.host,
port = args.port,
silent = args.silent,
api_only = args.api_only,
llama_parallel_slots = args.parallel,
cloudflare = args.cloudflare,
secure = args.secure,
enable_tools = args.enable_tools,
password = args.password,
)
if args.frontend is not None:
kwargs["frontend_path"] = Path(args.frontend)
try:
run_server(**kwargs)
except Exception:
sys.stderr.write("\n")
sys.stderr.write("=" * 60 + "\n")
sys.stderr.write("ERROR: Unsloth Studio failed to start.\n")
sys.stderr.write("=" * 60 + "\n")
traceback.print_exc(file = sys.stderr)
sys.stderr.write("\n")
sys.stderr.write("If a package is missing, try re-running: unsloth studio setup\n")
sys.stderr.flush()
sys.exit(1)
# Signal handler -- ensures subprocess cleanup on Ctrl+C.
def _signal_handler(signum, frame):
# Restore defaults so a second signal force-quits if shutdown stalls.
signal.signal(signal.SIGINT, signal.SIG_DFL)
signal.signal(signal.SIGTERM, signal.SIG_DFL)
if hasattr(signal, "SIGBREAK"):
signal.signal(signal.SIGBREAK, signal.SIG_DFL)
_graceful_shutdown(_server)
_shutdown_event.set()
signal.signal(signal.SIGINT, _signal_handler)
signal.signal(signal.SIGTERM, _signal_handler)
# On Windows, some terminals send SIGBREAK for Ctrl+C / Ctrl+Break.
if hasattr(signal, "SIGBREAK"):
signal.signal(signal.SIGBREAK, _signal_handler)
# Keep running until shutdown signal. Event.wait() without a timeout blocks at
# the C level on Linux, preventing SIGINT delivery; a short timeout in a loop
# lets the interpreter process pending signals.
while not _shutdown_event.is_set():
_shutdown_event.wait(timeout = 1)
_wait_for_server_shutdown()