CLI command reference

Interactive agent session (same as bare stimulir).

prompt

Positional argument. Optional first prompt — starts the session with this task.

--session

Resume an existing session by its session key.

Attach as the LOCAL EXECUTOR for a server-side agent session.

session_key

Positional argument. Chat session key (the web composer's session id). Omit to pick from recent sessions.

Explicitly export a bounded typed usage-event projection to a file.

--output, -o required

Destination JSON file.

--limit

See the installed command’s --help for accepted values.

--provider

See the installed command’s --help for accepted values.

--model

See the installed command’s --help for accepted values.

--mode

mi | byok

Report month-to-date meter reconciliation without exposing event bodies.

--json

Emit the typed result as JSON.

Month-to-date HybrIE billing snapshot for the active workspace.

--json

Emit raw JSON output.

Store an encrypted provider credential and validate it.

--provider, -p required

Upstream provider.

--label, -l required

Label for this credential.

--secret

Provider API key (omit to be prompted without echo).

--json

Emit raw JSON output.

List BYOK credentials in the active workspace.

--json

Emit raw JSON output.

Delete a stored credential.

credential_id required

Positional argument. Credential id (from stimulir byok list).

--yes, -y

Skip the confirmation prompt.

--json

Emit raw JSON output.

Re-validate a stored credential against its provider.

credential_id required

Positional argument. Credential id (from stimulir byok list).

--json

Emit raw JSON output.

Configure Codex App to use Stimulir as a custom model provider.

--model, -m

Stimulir model id Codex should use.

--project

Project id to send as X-Project-Id.

--staging

Use the Stimulir staging API base.

--api-base

Console API origin. Overrides --staging and ~/.stimulir config.

--inference-base-url

Full inference base URL. Defaults to {api_base}/api/v1/inference.

--config

Codex user config path.

--auth

Auth strategy for Codex provider config.

--session-id

Runtime session id used when minting short-lived Codex inference keys.

--ttl-seconds

TTL for short-lived Codex inference keys minted by the auth helper.

--dry-run

Print the patched Codex config instead of writing it.

--yes, -y

Skip confirmation before modifying Codex config.

Switch the Codex model in-place (Desktop's picker can't list custom models — openai/codex #19694 — so this is the supported switcher).

model required

Positional argument. See the installed command’s --help for accepted values.

--config

Codex user config path.

--relaunch

Quit and relaunch the Codex Desktop app so the change loads.

List model ids the Codex launcher advertises as first-class choices.

No command-specific arguments.

Remove the Stimulir-managed Codex provider config.

--config

Codex user config path.

--provider-id

Provider id to remove.

Make the Stimulir models resolvable by name in Codex.

--config

Codex user config path.

Print a bearer token for Codex command-backed auth.

--session-id

Runtime session id used for the short-lived inference key.

--ttl-seconds

TTL for the short-lived inference key.

--api-base

Console API origin used to mint the key.

Remove the application bootstrap so test sessions are refused.

--instance required

See the installed command’s --help for accepted values.

Configure where Console sends signed test-session requests for an instance.

--instance required

See the installed command’s --help for accepted values.

--url required

HTTPS endpoint the application exposes for signed test-session requests.

--secret required

Project agent secret ID shared with the application; never a value.

Show the configured URL and secret reference; the secret value is never returned.

--instance required

See the installed command’s --help for accepted values.

Permanently retire a deployment, drain workers, and retain history.

--instance required

See the installed command’s --help for accepted values.

Register a digest-pinned image manifest and bind it to an instance.

manifest required

Positional argument. See the installed command’s --help for accepted values.

--name required

See the installed command’s --help for accepted values.

--version required

See the installed command’s --help for accepted values.

--instance required

See the installed command’s --help for accepted values.

--key required

Stable deployment retry key; use a new key for a new version.

Start a worker in an existing room; inspect status for readiness.

--instance required

See the installed command’s --help for accepted values.

--room required

See the installed command’s --help for accepted values.

--key required

Stable key for this room attachment.

List deployment availability and worker counts.

--instance required

See the installed command’s --help for accepted values.

Resume a deployment without restarting old dispatches.

--instance required

See the installed command’s --help for accepted values.

Create or rotate a project agent secret from an environment variable.

--value-env required

Name of the environment variable that holds the secret value.

Read the deployment and bounded dispatch history.

--instance required

See the installed command’s --help for accepted values.

Request one worker to stop; shutdown is asynchronous.

--deployment required

See the installed command’s --help for accepted values.

--instance required

See the installed command’s --help for accepted values.

Prevent new dispatches and drain workers.

--instance required

See the installed command’s --help for accepted values.

Ask the instance's configured application for a signed test session link.

--instance required

See the installed command’s --help for accepted values.

Terminate an instance.

instance_id required

Positional argument. Instance id.

--yes, -y

Skip the confirmation prompt.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

List compute instances.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

List available GPU offers.

--provider

Compute provider. Omit to list every reachable provider.

--region

Only offers available in this region.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Pair local CUDA, Metal or NPU compute with a project using browser approval.

--runtime-url required

HybrIE node address reachable from Console, without /v1.

--api-base

Console API address.

--once

Register once without heartbeat polling.

--probe-url

Local runtime address for capability discovery; registration still uses --runtime-url.

--backend

auto, cuda, metal or npu.

--memory-gb

Declared memory for Metal/NPU when automatic detection is unavailable.

Register a runtime peer (at least one of --grpc / --realtime is required).

--name

Friendly peer name.

--grpc

Peer gRPC URL.

--realtime

Peer realtime URL.

--instance-id

Compute instance id backing the peer.

--no-probe

Register without probing reachability.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

List registered runtime peers.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Remove a runtime peer.

peer_id required

Positional argument. Peer id (from stimulir compute peers list).

--yes, -y

Skip the confirmation prompt.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Show a single instance.

instance_id required

Positional argument. Instance id.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Launch instance(s) from an offer.

offer_id required

Positional argument. Offer id (from stimulir compute offers).

--region required

Region to launch in.

--ssh-key required

SSH key name registered with the provider (repeatable).

--count

Number of instances.

--provider

Compute provider.

--spot

Request spot capacity.

--name

Instance name/tag.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Provision, bootstrap and register an exact managed training worker.

offer_id required

Positional argument. Offer id from stimulir compute offers.

--region required

Region with live capacity.

--ssh-key required

Provider SSH key name (repeatable).

--count

See the installed command’s --help for accepted values.

--provider

See the installed command’s --help for accepted values.

--spot

See the installed command’s --help for accepted values.

--model-repo required

Exact Hugging Face base-model repository.

--model-slug required

HybrIE on-disk family slug.

--hybrie-ref required

Reviewed HybrIE branch or commit to build.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

See the installed command’s --help for accepted values.

Stage multiple assets together for cleaning, snapshotting, Lab, SFT, or eval.

--ids required

Comma-separated data asset IDs.

--stage required

See the installed command’s --help for accepted values.

--status

See the installed command’s --help for accepted values.

--target

eval | sft | preference.

--json

Emit raw JSON output.

Register a data asset (metadata only; use upload to attach a file).

--name, -n required

Asset display name.

--kind

See the installed command’s --help for accepted values.

--stage

See the installed command’s --help for accepted values.

--target

eval | sft | preference.

--description

See the installed command’s --help for accepted values.

--content-type

See the installed command’s --help for accepted values.

--json

Emit raw JSON output.

Curate a data asset directly from a trace.

trace_id required

Positional argument. Trace UUID (from stimulir lab eval ... or the traces panel).

--source required

See the installed command’s --help for accepted values.

--name, -n

See the installed command’s --help for accepted values.

--target

See the installed command’s --help for accepted values.

--json

Emit raw JSON output.

List data assets in the active workspace.

--cursor

Pagination cursor.

--limit

See the installed command’s --help for accepted values.

--target

Filter by eval | sft | preference.

--json

Emit raw JSON output.

Remove a data asset (irreversible).

asset_id required

Positional argument. Asset ID.

--yes, -y

Skip the confirmation prompt.

--json

Emit raw JSON output.

Snapshot the current state of an asset (immutable point-in-time copy).

asset_id required

Positional argument. Asset ID.

--json

Emit raw JSON output.

Transition an asset's lifecycle stage (raw → cleaning → … → lab).

asset_id required

Positional argument. Asset ID.

stage required

Positional argument. See the installed command’s --help for accepted values.

--status

See the installed command’s --help for accepted values.

--json

Emit raw JSON output.

Move a staged asset back to raw so it can be cleaned or re-targeted.

asset_id required

Positional argument. Asset ID.

--json

Emit raw JSON output.

Update data asset metadata without changing the underlying stored payload.

asset_id required

Positional argument. Asset ID.

--name, -n

Asset display name.

--description

See the installed command’s --help for accepted values.

--target

eval | sft | preference.

--status

See the installed command’s --help for accepted values.

--json

Emit raw JSON output.

Upload a file as a new data asset.

file required

Positional argument. File to upload.

--name, -n

Override the asset name.

--stage

See the installed command’s --help for accepted values.

--target

See the installed command’s --help for accepted values.

--description

See the installed command’s --help for accepted values.

--json

Emit raw JSON output.

Ask a panel of models and stream the judge-fused answer.

prompt required

Positional argument. The question to answer with the fusion panel.

--panel

Comma-separated panel model ids (overrides the default panel).

--judge

Judge model id (overrides the default judge).

--max

Use stimulir/fusion-max (claude-sonnet-4-6 + claude-opus-4-6, judged by claude-opus-4-6).

--staging

Use the Stimulir staging API base.

--project

Project id to send as X-Project-Id.

--no-stream

Wait for the full answer instead of streaming (SSE).

Send a chat completion request.

prompt required

Positional argument. The user prompt.

--model, -m required

Model id (see stimulir models).

--stream

Stream tokens live (SSE).

--temperature, -t

See the installed command’s --help for accepted values.

--max-tokens

See the installed command’s --help for accepted values.

--json

Emit the raw response JSON (non-stream only).

Create an API key — the plaintext key is shown exactly once.

--name, -n required

Human-readable key name.

--env

Key environment.

--expires-in-days

Expire the key after N days.

--credit-limit-pence

Spend cap in pence.

--save

Store the key in ~/.stimulir/credentials.json for stimulir infer.

--json

Emit raw JSON output.

List API keys in the active workspace.

--include-revoked

Include revoked keys.

--json

Emit raw JSON output.

Revoke an API key (irreversible).

key_id required

Positional argument. API key id (from stimulir keys list).

--yes, -y

Skip the confirmation prompt.

--json

Emit raw JSON output.

Show an adapter's details.

adapter_id required

Positional argument. Adapter id.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

List registered adapters.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Load an adapter into the runtime.

adapter_id required

Positional argument. Adapter id.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Unload an adapter from the runtime.

adapter_id required

Positional argument. Adapter id.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Internalize one document on explicitly selected managed compute.

document required

Positional argument. See the installed command’s --help for accepted values.

--compute-offer required

Managed GPU offer id.

--region required

Managed GPU region.

--gpus

Managed GPU instance count.

--spot

Use spot pricing when supported.

--compute-provider

See the installed command’s --help for accepted values.

--max-duration-hours

See the installed command’s --help for accepted values.

--max-cost-gbp

See the installed command’s --help for accepted values.

--name

See the installed command’s --help for accepted values.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Record that a steer was consumed. Write-once and idempotent.

run_id required

Positional argument. Eval run id.

steer_id required

Positional argument. Steer id to acknowledge.

--consumed-by required

Who consumed it — an agent session id, not a user id.

--note

What was DONE, e.g. 'applied as derive on run X'.

--json

Emit raw JSON output.

Score a PEFT LoRA adapter directory on held-out retrieval (vs base model).

--family, -f required

Base model family: qwen3-4b, qwen3-0.6b, or mistral-7b.

--adapter-dir, -a required

Adapter directory with PEFT LoRA weights to score.

--examples

Held-out examples to score.

--seed

Seed for reproducible example generation.

--device

auto, metal, cuda, or cpu.

--model-dir

Optional base-model directory override.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Emit lineage identifiers, fingerprints, counts, and outcomes only.

run_id required

Positional argument. Explicit eval run id.

Emit bounded machine-readable status without raw eval or steer content.

run_id required

Positional argument. Explicit eval run id.

List promoted champions, newest promotion first.

--comparability-key

Only the champion(s) of one comparability bucket.

--limit

See the installed command’s --help for accepted values.

--json

Emit raw JSON output.

Create a durable prompt/data/inference eval run.

--name, -n

Eval suite display name.

--data-asset-id

Staged data asset or trace snapshot id. Repeat to use one exact multi-case cohort.

--prompt

Prompt ref as KEY, KEY:VERSION, or KEY:LABEL. Can be repeated.

--provider

Baseline provider.

--model

Baseline model or endpoint model name.

--adapter-id

Optional adapter id to hot-swap as a candidate.

--adapter-version

Exact registered adapter version.

--base-model-id

Exact base model recorded by the adapter manifest.

--adapter-format

Registered adapter format, for example peft-lora.

--evaluator-id

Attach an evaluator. Its definition governs scoring, and its invariants gate every candidate BEFORE any judge spend (see evaluator-create).

--execute

Create the run and start it now.

--leave-queued

Create the run QUEUED and do NOT start it. You start it yourself with execute-run.

--json

Emit raw JSON output.

Archive (default) or hard-delete eval runs.

run_ids required

Positional argument. One or more eval run ids.

--hard

Destroy the rows instead of archiving. Gated on lineage.

--include-descendants

With --hard, also destroy every run branched below the selection.

--yes, -y

Skip the confirmation prompt.

--json

Emit raw JSON output.

Branch a child run from one candidate of a parent run.

run_id required

Positional argument. Parent run id to branch from.

--rationale required

The hypothesis this branch tests. Required — a branch with no stated hypothesis is a rerun.

--prompt-file

File holding the new prompt content. Pushed as a new version of the SOURCE arm's prompt key.

--prompt-ref

Existing prompt as KEY, KEY:VERSION or KEY:LABEL. Its key must match the source arm's.

--kind

Derive kind. adapter_warm_start is the only unavailable training operation.

--source-candidate-key

Branch from this arm instead of the parent's best-scoring one.

--adapter-id

Adapter id for an adapter_hot_swap derive.

--adapter-version

Exact adapter version for an adapter_hot_swap derive.

--base-model-id

Exact base model recorded by the adapter manifest.

--adapter-format

Adapter format for an adapter_hot_swap derive.

--candidate-provider

Explicit provider for a model_substitution or prompt_and_model derive.

--candidate-model

Explicit model for a model_substitution or prompt_and_model derive.

--instruction

Immutable provenance: what this child was steered with. Usually a steer body.

--allow-rejected

Override the rejection ledger: derive even though this exact candidate (same content, key, provider, model and route) was already rejected in this measurement context. Without it the API refuses with a 409.

--stop-parent

Skip the parent's PENDING results. PERMANENT: the parent becomes a partial measurement.

--start/--no-start

Queue and spawn the child now (default), or leave it DRAFT.

--idempotency-key

Retry-safe key. A repeat under the same key returns the first child and spends nothing.

--max-cases

Cap cases drained by the first execution.

--max-candidates

Cap candidates drained by the first execution.

--json

Emit raw JSON output.

Create a reusable evaluator whose invariants gate candidates pre-spend.

--name, -n required

Evaluator display name.

--forbidden-phrase

A candidate whose prompt contains this exact text is SKIPPED before any judge spend and recorded as a rejection. Repeatable.

--required-phrase

A candidate whose prompt is MISSING this text is skipped. Repeatable.

--max-prompt-chars

Skip a candidate whose prompt exceeds this length.

--invariant-key

Identifier recorded on rejections this invariant produces.

--description

Human note stored with the evaluator.

--json

Emit raw JSON output.

List evaluators visible to the active workspace.

--include-archived

Include archived evaluators.

--json

Emit raw JSON output.

Execute a queued durable Lab eval run.

run_id required

Positional argument. Eval run id.

--json

Emit raw JSON output.

Show one durable Lab eval run: best arm, lineage, review and pending work.

run_id required

Positional argument. Eval run id.

--json

Emit raw JSON output.

Score a trained adapter checkpoint on held-out NIAH retrieval (vs zero-LoRA baseline).

--family, -f required

Base model family: qwen3-4b, qwen3-0.6b, or mistral-7b.

--checkpoint-dir, -c required

Trained checkpoint directory (metadata.json + model.safetensors).

--examples

Held-out examples to score (default 200).

--seed

Seed for reproducible example generation.

--device

auto, metal, cuda, or cpu.

--model-dir

Optional base-model directory override.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Apply a promotion proposal: move the production label and pin the champion.

proposal_id required

Positional argument. Proposal id to apply (see proposals).

--reason

Note recorded with the promotion.

--yes, -y

Skip the confirmation prompt.

--json

Emit raw JSON output.

Prepare a proposal for human review without applying or promoting it.

run_id required

Positional argument. Completed eval run id whose winner should be reviewed.

Emit one bounded proposal action without narratives or arbitrary payloads.

proposal_id required

Positional argument. Explicit proposal id to review.

List promotion proposals — a completed run's recommended label move.

--status

Filter by status (proposed, applied, dismissed).

--limit

See the installed command’s --help for accepted values.

--json

Emit raw JSON output.

List the rejection ledger, newest first.

--comparability-key

Only rejections in one comparability bucket.

--identity-hash

Only rejections of one candidate identity.

--run

Only rejections recorded on one run.

--limit

See the installed command’s --help for accepted values.

--json

Emit raw JSON output.

Score a policy on an RL environment and report reward statistics.

--family, -f required

Base model family: qwen3-4b, qwen3-0.6b, or mistral-7b.

--environment

RL environment (e.g. niah).

--policy

Policy to score: base, hypernet, or peft-lora.

--checkpoint-dir, -c

Trained checkpoint directory (required for non-base policies).

--tasks

Number of eval tasks (sent as num_tasks).

--max-new-tokens

Generation cap per task.

--temperature

Sampling temperature.

--seed

Seed for reproducible task generation.

--pass-threshold

Reward threshold above which a task counts as passed.

--device

auto, metal, cuda, or cpu.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output (includes the per_task list).

List durable Lab eval runs, ranked by their best ARM rather than a run mean.

--limit

See the installed command’s --help for accepted values.

--status

Filter by draft, queued, running, completed, or failed.

--include-archived

Also list archived runs (hidden by default).

--json

Emit raw JSON output.

Attach an instruction to a run for whichever agent polls it next.

run_id required

Positional argument. Eval run id.

--body required

The instruction, in plain text.

--kind

instruction or note.

--idempotency-key

Retry-safe key; a repeat returns the existing steer instead of a second one.

--json

Emit raw JSON output.

List a run's steers, newest first, with each consumption note.

run_id required

Positional argument. Eval run id.

--pending

Only steers nobody has consumed yet.

--include-archived

Allow reading an archived run's steers.

--json

Emit raw JSON output.

Show the whole lineage tree containing a run: nodes, best-per-bucket, pending work.

run_id required

Positional argument. Any run id in the tree — root, leaf or middle.

--include-archived

Include archived member runs.

--json

Emit raw JSON output.

Let the server diagnose, propose and start one measured next step.

run_id required

Positional argument. Improve run id.

--instruction

Optional operator constraint for the next proposal.

--idempotency-key

Retry-safe continuation key.

--json

Emit raw JSON output.

--env-file

Adopter dotenv file containing Stimulir app context.

Read searchable Improve project inventory; this does not start a run.

--env-file

Adopter dotenv file containing Stimulir app context.

Show diagnosis, lineage, hypotheses and pending work.

run_id required

Positional argument. Improve run id.

--json

Emit raw JSON output.

--env-file

Adopter dotenv file containing Stimulir app context.

Resolve cohort, incumbent, evaluator, candidates and cost without spend.

--source-window

Relative trace window; overridden by explicit dates or a folder.

--prompt

Prompt target; auto resolves it from trace provenance.

--max-iterations

Maximum diagnostic hill-climb iterations.

--trace-tag

Require production traces carrying this exact tag. Repeatable.

--env-file

Adopter dotenv file containing Stimulir app context.

--from

Inclusive ISO-8601 trace start.

--to

Exclusive ISO-8601 trace end.

--folder-id

Immutable named cohort id.

--folder-name

Immutable named cohort name.

--count

Maximum selected eligible cases.

--sampling

newest or deterministic_random.

--sampling-seed

Stable random sampling seed.

--prompt-key

Exact source prompt key. Repeatable.

--source-model

Exact inference model. Repeatable.

--budget-gbp

Overall budget ceiling.

--experiment-mode

prompt_only, model_only, or prompt_and_model.

--candidate-model

Exact PROVIDER:MODEL candidate. Repeatable.

--baseline-model

Explicit PROVIDER:MODEL execution baseline for an imported benchmark cohort.

--allow-managed-inference/--no-managed-inference

Allow Stimulir managed models. Unset means the project default: on for demo projects, off otherwise.

--evaluator-id

Exact published evaluator id.

--evaluator-key

Exact published evaluator key.

--evaluator-version

Published evaluator version (with --evaluator-key).

--judge-provider

Explicit provider for an LLM evaluator's judge.

--judge-model

Explicit model for an LLM evaluator's judge.

Show the terminal outcome and evidence; never apply promotion.

run_id required

Positional argument. Improve run id.

--json

Emit raw JSON output.

--env-file

Adopter dotenv file containing Stimulir app context.

Submit one agent-authored diagnostic candidate and detach immediately.

run_id required

Positional argument. Improve run id.

--rationale required

Evidence-backed hypothesis for this iteration.

--prompt-file required

Complete candidate prompt.

--instruction

Optional immutable provenance.

--json

Emit raw JSON output.

--env-file

Adopter dotenv file containing Stimulir app context.

Start a durable diagnostic Improve run and detach; promotion is disabled.

--source-window

Captured-trace window, resolved in the workspace timezone.

--prompt

Prompt target; auto discovers it from the cohort.

--max-iterations

Maximum diagnostic hill-climb iterations.

--idempotency-key

Retry-safe creation key.

--preview-token

Token returned by the reviewed Improve preview.

--cohort-fingerprint

Eligible-cohort fingerprint returned by the reviewed Improve preview.

--trace-tag

Require production traces carrying this exact tag. Repeatable.

--tag

Attach metadata tag to the Improve run. Repeatable.

--env-file

Adopter dotenv file containing Stimulir app context.

--json

Emit raw JSON output.

--from

Inclusive ISO-8601 trace start.

--to

Exclusive ISO-8601 trace end.

--folder-id

Immutable named cohort id.

--folder-name

Immutable named cohort name.

--count

Maximum selected eligible cases.

--sampling

newest or deterministic_random.

--sampling-seed

Stable random sampling seed.

--prompt-key

Exact source prompt key. Repeatable.

--source-model

Exact inference model. Repeatable.

--budget-gbp

Overall budget ceiling.

--experiment-mode

prompt_only, model_only, or prompt_and_model.

--candidate-model

Exact PROVIDER:MODEL candidate. Repeatable.

--baseline-model

Explicit PROVIDER:MODEL execution baseline for an imported benchmark cohort.

--allow-managed-inference/--no-managed-inference

Allow Stimulir managed models. Unset means the project default: on for demo projects, off otherwise.

--evaluator-id

Exact published evaluator id from preview.

--evaluator-key

Exact published evaluator key from preview.

--evaluator-version

Published evaluator version (with --evaluator-key).

--judge-provider

Exact judge provider from preview.

--judge-model

Exact judge model from preview.

Show the current durable Improve state and progress.

run_id required

Positional argument. Improve run id.

--json

Emit raw JSON output.

--env-file

Adopter dotenv file containing Stimulir app context.

Cancel a running training job (RL and PEFT SFT jobs — D2L SFT runs to completion).

job_id required

Positional argument. Training job id.

--yes, -y

Skip the confirmation prompt.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Show a training job's status and latest progress points.

job_id required

Positional argument. Training job id.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

List training jobs.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Create a multimodal projector tuning run — the third training route.

--modality, -m required

Input modality: image (vision tower) or audio (Whisper encoder).

--encoder

Frozen encoder tap (default per modality: SigLIP for image, Whisper for audio).

--stages required

stage1 trains the projector only; stage1+2 also jointly tunes a LoRA adapter.

--lora-rank

Stage-2 LoRA rank (stage1+2 only).

--lora-alpha

Stage-2 LoRA alpha (stage1+2 only).

--dataset

Workspace data asset id with paired multimodal examples.

--name, -n

Display name for the run.

--json

Emit raw JSON output.

Delete a projector tuning run record (in-flight HybrIE jobs run to completion).

run_id required

Positional argument. Projector tuning run id.

--yes, -y

Skip the confirmation prompt.

Show a projector tuning run — status, encoder, stages, and HybrIE job linkage.

run_id required

Positional argument. Projector tuning run id.

--json

Emit raw JSON output.

List multimodal projector tuning runs.

--modality

Filter by modality: image or audio.

--status

Filter by queued, running, completed, failed, or cancelled.

--limit

See the installed command’s --help for accepted values.

--json

Emit raw JSON output.

Submit a D2L (Doc-to-LoRA) training job — hypernetwork context internalisation.

--family, -f required

See the installed command’s --help for accepted values.

--examples

Number of training examples (sent as num_examples).

--eval-examples

Held-out eval examples scored after training.

--min-blocks

Minimum synthetic context blocks.

--max-blocks

Maximum synthetic context blocks.

--log-every

Progress interval; 0 disables intermediate reports.

--epochs

Training epochs.

--lr

Learning rate.

--seed

Seed for reproducible example generation.

--checkpoint-dir

Directory to write the trained checkpoint into.

--device

Target device (auto, cuda, mps, cpu).

--compute-offer

Managed GPU offer id.

--region

Managed GPU region.

--gpus

Managed GPU instance count.

--spot

Use spot pricing when the offer supports it.

--compute-provider

Managed compute provider.

--max-duration-hours

Hard managed-run time ceiling.

--max-cost-gbp

Optional GPU-spend ceiling.

--preview

Validate and price only; create no job and launch nothing.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Submit an RL training job.

--family, -f required

See the installed command’s --help for accepted values.

--environment

RL environment (e.g. niah).

--prompts

Number of training prompts.

--group-size

Rollout group size per prompt.

--policy

Policy to train: hypernet or peft-lora.

--lora-rank

LoRA rank for peft-lora policies.

--init-checkpoint-dir

Checkpoint directory to initialise the policy from.

--lr

Learning rate.

--kl-beta

KL penalty coefficient.

--seed

Seed for reproducible rollouts.

--device

Target device (auto, metal, cuda, cpu).

--checkpoint-dir

Directory to write the trained checkpoint into.

--compute-offer

Managed GPU offer id.

--region

Managed GPU region.

--gpus

Managed GPU instance count.

--spot

Use spot pricing when the offer supports it.

--compute-provider

Managed compute provider.

--max-duration-hours

Hard managed-run time ceiling.

--max-cost-gbp

Optional GPU-spend ceiling.

--preview

Validate and price only; create no job and launch nothing.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Submit an SFT job training a plain PEFT LoRA adapter (job kind sft-peft-lora).

--family, -f required

See the installed command’s --help for accepted values.

--lora-rank

LoRA rank of the trained adapter (default: family preset).

--init-adapter-dir

PEFT adapter directory to warm-start from.

--examples

Number of training examples (sent as num_examples).

--epochs

Training epochs.

--lr

Learning rate.

--l1-coef

L1 regularisation coefficient.

--eval-examples

Held-out eval examples scored after training.

--min-blocks

Minimum synthetic context blocks.

--max-blocks

Maximum synthetic context blocks.

--log-every

Progress interval; 0 disables intermediate reports.

--seed

Seed for reproducible example generation.

--device

Target device (auto, metal, cuda, cpu).

--checkpoint-dir

Directory to write the trained adapter into.

--dataset

Immutable Lab data asset id.

--split-ratio

Fraction used for training.

--split-seed

Deterministic case split seed.

--eval-cohort

Held-out comparable eval run id.

--compute-offer

Managed GPU offer id.

--region

Managed GPU region.

--gpus

Managed GPU instance count.

--spot

Use spot pricing when the offer supports it.

--compute-provider

Managed compute provider.

--max-duration-hours

Hard managed-run time ceiling.

--max-cost-gbp

Optional GPU-spend ceiling.

--preview

Validate and price only; create no job and launch nothing.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Configure and open Codex App with Stimulir as the model provider.

--model, -m

Stimulir model id Codex should use.

--project

Project id to send as X-Project-Id.

--staging

Use the Stimulir staging API base.

--api-base

Console API origin. Overrides --staging and ~/.stimulir config.

--inference-base-url

Full inference base URL. Defaults to {api_base}/api/v1/inference.

--config

Codex user config path.

--auth

Auth strategy for Codex provider config.

--session-id

Runtime session id used when minting short-lived Codex inference keys.

--ttl-seconds

TTL for short-lived Codex inference keys minted by the auth helper.

--dry-run

Print the patched Codex config instead of writing it.

--restore

Remove the Stimulir-managed Codex config instead of configuring.

--no-launch

Configure Codex without opening the app.

--yes, -y

Skip confirmation before modifying Codex config.

Log in to Stimulir.

--token

Paste a stim_cli_ token (Settings -> CLI Tokens) instead of the browser flow.

--device

Force the browser device flow (the default when no --token is given).

--headless

Device flow without auto-opening the browser — prints the link (SSH/remote).

Log out and delete stored credentials.

--remote

Also revoke the device-auth (stim_cli_) token server-side before clearing it.

List available models.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Show supported anonymization methods and recognizer count.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

De-identify PII in text. Only the redacted text is returned — no raw PII.

--text, -t required

Text to de-identify.

--method, -m

replace | redact | mask | hash_sha256 | hash_sha512 | keep | replace_consistent.

--mask-char

Mask character (for --method mask).

--keep-tail

Trailing chars to keep visible (for --method mask).

--placeholder

Replacement placeholder (for --method replace).

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

De-identify bounded UTF-8 stdin; sensitive source text never enters argv.

--method, -m

replace | redact | mask | hash_sha256 | hash_sha512 | keep | replace_consistent.

--mask-char

See the installed command’s --help for accepted values.

--keep-tail

See the installed command’s --help for accepted values.

--placeholder

See the installed command’s --help for accepted values.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit the typed result as JSON.

Detect PII entities in text.

--text, -t required

Text to scan for PII.

--endpoint

HybrIE runtime URL (bypasses the console proxy).

--json

Emit raw JSON output.

Clear the active project — attribute at the workspace level again.

No command-specific arguments.

List projects in the active workspace (the active one is marked).

--json

Emit raw JSON output.

Set the active project (sent as X-Project-Id, saved to ~/.stimulir/config.json).

project_id required

Positional argument. Project id to attribute requests to.

Archive a prompt version by marking it inactive.

key required

Positional argument. Prompt key.

version required

Positional argument. Version number to archive.

--json

Emit raw JSON output.

Create a prompt (or a new version of an existing key).

--key, -k required

Stable prompt key.

--content

Prompt body (or use --file / stdin).

--file, -f

Read prompt body from this file.

--name

Display name.

--type

Optional prompt type tag.

--label

Label to attach to this version (e.g. prod).

--notes

Change notes for this version.

--json

Emit raw JSON output.

Fetch a prompt by key plus label/version.

key required

Positional argument. Prompt key.

--label

Label to fetch, e.g. prod or staging.

--version

Specific version to fetch.

--json

Emit raw JSON output.

Move a label to a specific version (promote or roll back).

key required

Positional argument. Prompt key.

version required

Positional argument. Version number to label.

label required

Positional argument. Label to set on this version.

--json

Emit raw JSON output.

List versioned prompts in the active workspace.

--json

Emit raw JSON output.

Update prompt metadata. Prompt content remains immutable; create a new version to edit content.

key required

Positional argument. Prompt key.

version required

Positional argument. Version number to update.

--name

Display name.

--notes

Change notes.

--active/--inactive

Mark this version active or inactive.

--json

Emit raw JSON output.

Show the full version history for a prompt.

key required

Positional argument. Prompt key (from stimulir prompts list).

--json

Emit raw JSON output.

List past dedicated-task runs for the active workspace.

--json

Print raw JSON.

Pipe skills so each step's output feeds the next (synchronous, per step).

--step required

Ordered skill step: 'slug' or 'slug:k=v,k2=v2'. Repeat for each step, in order.

--input-key required

Storage key of the initial input.

--output-key-only

On success, print ONLY the bare final output_key (for KEY=$(...) shell capture).

--json

Print raw JSON (ordered per-step results).

Run a skill against a stored input in an isolated sandbox (synchronous).

skill required

Positional argument. Imported skill slug, e.g. evidence-clip.

--input-key required

Storage key of the input (must be under your workspace prefix).

--param, -P

key=value skill param (repeatable).

--output-key-only

On success, print ONLY the bare output_key (for KEY=$(...) shell capture).

--json

Print raw JSON.

Connect this project to Stimulir, from sign-in to the first trace.

--tool

Coding agent to connect. auto detects Claude Code, Cursor or Codex; none skips the MCP server.

--project

Project name or id to send traces to. Defaults to your current or default project. A name that does not exist yet is created.

--key-name

Name for the new API key. Defaults to <tool>-<your name>.

--env-file

Env file to write the gateway settings to. Defaults to .env.local or .env in this folder.

--no-test

Skip the test request.

--yes, -y

Accept every default without asking. Coding agents should always pass this.

--base-url

API to set up against, e.g. staging or https://api.staging.stimulir.com. Defaults to the one you signed in to.

--model

Model for the test request.

List the skills a repo exposes (SKILL.md files), without importing.

repo required

Positional argument. GitHub repo as owner/name, e.g. tosi-n/agentic-frame.

--ref

Branch, tag, or commit.

--json

Print raw JSON.

Import selected skills into this workspace's sandbox (pinned to the commit).

repo required

Positional argument. GitHub repo as owner/name.

--path, -p required

Skill path in repo (repeatable), e.g. skills/evidence-clip.

--ref

Branch, tag, or commit.

--json

Print raw JSON.

List skills imported into the active workspace.

--json

Print raw JSON.

Delete a stored object (removes the S3 file and its record).

object_id required

Positional argument. Object id to delete (S3 + row).

List storage folders for the active workspace.

--json

Print raw JSON.

Resolve a storage KEY to a signed URL (or download it) — the loop-closer for a sandbox output_key, which is a key, not an object id.

--key required

Storage key (e.g. a sandbox run/sandbox pipe output_key).

--expires-in

Signed URL TTL in seconds.

--output, -o

Download the bytes to this file instead of printing the signed URL.

Mint a time-limited signed URL for a stored object.

object_id required

Positional argument. Object id (from storage list).

--expires-in

Signed URL TTL in seconds.

--download

Content-Disposition: attachment.

List stored objects for the active workspace.

--task-id

Filter to one dedicated-task run.

--json

Print raw JSON.

Upload an input file into tenant storage; prints its storage key.

file required

Positional argument. Local file to upload.

--name

Storage filename (defaults to the file's name).

Export workspace traces (JSONL/JSON/CSV) for the Stimulir cost estimator.

--output, -o

Write to this file instead of stdout.

--format, -f

jsonl | json | csv.

--source

agent | usage.

--status

completed | failed | running.

--context

Agent trace context_type filter (agent traces only).

--api-key-id

Restrict to usage events for this key.

--tag

Case-insensitive tag filter.

--from

Inclusive lower bound (ISO date or datetime, UTC).

--to

Upper bound; a bare date includes the whole day.

--include-content

Include prompt/response bodies and tool calls.

--limit

Max traces to export.

Persist frps connection settings for tunnel start (once per environment).

--control-host required

frps control host (tunnel_control CNAME for this environment).

--control-port

frps control port.

--domain required

Tunnel wildcard domain, e.g. tunnel.stimulir.com.

--token required

frps auth.token for this environment (from Secrets Manager).

List tracked tunnels (running or stale).

No command-specific arguments.

Start a public HTTPS tunnel to a local port via our self-hosted frps.

--port, -p required

Local port to expose.

--label, -l required

Name for this tunnel (used by stop/list, and as the subdomain).

Stop a running tunnel.

label required

Positional argument. Tunnel label to stop.

Show aggregated inference usage for the active workspace.

--window, -w

Time window.

--group-by, -g

Aggregation dimension.

--json

Emit raw JSON output.

Speak bounded stdin over Stimulir realtime; requires stimulir[realtime].

--output, -o required

See the installed command’s --help for accepted values.

--provider

See the installed command’s --help for accepted values.

--model

See the installed command’s --help for accepted values.

--voice

See the installed command’s --help for accepted values.

--timeout

See the installed command’s --help for accepted values.

Send one 16kHz mono PCM16 WAV utterance through realtime speech-to-speech.

audio required

Positional argument. See the installed command’s --help for accepted values.

--output, -o required

See the installed command’s --help for accepted values.

--provider

See the installed command’s --help for accepted values.

--model

See the installed command’s --help for accepted values.

--voice

See the installed command’s --help for accepted values.

--timeout

See the installed command’s --help for accepted values.

Transcribe one bounded local audio file and emit only the typed transcript.

audio required

Positional argument. See the installed command’s --help for accepted values.

--model

See the installed command’s --help for accepted values.

--language

See the installed command’s --help for accepted values.

--json

See the installed command’s --help for accepted values.

Synthesize bounded stdin to an explicit output file.

--output, -o required

Audio output file.

--model

See the installed command’s --help for accepted values.

--voice

See the installed command’s --help for accepted values.

--format

See the installed command’s --help for accepted values.

Show the current identity, and WHICH credential produces it.

--json

Emit raw JSON output.

List workspaces you belong to (the active one is marked).

--json

Emit raw JSON output.

Set the active workspace (saved to ~/.stimulir/config.json).

workspace_id required

Positional argument. Workspace (business profile) id.