analyze_workflow
ChatGPTList the custom nodes and model files a ComfyUI workflow needs, and how each would be found. Free. Each model comes with its size and a resolution: already on RunComfy, a known download, or one the user must supply (for example a gated Hugging Face file they have no access to). Use it before start_workflow_setup so the user knows what the setup will install.
cancel_model_request
ChatGPTCancel a queued Model API request. Backs `POST /v1/requests/{request_id}/cancel. Returns cancelled if accepted, not_cancellable` if the request is already in progress or finished.
cancel_request
ChatGPTCancel a queued or running request. Backs `POST /prod/v1/deployments/{deployment_id}/requests/{request_id}/cancel. Returns cancelled if accepted, not_cancellable` if already done.
cancel_training_job
ChatGPTCancel a queued or running training job. Backs `POST /prod/v1/trainers/ai-toolkit/jobs/{job_id}/cancel. Progress stops, but get_training_job_result` still returns any checkpoints produced so far.
create_dataset
ChatGPTCreate an empty training dataset. Backs `POST /prod/v1/trainers/datasets. The new dataset starts in DRAFT; upload files into it, then poll get_dataset_status until it reaches READY before submitting a training job. Args: name: Human-readable name, unique within the account. This is the dataset_name an AI Toolkit config references as /app/ai-toolkit/datasets/{dataset_name}`. Omit to let RunComfy generate one.
create_deployment
ChatGPTCreate a Serverless API (ComfyUI) deployment. Backs `POST /prod/v2/deployments`. For LoRA deployments, create via the runcomfy.com UI instead. Args: name: Human-readable name. workflow_id: UUID of the ComfyUI workflow. workflow_version: Version label, e.g. "v1". hardware: One of TURING_16, AMPERE_24, AMPERE_48, ADA_48_PLUS, AMPERE_80, ADA_80_PLUS, HOPPER_141. min_instances: 0..30. Warm instance floor (billable if > 0). max_instances: 1..60. Concurrency ceiling. queue_size: >= 0. Pending requests before scaling up. keep_warm_duration_in_seconds: >= 0. Idle timeout.
delete_assets
ChatGPTDelete files or folders from My Assets. This cannot be undone; confirm with the user first.
delete_dataset
ChatGPTPermanently delete a training dataset. Backs `DELETE /prod/v1/trainers/datasets/{dataset_id}`. This cannot be undone.
delete_deployment
ChatGPTPermanently delete a deployment. Backs `DELETE /prod/v2/deployments/{deployment_id}. This cannot be undone. Consider update_deployment(is_enabled=false)` to pause instead.
delete_workflow_setup
ChatGPTDelete a finished Auto Setup submission and the workflow it created from My Workflows. Queued or running setups cannot be deleted, nor can a workflow a deployment still uses. Confirm with the user first: the workflow is removed from My Workflows.
edit_flux_kontext_pro_image
ChatGPTEdit one image from a text instruction with FLUX Kontext Pro. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="blackforestlabs/flux-1-kontext/pro/edit") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
edit_gpt_image_2_5_image
ChatGPTEdit or combine reference images from a text instruction with GPT Image 2.5 Sunburst. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="openai/gpt-image-2.5/sunburst/edit") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
edit_nano_banana_2_1_image
ChatGPTEdit or combine reference images from a text instruction with Nano Banana 2.1. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="google/nano-banana-2.1/edit") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
edit_nano_banana_2_image
ChatGPTEdit or combine reference images from a text instruction with Nano Banana 2. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="google/nano-banana-2/edit") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
edit_nano_banana_pro_image
ChatGPTEdit or combine reference images from a text instruction with Nano Banana Pro. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="google/nano-banana/pro/edit") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
edit_qwen_image_3_pro_image
ChatGPTEdit images from a text instruction with Qwen Image 3.0 Pro. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="qwen/qwen-image-3.0/pro/edit") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
edit_seedream_5_pro_image
ChatGPTEdit or combine reference images from a text instruction with Seedream 5.0 Pro. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="bytedance/seedream-5.0-pro/image-to-image") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
edit_training_job
ChatGPTReplace the config of a non-running training job. Backs `POST /prod/v1/trainers/ai-toolkit/jobs/{job_id}/edit. Only works while the job is STOPPED, CANCELED, or FAILED, and config.name in the new YAML must still match the original job's name. GPU type and count are chosen at resume time, so call resume_training_job` afterwards to re-queue with the new config.
generate_3d
ChatGPTGenerate a 3D model with a RunComfy 3D model, using RunComfy balance. Find a model_id with list_models(category="image-to-3d"); list_model_categories lists the other categories that end in -3d. Read the model's input_schema and price with get_model, and obtain the user's approval of the model and the price before this billable submission. Put the model's own fields in inputs. A model that does not make 3D models is refused. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_ace_step_music
ChatGPTGenerate a song or instrumental track from style tags and optional lyrics with ACE-Step 1.5. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="acestep-ai/ace-step-1.5/text-to-audio") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_audio
ChatGPTGenerate audio with a RunComfy audio model, using RunComfy balance. Find a model_id with list_models(category="text-to-audio"); list_model_categories lists the other categories that end in -audio. Read the model's input_schema and price with get_model, and obtain the user's approval of the model and the price before this billable submission. Put the model's own fields in inputs. A model that does not make audio is refused. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_flux_schnell_image
ChatGPTGenerate one image with FLUX.1 Schnell text-to-image using RunComfy balance. The model is fixed; no deployment is required. seed=-1 requests a random seed. Read this model's current price with get_model and obtain approval before the billable submission. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first. A key echo does not confirm backend deduplication. A new generation needs approval.
generate_gpt_image_2_5_image
ChatGPTGenerate one or more images from a text prompt with GPT Image 2.5 Sunburst, which keeps words on labels and posters accurate. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="openai/gpt-image-2.5/sunburst/text-to-image") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_heygen_voice_speech
ChatGPTTurn text into a spoken voiceover with HeyGen Voice. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="heygen/heygen-voice/text-to-audio") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_image
ChatGPTGenerate an image with a RunComfy image model, using RunComfy balance. Find a model_id with list_models(category="text-to-image"); list_model_categories lists the other categories that end in -image. Read the model's input_schema and price with get_model, and obtain the user's approval of the model and the price before this billable submission. Put the model's own fields in inputs. A model that does not make images is refused. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_kling_3_pro_video
ChatGPTTurn a starting image into a video with Kling 3.0 Pro. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="kling/kling-3.0/pro/image-to-video") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_qwen_image_3_pro_image
ChatGPTGenerate an image from a text prompt with Qwen Image 3.0 Pro. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="qwen/qwen-image-3.0/pro/text-to-image") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_seedance_2_5_video
ChatGPTGenerate a 720p video from a text prompt and optional reference images, video or audio with Seedance 2.5. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="bytedance/seedance-2.5/reference-to-video/720p") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_seedance_pro_fast_video
ChatGPTGenerate one video from an image with Seedance 1.0 Pro Fast, using RunComfy balance. The model is fixed. duration is seconds; resolution and duration affect cost. Read this model's current pricing with get_model and obtain approval before submission. The image must be available through an unauthenticated HTTPS GET; do not include credentials in its URL. Backend validation of image bytes is separate from URL syntax validation here. Prefer wait_for_completion=false. Keep the returned request ID and use get_model_request_status/get_model_request_result; never generate again to check progress. After an unknown submission outcome, verify history before any retry. An echoed idempotency key is not backend confirmation.
generate_seedream_5_pro_image
ChatGPTGenerate an image from a text prompt with Seedream 5.0 Pro. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="bytedance/seedream-5.0-pro/text-to-image") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_veo_3_1_fast_video
ChatGPTGenerate a video from a text prompt with Veo 3.1 Fast. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="google-deepmind/veo-3-1/fast/text-to-video") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_video
ChatGPTGenerate a video with a RunComfy video model, using RunComfy balance. Find a model_id with list_models(category="image-to-video"); list_model_categories lists the other categories that end in -video. Read the model's input_schema and price with get_model, and obtain the user's approval of the model and the price before this billable submission. Put the model's own fields in inputs. A model that does not make videos is refused. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
generate_wan_3_video
ChatGPTTurn a starting image into a video with Wan 3.0. Uses RunComfy balance. The model is fixed and no deployment is required. Read this model's current price with get_model(model_id="wan-ai/wan-3.0/image-to-video") and obtain approval before the billable submission. File inputs must be public HTTPS URLs without credentials. Prefer wait_for_completion=false: keep the returned request ID, poll get_model_request_status, and fetch outputs with get_model_request_result. Never call this tool to check progress or blindly retry after a timeout; verify request history first.
get_asset_download_url
ChatGPTGet a link that downloads one My Assets file, valid for an hour.
get_asset_import_status
ChatGPTCheck a My Assets import started by import_asset_from_url.
get_asset_upload_url
ChatGPTGet a link to upload one file into My Assets with an HTTP PUT, valid for an hour. An existing file at the path is replaced.
get_balance
ChatGPTGet the account's remaining RunComfy balance. Backs `GET /prod/v2/balance. One wallet funds every product, so this is the figure Serverless deployments, hosted generation requests, and training jobs all draw down — and the one that gets checked before work is allowed to start. Returns balance_usd for reading and balance_microdollars` (millionths of a dollar) for exact arithmetic.
get_dataset_status
ChatGPTGet a dataset's status and its successfully uploaded files. Backs `GET /prod/v1/trainers/datasets/{dataset_id}/status. Lifecycle: DRAFT → UPLOADING → READY (or FAILED, which sets error). Only READY datasets can be mounted by a training job. Files still uploading or failed do not appear in files`.
get_dataset_upload_urls
ChatGPTGet signed upload URLs for dataset files the server cannot fetch. Backs `POST /prod/v1/trainers/datasets/{dataset_id}/get-upload-endpoint. Use this for local files and for anything over 150 MB: the caller PUTs each file's bytes to the returned upload_url using the returned method and headers. Args: dataset_id: Target dataset. filename_to_byte_size: Map of filename → exact size in bytes, e.g. {"img_0001.jpg": 2000000, "img_0001.txt": 12000}. The signature is derived from the size, so a wrong value is rejected by storage at PUT time. Signed URLs are short-lived; call this again for a fresh one if it expires. After every PUT succeeds, poll get_dataset_status until the dataset is READY`.
get_deployment
ChatGPTGet one deployment by ID. Backs `GET /prod/v2/deployments/{deployment_id}. Set include_payload=true to inspect the deployed workflow graph (workflow_api_json) and default overrides — use the node IDs and input names to build the overrides for submit_request`.
get_library_workflow
ChatGPTWhat running a workflow from the public library takes, with no copy in My Workflows. Free. workflow_version_id goes to request_machine_start, and workflow_id with workflow_version to create_deployment. workflow_api_json is the prompt for run_on_machine; its sample files are on the machine, so it runs unchanged. inputs lists the fields to change for other files or text. Finished files land in My Assets under output/, and get_asset_download_url gives a link to them.
get_machine
ChatGPTGet one ComfyUI machine: status, type, ComfyUI link and when it stops.
get_machine_history
ChatGPTWhat finished on a machine's ComfyUI: each prompt's status and its output files.
get_machine_queue
ChatGPTHow many prompts are running and waiting on a machine's ComfyUI.
get_machine_start_request
ChatGPTCheck a machine start request: waiting for the user's approval, started (with the machine), denied, expired or failed. While it waits, check again only after the user says they approved it, or in about two minutes.
get_model
ChatGPTGet current metadata and pricing for one of the two supported models. Accepts only FLUX Schnell text-to-image or Seedance Pro Fast image-to-video. Returns the explicit generation tool name, description, model URL and current pricing. Input schemas are already declared on those generation tools and are not fetched for use with a generic executor.
get_model_request_result
ChatGPTFetch a completed Model API request's outputs. Backs `GET /v1/requests/{request_id}/result. The output shape is defined by the model's Output schema; any hosted asset URLs found in it are also flattened into output_urls. This is for hosted generation requests. Serverless deployment requests use get_request_result` instead.
get_model_request_status
ChatGPTPoll a Model API request's current status. Backs `GET /v1/requests/{request_id}/status. Lifecycle: in_queue → in_progress → completed / cancelled. While in_queue the payload also carries queue_position. This is for hosted generation requests. Serverless deployment requests use get_request_status` instead.
get_request_result
ChatGPTFetch a completed request's outputs. Backs `GET /prod/v1/deployments/{deployment_id}/requests/{request_id}/result. Output URLs are hosted for 7 days. This is for submit_request requests on a deployment. Model API requests from the hosted generation tools use get_model_request_result`.
get_request_status
ChatGPTPoll a request's current status. Backs `GET /prod/v1/deployments/{deployment_id}/requests/{request_id}/status. Lifecycle: in_queue → in_progress → completed / cancelled. This is for submit_request requests on a deployment. Model API requests from the hosted generation tools use get_model_request_status`.
get_training_job_result
ChatGPTFetch a training job's artifacts as hosted URLs. Backs `GET /prod/v1/trainers/ai-toolkit/jobs/{job_id}/result. Returns checkpoints (.safetensors), the resolved config, and sample outputs. Safe to call while the job is still RUNNING — the artifact list grows over time — and after a FAILED or CANCELED job to recover whatever was produced. Config publication is separate from training completion. Only explicit config_publication_status=published confirms backend publication; pending means publication metadata has not arrived, and unconfirmed includes older backends without this field. Never infer publication from STOPPED or a URL's hostname. Raw result.artifacts.config is retained for compatibility, but its URLs are provisional unless published: they appear in provisional_config_urls, not artifact_urls`. Checkpoint and sample URLs remain available unchanged. Publication does not itself verify a download or the file's contents. For pending or unconfirmed publication, the caller may refresh this read-only tool with the SAME job_id up to three times, 30 seconds apart. Each call fetches once; there is no internal publication wait or polling. If still unavailable, report the limitation and retain the job ID. Never submit, resume, edit, rename or retrain to publish a config, and never invent or rewrite an artifact URL. Use checkpoints in a compatible saved ComfyUI workflow. The two hosted generation tools do not accept arbitrary LoRAs.
get_training_job_status
ChatGPTPoll a training job's status and step progress. Backs `GET /prod/v1/trainers/ai-toolkit/jobs/{job_id}/status. Lifecycle: IN_QUEUE → RUNNING → STOPPED (finished or preempted), FAILED (error explains why), or CANCELED`.
get_workflow_save_status
ChatGPTWhether a save finished. saved means the new version is ready to use; failed means it did not work. version is the name it was saved with. workflow_version, once saved, is the label create_deployment takes.
get_workflow_setup_status
ChatGPTCheck one Auto Setup submission: queued, running, ready or rejected. Setup usually takes 30-60 minutes, and 2-3 hours when it needs a different machine type. While it is queued or running the result says when to check again (check_again_in_minutes); do not check more often, and let the user carry on meanwhile: RunComfy emails them when it finishes. When it is ready it returns the workflow_id and workflow_version that create_deployment takes, and the workflow_version_id that starting a machine takes; when it is rejected, the reason.
import_asset_from_url
ChatGPTDownload a file from a link straight into My Assets, without passing it through the chat. Returns an import to check with get_asset_import_status.
list_datasets
ChatGPTList training datasets in the caller's account. Backs `GET /prod/v1/trainers/datasets. Use it to find the id (for upload/status/delete calls) and the name (for the folder_path in an AI Toolkit config). The listing carries no per-file detail — use get_dataset_status` for a dataset's files. Args: include_raw: Return each dataset's unabridged payload instead of the compact summary. Larger response.
list_deployments
ChatGPTList Serverless API deployments in the caller's account. Backs `GET /prod/v2/deployments`. Args: ids: Optional list of deployment IDs to filter to. include_payload: Include workflow_api_json, overrides, and object_info_url for each deployment. Larger response. include_readme: Include the deployment's README markdown.
list_library_models
ChatGPTBrowse RunComfy's built-in ComfyUI model library, which every machine has without downloading. Free, no machine needed.
list_library_workflows
ChatGPTSearch RunComfy's public ComfyUI workflows, the ones on runcomfy.com/comfyui-workflows. Free. Call get_library_workflow for what running one takes.
list_machines
ChatGPTList the user's ComfyUI machines that are starting, running or stopping, with when each stops.
list_model_categories
ChatGPTList the two exposed generation categories: text-to-image and image-to-video. This is a fixed capability summary, not discovery of additional operations.
list_models
ChatGPTList the two supported generation operations and their current pricing. Returns metadata for FLUX Schnell text-to-image and Seedance Pro Fast image-to-video, including each explicit tool name. Does not discover additional callable operations or return input schemas. The complete generation inputs are declared on the two generation tools themselves.
list_my_assets
ChatGPTList the user's own files in My Assets: ComfyUI inputs, outputs and models every machine of theirs can use.
list_my_generations
ChatGPTList the account's finished model generations, newest first, from the website and the API alike. Each entry has the request_id, model_id, the inputs it ran with and its output_urls. Those links work as image, video or audio inputs to a new generation, so a past result can be the reference for a new one. Files are kept for a limited time: an entry past that time is not listed. A generation still running has no entry yet: use get_model_request_status.
list_my_workflows
ChatGPTList the saved versions in My Workflows. workflow_version is the label create_deployment takes, and workflow_version_id is what starting a machine takes.
list_workflow_setups
ChatGPTList the account's ten most recent Auto Setup submissions and the price of the next one.
open_upload_window
ChatGPTOpen an upload window in the conversation so the user can send a file from their device to RunComfy. Use it when the user has a local file for a model input or a workflow: files attached to the chat do not reach RunComfy. The file goes to My Assets under input/, and you are told its path and a link that works as a model input. Wait for that; nothing is uploaded until the user picks a file. End your turn with one short sentence pointing to the window: more text pushes it out of the user's view.
prepare_asset_upload
ChatGPTFor the upload window: a link to put the file the user chose under a new name in input/.
request_machine_start
ChatGPTAsk to start a ComfyUI machine for one of the user's workflows or one from the public library. It does not start until the user approves it. For a workflow the user names by its page, title or number, call get_library_workflow first: it needs no copy in My Workflows. The user approves on a card shown in the conversation, or at approve_url in an AI app that cannot show cards. The result has the most the machine can cost (the user's own hourly price for the whole time limit). The request expires after 30 minutes. You cannot approve it. Where the card shows, end your turn with one short sentence pointing to it: more text pushes it out of the user's view. When the user has answered, check get_machine_start_request. Machines cost money while they run: choose the smallest type and time limit that fit the job. When the work is done, call stop_machine, or set stop_when_done on run_on_machine for the last prompt. If neither happens the machine runs to its time limit, even after the conversation has ended.
respond_machine_start
ChatGPTFor the approval card: the user's answer to a machine start request.
resume_training_job
ChatGPTResume a stopped training job from its latest checkpoint. Backs `POST /prod/v1/trainers/ai-toolkit/jobs/{job_id}/resume. Reuses the same job_id rather than creating a new job, and restarts from the highest-step checkpoint (from step 0 if none exists). Useful after a preemption; for a FAILED job, read error from the status first and fix the cause — often via edit_training_job` — before resuming.
save_machine_workflow
ChatGPTSave a running machine as a new version in My Workflows, with its nodes and models. Saving takes a while, and the machine pauses during it. Check get_workflow_save_status every few minutes. Stop the machine only after it says the save is done: stopping earlier loses the save.
start_workflow_setup
ChatGPTHave RunComfy set up a ComfyUI workflow: a machine installs its custom nodes and models and saves it as a new version in My Workflows. This costs money. Call it first without confirmed_price_usd. That only returns the quote: the price ($0.20 for the account's first setup, $2.99 after that), the balance, and whether a pay-as-you-go account has room for another workflow. Tell the user the price, and that setup usually takes 30-60 minutes, or 2-3 hours when the workflow needs a different machine type. Only after they agree, call again with confirmed_price_usd set to the quoted price. Call it once per confirmation; if a call ends without an answer, check list_workflow_setups before trying again. After submitting, do not poll: check get_workflow_setup_status when it says to. RunComfy also emails the user when the setup finishes. When it is ready, create_deployment with its workflow_id and workflow_version publishes it as a Serverless API.
stop_machine
ChatGPTStop a running ComfyUI machine now, which ends its billing. Unsaved changes on the machine are lost; the saved workflow stays in My Workflows.
submit_request
ChatGPTSubmit an async inference request to a deployment. Backs `POST /prod/v1/deployments/{deployment_id}/inference. Args: deployment_id: Target deployment. overrides: Partial graph keyed by node_id, e.g. {"6": {"inputs": {"text": "a cat"}}}. Use get_deployment(include_payload=true) to discover node IDs and input names. workflow_api_json: Advanced — run a different workflow without updating the deployment. Omit overrides in this mode. webhook_url: Push-based updates instead of polling. webhook_intermediate_status: Fire webhooks on every status change, not just terminal. wait_for_completion: If true, poll until done and return the result inline. timeout_seconds: Max wait when wait_for_completion=true. File inputs: pass a public HTTPS URL or Base64 data URI directly in the overrides value, e.g. {"189": {"inputs": {"image": "https://example.com/photo.jpg"}}} or {"189": {"inputs": {"image": "data:image/jpeg;base64,/9j..."}}}`.
submit_training_job
ChatGPTSubmit an AI Toolkit training job (typically LoRA training). Backs `POST /prod/v1/trainers/ai-toolkit/jobs. The job mounts a READY dataset and runs the config you supply. Training runs for hours — this returns as soon as the job is queued; track it with get_training_job_status and pull artifacts with get_training_job_result. Args: config_file: The complete AI Toolkit YAML config as a string. Two paths in it are fixed by the platform: training_folder must be /app/ai-toolkit/output, and the dataset's folder_path must be /app/ai-toolkit/datasets/{dataset_name} where dataset_name is the dataset's name (not its id). gpu_type: ADA_80_PLUS (H100) or HOPPER_141 (H200). gpu_count: 1 for single-GPU (default), or 8 for multi-GPU. Multi-GPU is only supported on ADA_80_PLUS. gpu_id: Optional specific GPU selector, e.g. "#1"`.
update_deployment
ChatGPTPartially update a deployment. Backs `PATCH /prod/v2/deployments/{deployment_id}`. Only pass the fields you want to change. Set is_enabled=false to pause; true to re-enable.
upload_dataset_file_from_url
ChatGPTAdd one file to a dataset by fetching it from a public URL. Downloads `source_url and forwards the bytes to POST /prod/v1/trainers/datasets/{dataset_id}/upload. Args: dataset_id: Target dataset. source_url: Publicly reachable HTTPS URL for an image, video, or caption .txt file. Must be under 150 MB. filename: Name to store it under. Defaults to the URL's basename. For LoRA training each image/video needs a caption .txt with the *same base name* — img_0001.jpg pairs with img_0001.txt. Re-uploading the same filename overwrites the previous copy. For files the caller holds locally, or anything over 150 MB, use get_dataset_upload_urls` and PUT the bytes directly instead.
upload_dataset_text_file
ChatGPTAdd a text file — normally a caption — to a dataset. Backs `POST /prod/v1/trainers/datasets/{dataset_id}/upload with inline text, so captions can be written without hosting a file. Args: dataset_id: Target dataset. filename: Must share the base name of the media it captions: img_0001.jpg → img_0001.txt`. text: Caption body.
whoami
ChatGPTShow which RunComfy account this connection is signed in to: its email, username and account id. Use it when the user asks which account they are connected as, or when a balance, file, workflow or job they expect is missing: an AI app can hold connections to more than one RunComfy account.
get_balance
Claudeget_deployment
Claudeget_model
Claudeget_request_status
Claudelist_deployments
Claudelist_model_categories
Claudelist_models
Claude