diagnose_automation_run
ChatGPTExplain one run from its immutable workflow version, captured runtime inputs and asset snapshot, exact failed node attempts, error codes, outputs, and events. Returns targeted repair guidance and never retries automatically.
download_canvas_node_output
ChatGPTGet a temporary download URL for the full original image or video from a generator node, matching its Canvas Download button. No browser login is needed. Omit output_index for the currently selected result, or use a 1-based index into generatedOutputs from get_canvas. Download the returned URL as a file; it is not a preview. Links expire within 10 minutes and may be invalidated by reconnecting. Does not generate media, spend credits, or change the canvas.
export_automation_workflow
ChatGPTExport a draft or published workflow as an integrity-checked credential-free package. Instance credentials and deployment bindings are never embedded.
export_canvas_document
ChatGPTExport one approved canvas as a credential-free portable .scenelith.json document. Instance asset IDs, media URLs and generated outputs are deliberately removed.
get_automation_capabilities
ChatGPTRead the canonical versioned node catalog: exact inputs, outputs, settings, defaults, RUN INPUTS sidebar bindings, prompt variables, dynamic ports, help, triggers, edge roles, and limits. Pass canvas_id to include the live model/ratio/resolution catalogue.
get_automation_run
ChatGPTGet one automation run with captured inputs, node outputs, events, costs, warnings, and exact immutable workflow version.
get_automation_workflow
ChatGPTGet a workflow, its draft and published immutable versions, graph validation, and system-model issues.
get_canvas
ChatGPTRead a complete canvas once to discover nodes and readSet. Reuse its readSet as read_set for subsequent writes; unrelated edits and layout changes do not require another full read. Refresh only affected nodes with get_canvas_nodes.
get_canvas_capabilities
ChatGPTRead the exact node types, generation and Assistant models, model settings, semantic input ports, reference roles, and supported agent operations for one canvas. Call this before creating or configuring nodes.
get_canvas_nodes
ChatGPTRead only requested nodes, their incoming sources and incident connections, current scene revisions and a scoped readSet. Missing nodes are reported explicitly. Use after a scoped conflict or when those nodes may have changed; use the returned readSet token for these nodes; retain previous tokens for untouched nodes. No whole-canvas reread is needed.
inspect_canvas_node_inputs
ChatGPTResolve one node's exact connected Assistant text and concrete image, video or audio asset references, including identity cards and attached references. Pass clip_id for one Video Master scene. Only explicit generation inputs are returned; ORIGINAL is timeline context, not an automatic model input. Reports video segments that still need materialization.
inspect_identity_reference
ChatGPTReturn a bounded image preview plus metadata for one Character, Before, or After reference returned by list_identities. Identity references are workspace-level and do not expose their original Library source when that canvas was not approved.
inspect_library_asset
ChatGPTReturn a bounded visual preview plus exact metadata for one approved Library asset. Images return an image thumbnail; videos return a representative frame. Use this before selecting visual references.
list_automation_deployment_bindings
ChatGPTList credential and child-workflow slots connected to one workflow. Without the separately approved automation:credentials scope, credential names and IDs stay redacted while connection status remains visible.
list_automation_fixtures
ChatGPTList pinned test fixtures for single-node previews, including immutable version, runtime inputs, and captured per-node inputs.
list_automation_runs
ChatGPTList recent top-level automation runs for a canvas, optionally filtered to one workflow.
list_automation_trigger_deliveries
ChatGPTInspect automatic trigger delivery attempts, immutable version and input snapshots, retry state, errors, resulting run IDs, and open dead-letter alerts for one approved canvas.
list_automation_triggers
ChatGPTList paused or active schedules, webhooks, and Canvas-event triggers for one workflow, including pinned version, fixed run inputs, overlap policy, concurrency, and next/last fire times. Webhook secrets are never re-exposed.
list_automation_versions
ChatGPTList immutable draft, published, superseded, restored, and named-checkpoint version metadata for one workflow.
list_automation_workflows
ChatGPTList workflows available to a canvas, including system workflows and user workflows. Use get_automation_workflow for the graph and current draft version ID.
list_canvases
ChatGPTList accessible canvases with IDs, current revisions, status, and compact graph summaries. Optionally limit the result to one workspace.
list_identities
ChatGPTList reusable Scenelith identities with explicit single or before_after type, separated Character/Before/After groups, copied identity asset IDs, and accessible Library source lineage.
list_library_assets
ChatGPTList one page of generated or uploaded images and videos from the canvases approved for this connection. Returns stable asset IDs, media metadata, counts, and next_cursor. Browser URLs require a signed-in Scenelith session; call inspect_library_asset when the agent needs to see the media.
list_workspaces
ChatGPTList the Scenelith workspaces this connection can access. Use the returned workspace IDs for canvas, library, and identity tools.
studio_capabilities
ChatGPTLive Library, Image GEN, Video GEN, Edit, character and Reviral capabilities: exact UI schemas, all current models, reference roles, tools and section links.
studio_download
ChatGPTGet a short-lived authenticated download for a saved Studio asset. Revocation and live workspace access are checked at download time.
studio_generate
ChatGPTStart a durable Studio task using the exact quoted snapshot. Requires explicit generation scope; uses credits. A repeated UUID key cannot duplicate a task or charge.
studio_generator_draft
ChatGPTRead the current Image GEN or Video GEN draft, revision and saved history. Does not save settings or generate media.
studio_history
ChatGPTRecent tasks from a Studio section. Use surface image-gen/video-gen for standalone generators, or kind character/video for legacy character/Reviral tasks.
studio_image_edit
ChatGPTCreate a separate edited version of an image after approval of studio_quote_image_edit. Requires expectedCredits from that quote. Uses credits and the same Edit pipeline as the viewer; preserves the original and its character membership. Retry identical input with the same requestKey.
studio_inspect
ChatGPTInspect actual pixels of an authorized image or a selected video frame. Video inspection saves a frame asset to the workspace; image inspection only returns pixels. Catalog images and authorized workspace media only.
studio_library
ChatGPTBrowse saved images and videos across Library, with dimensions, character membership and pagination. Canvas media is included only with a separate Library/Canvas grant. Assigned media needs no Add to identity.
studio_media_job
ChatGPTRead a saved media preparation operation. When completed, repeat the original request with the same key. Does not submit a paid generation.
studio_quote
ChatGPTValidate and quote one media generation request. May prepare cached media analysis and return a media job. Read studio_media_job, then repeat the same quote. Never starts paid generation or spends credits.
studio_quote_image_edit
ChatGPTValidate and quote a separate edited image. May save reference copies; never starts generation or spends credits. Original is preserved. For a character-owned image, character read permission is also required.
studio_run
ChatGPTRead saved Studio, Image GEN, Video GEN or character task status, source/settings snapshot and its correct screen link.
validate_automation_connection
ChatGPTValidate one proposed typed connection against the current graph before saving it, including exact ports, compatibility, edge role, duplicate inputs, disabled nodes, cycles, terminal nodes, error routes, and bounded retry rules.
validate_automation_workflow
ChatGPTValidate a complete draft without saving it. Returns every structural, node-setting, prompt-variable, port, route, retry, secret, and terminal-path issue plus ordered nodes and the exact run-input contract when parseable.
