check_design_plan
ChatGPTChecks the content and layout of a MiriCanvas design plan by applying it to the template before saving a design. Takes the plan (the content and layout to apply) and returns check results (findings). Stores the plan on the server and returns a reference (planRef) for subsequent calls. Overlap and overflow only show up once the plan is applied. Call it before create_design for: - A plan that adds a chart or changes a chart's data, series, or type - A plan that adds a table or moves the elements around a table - A plan whose text exceeds a box's estimated capacity, or that changed wording or page layout because space ran short - A plan that changes the page structure substantially, or when the user wants to review before creation For other plans, create_design also returns the same check results without saving when there are errors. valid=true only means there are no errors; it does not mean tables and charts do not overlap. The design/plan-check guide section "plan만 검증하기" (get_miricanvas_guide) explains how to handle the results.
complete_image_upload
ChatGPTChecks whether uploaded images are ready to use in a MiriCanvas design. Call it after uploading files to the addresses from create_image_upload. Returns image reference information (resource) for each registered image. Put it as is into an image replacement entry (replace_image) in the design plan. If the registration status (status) is still processing (PROCESSING), call it again with the same upload identifier (uploadId). Checking can take up to 20 seconds. If the file upload (PUT) failed, do not retry it; call this tool with the error you received in uploadError, and it tells you the next method right away without waiting. The design/user-image guide (get_miricanvas_guide) explains alternatives for a blocked upload.
create_design
ChatGPTCreates a new MiriCanvas design by applying the plan (the content and layout to apply) to the selected template, and saves it to the team chosen when connecting. Call it once the plan is complete: the outline from prepare_design_plan, filled in using the template structure information (manifest). For plans that match a condition listed in check_design_plan, that tool checks the plan first without saving a design. Every call creates a new design. saved=true means the design was saved and the response includes the link to share. saved=false means nothing was saved and the check results (findings) list what to fix; items with severity error must be fixed before calling again. Instead of the whole plan, you may send the reference to the plan stored on the server (planRef) and a planPatch with the parts to change. The design/plan-check guide section "create_design 호출과 오류 처리" (get_miricanvas_guide) explains how to handle the results, and the design/plan-text section "색 바꾸기" covers recoloring.
create_image_upload
ChatGPTPrepares a place to upload the user's own images (PNG, JPEG, WebP, GIF, or SVG) to MiriCanvas so they can be used in a design. The images are saved to the user's MiriCanvas drive. Returns an upload address (uploadUrl) and headers for each file. Within 3 minutes, send each file's bytes to its address with an HTTP PUT, using exactly the given headers. Then pass the upload identifier (uploadId) to complete_image_upload to finish the registration. Do not send file contents to this tool. For an SVG graphic you drew yourself, use register_svg_image instead. The design/user-image guide (get_miricanvas_guide) explains the available ways to add images.
get_design_preview
ChatGPTGets preview images of the pages of a created design. Only the design result screen calls this tool; the model does not.
get_miricanvas_guide
ChatGPTGets a guide explaining how to do the requested MiriCanvas task and what to watch out for. The overview guide describes what users can do with MiriCanvas. The design/create-from-template entry guide covers template search and design creation, including the tool sequence and the guide for each stage of writing the design plan.
get_my_image_previews
ChatGPTShows the user's images registered in MiriCanvas as small preview images. Use it to see what an image shows before deciding where it goes in the design. Pass only image identifiers (keys) you received in this conversation, such as from an image registration result or a message from the image picker screen. Each image is preceded by a label with its image identifier (key) and width and height, and an image that cannot be shown comes with a reason instead of an image.
get_template_page_previews
ChatGPTShows template pages as individual images or a combined preview. Before selection, use type="grid" for up to 10 candidates. Each returns one image within 768x768 pixels of up to its first 20 pages in page order. Omit pageNoList. If the combined preview cannot be created, the first page is returned alone with a note. The size of each page within a preview does not represent its original size. After selection, use type="page" (the default) with pageNoList to inspect individual pages. Results follow template key order, then requested page order. Each image has a template key label; individual images also have a page number. Unavailable previews come with a reason instead of an image. The guide sections design/create-from-template "전체 작업 절차" and design/plan-structure "콘텐츠와 페이지 구성하기" (get_miricanvas_guide) explain when to call it and how to use the images.
get_template_structure
ChatGPTGets the page layout of the selected MiriCanvas template and its editable text, images, charts, and colors. This structure information (manifest) includes slots (the places where text goes) and groups of elements, as the basis for writing the plan (the content and layout to apply). The design/plan-structure guide (get_miricanvas_guide) explains how to write the plan using this information.
list_my_images
ChatGPTShows the images the user has already uploaded to their MiriCanvas drive root folder, most recently updated first, 30 at a time. Call it when the user wants to use images they uploaded before and the conversation cannot show an image picker screen, the user said the screen does not appear, or the user left the choice to you. Otherwise use open_image_picker. Each image comes with a number, file name, width and height, last updated date, and image reference information (resource). Preview images are included when available. Put the reference information of each chosen image as is into an image replacement entry (replace_image) in the design plan. The response includes guidance for choosing images.
open_image_picker
ChatGPTShows a screen in the conversation where the user browses thumbnails of images they already uploaded to their MiriCanvas drive and chooses some. Use it when the user wants images they uploaded before. The design/user-image guide (get_miricanvas_guide) describes alternatives when the conversation cannot show a screen. When the user chooses images, the screen sends a message with image reference information (resource) for each image. The design/user-image guide (get_miricanvas_guide) explains how to continue after image selection.
open_image_upload
ChatGPTShows a screen in the conversation where the user picks images from their device to upload to MiriCanvas. The images are saved to the user's MiriCanvas drive. Use it when you cannot upload the files yourself: there is no attached file object or image address and you have no local file path and shell, or the app blocked your direct upload. To choose images the user already uploaded, use open_image_picker instead. When the user picks images, the screen finishes the registration and sends a message with image reference information (resource) for each image. The design/user-image guide (get_miricanvas_guide) explains how to continue after image selection. The design/user-image guide (get_miricanvas_guide) explains the available ways to add images.
ping
ChatGPTChecks that MiriCanvas is connected and that the sign-in is valid. Returns pong on success.
prepare_design_plan
ChatGPTBuilds the base outline of a design plan, ready to be filled with content, from the selected template pages. The plan holds the content and layout to apply to the design. Returns a plan whose slots (the places where text goes) are filled with the template's original text, along with plan-writing guidance (hints) and per-slot writing guidance (slotHints). The design/plan-text guide (get_miricanvas_guide) explains how to fill the outline. Call it after choosing the pages to use from the template structure information (manifest) and the preview images from get_template_page_previews. The design/plan-structure guide section "콘텐츠와 페이지 구성하기" (get_miricanvas_guide) explains page composition. A slot whose original text you neither change nor mark to keep (keep) becomes blank when saved, and the save is held back (expose error). The design/plan-text guide section "골격의 hints와 slotHints 읽기" explains how to fill every slot.
register_attached_images
ChatGPTAdds images attached in the chat to MiriCanvas so they can be used in a design. The images are saved to the user's MiriCanvas drive. Supports PNG, JPEG, WebP, GIF, and SVG. Use it only when the host supplies a temporary download address (download_url) and file identifier (file_id) in the attachment list (images); never build these values yourself. Call it before choosing a template when the user attaches images to use in a design. Up to 10 images at a time. Returns image reference information (resource) for each added image. Put it as is into an image replacement entry (replace_image) in the design plan. The design/user-image guide (get_miricanvas_guide) explains the available ways to add images.
register_image_from_url
ChatGPTDownloads a public image (PNG, JPEG, WebP, GIF, or SVG) from the web and adds it to MiriCanvas so it can be used in a design. The image is saved to the user's MiriCanvas drive. Use it for an image address the user gives, or for a web image found for an image in the design. The design/web-image guide section "웹 이미지" (get_miricanvas_guide) explains how to check and add web candidates, the design/plan-image guide section "이미지 교체" explains when to use web images, and the design/user-image guide section "새 이미지 등록 방법" covers an image address the user gives. Pass the direct address of an image file. Web page addresses are not accepted. Returns image reference information (resource) for the added image. Put it as is into an image replacement entry (replace_image) in the design plan.
register_svg_image
ChatGPTAdds a graphic you drew as SVG, such as a diagram or flowchart, to MiriCanvas so it can be placed in the design. The graphic is saved to the user's MiriCanvas drive. Use it for structure, flow, relationships, order, or comparison that the template has no page to show, or when the user asks for a drawing. Examples: a system diagram, a work process, an org chart, a timeline, a comparison matrix. Do not use it for photos or realistic images, numeric charts (use the chart operations of the design plan), sentences (use text slots), or icons and decorations (use search elements). Pass the SVG markup as is. Give the root element positive width and height and a viewBox with the same ratio, with no external references or document type declaration (DTD), up to 1 MiB. Returns image reference information (resource) for the added graphic. Put it as is into an image replacement entry (replace_image) in the design plan. The design/plan-image guide section "직접 그릴 그림" (get_miricanvas_guide) explains how to draw and place graphics.
report_app_event
ChatGPTRecords a usage event that occurred in the MiriCanvas app UI. Only the app UI calls this tool; the model does not. The event is sent to MiriCanvas usage analytics (Google Analytics) together with the signed-in MiriCanvas account, the host product, a conversation identifier when the host provides one, and the language and country when available. Conversation content and user-written text are never sent.
search_design_elements
ChatGPTSearches for elements to use in a design, such as photos, illustrations, and icons. Finds new candidates for photos[], the list of replaceable images in the template structure information (manifest). Call it after viewing the preview images of the original pages. The design/plan-image guide section "이미지 교체" (get_miricanvas_guide) explains search targets and methods, re-searching, and choosing candidates. Which targets to search, and when, follows the per-type strategy and transition conditions in the design/plan-image guide section "일반 교체 이미지". Up to 5 targets that share the same search method (searchMethod) can be searched in one call. Put the replacement element data (resource) from the response into the image replacement entry (replace_image) of the design plan as is.
search_templates
ChatGPTFinds MiriCanvas template candidates that match the user's request. Returns up to 30 candidates by default. structuredContent.templateKeys lists the identifiers; templates contains each templateKey, pageCount, and curation categories (curatedInfo): HIGH_QUALITY for high quality, REPORT for reports, or an empty array when neither applies. The design/create-from-template guide section "템플릿 검색과 선택" (get_miricanvas_guide) explains template types and search keywords. When the first request explicitly supplies a template identifier (templateKey), keywords=["key:<templateKey>"] looks up that template without templateType or language. A matching key is already selected; an absent key means the requested template was not found.
show_template_choices
ChatGPTShows previews of the specified templates for the user to choose from. Pass identifiers (templateKey) selected as described in the design/create-from-template guide section "후보 썸네일 비교", in the order to display them. Templates that are no longer available are omitted without replacement.
