add_reference
ChatGPTTurn an image, video or audio file into a reference id that estimate_cost and generate accept in reference_ids. Give exactly one of: chatgpt_file (a file the user attached in ChatGPT), url (a direct https link; use it whenever the file is already public, instead of uploading), upload_id (after sending a file with start_upload), asset_id (one of this brand's outputs) or library_id (a file the brand already uploaded to Predis). A new file is saved to the brand's Predis assets.
check_run
ChatGPTGet a run's status and, once it has finished, its output links and what it cost.
choose_from_uploads
ChatGPTShow the user the files already uploaded to this brand in Predis (images, videos, audio, and the brand logo) in a picker inside the chat, so they can pick references instead of uploading again. Use it when the user says the file is already in Predis, or before asking them to upload one. When they pick, the reference ids reach your context: ask them to send a message once they have chosen, then pass the ids in reference_ids. If no picker appears, call list_uploads and ask the user which file to use.
describe_model
ChatGPTShow one model's options (with allowed values and defaults), how it uses reference media, and what it needs. Use the option names as keys in config.
estimate_cost
ChatGPTPrice a generation without starting it, check its references suit the model, and report the credit balance. Nothing is charged. Call this before any video or multi-output generation.
generate
ChatGPTStart an image or video generation in Predis Model Studio. Credits for the quote are held now and the unused part is returned when the run ends. Returns straight away with a run id; the result says whether the app's result card will show the outcome by itself or you should wait with wait_for_run.
get_account
ChatGPTWhether Model Studio can create for this connection right now and, if not, the sentence to tell the user and what they can do; the plan and the credit balance; and who is signed in. Call it when the user asks about their Predis account or connection, or before explaining why a generation was refused.
list_models
ChatGPTList the image and video models available in Predis Model Studio, with their typical cost once the brand is on a paid plan. Call describe_model for a model's options before estimating or generating.
list_runs
ChatGPTList this brand's recent generations, newest first, including ones made in the Predis app.
list_uploads
ChatGPTOne page of the files already uploaded to this brand in Predis, newest first, with the brand logo on the first page. Pass a file's library_id to add_reference. The uploads picker (choose_from_uploads) calls this itself.
run_status
ChatGPTA run's status and, once it has finished, its outputs. The result card calls this itself while a run is going; assistants use check_run.
start_upload
ChatGPTGet a one-time link for sending a file the user attached, or one you have locally, to Predis. Not needed for a file that is already at a public https link: pass that to add_reference(url=...); nor for one already uploaded to Predis: call choose_from_uploads. PUT the file's bytes to upload_url, for example with the curl command returned; the response holds the reference id. Apps that show tool results inline also show the user an upload box, so if you cannot reach the link, ask them to choose the file there. Otherwise give them page_url. Then call add_reference with the upload_id.
wait_for_run
ChatGPTWait until a run finishes, up to max_seconds, then report it as check_run does. Call it again if the run is still going. Images usually take under a minute; videos take a few minutes.
