append_explainer_chunk
ChatGPTAppend a chunk of explainer markup to an open explainer stream. Chunks accumulate into one growing document — never resend earlier content. Viewers see it render live. Returns remaining budget. Your first successful append returns the explainer's link, so make it a real playback buffer — an opening section, not just a title slide. Send the link to the user as a non-blocking progress update, then continue appending immediately.
attach_explainer_image
ChatGPTAttach a picture the user supplied or you generated in the conversation to an open explainer stream, so an ordinary image slide can show it. You need a file reference with a download link. Returns the attachment id to write on an image item: <item type="image" attachment="<id>" anchor="..." title="..."></item>, with no prompt. Call it before the append_explainer_chunk that names the slide — an append naming an id that is not an upload of this explainer is refused. Up to 12 pictures per explainer (each reply says how many more it can take), 8 MB each, png/jpeg/webp/gif; they do not count toward the generated-image cap. If the file only comes as a bare reference with no download link (uploads from the ChatGPT mobile app do), the call fails: build the explainer without that picture, never invent an id.
create_guide
ChatGPTStart generating a narrated Scrimba guide from your briefing and get back a URL for the user. Use this whenever the user wants — or would clearly benefit from — a guide that walks through, explains, tours, or breaks down something: code, a diff, a file, an architecture, a concept, a bug, anything. Scrimba authors a spoken, narrated guide and returns a URL where it builds live. You are the Teacher briefing a Teaching Assistant. The TA wraps and narrates the material with its own teaching judgment and may add context, but it only knows what you give it and will not fetch anything afterwards — there is no follow-up data exchange. So the briefing must be complete and directive: - Include ALL material the guide needs, inline (code, diffs, file contents, errors, data — the prompt is markdown). - State exactly what must be taught: what to explain, what to focus on, which connections to make ("make sure they understand X causes Y"), what misconceptions to head off. - Describe the structure you want, and what the learner already knows, so the TA pitches it right. Everything you direct will be covered; anything you omit will be missing from the guide. All of the above is internal guidance for composing the call — don't narrate it to the user (no explaining briefings, TAs, or why you're including material). Just tell them you're creating their Scrimba guide. IMPORTANT: As soon as this returns, give the user the URL. They must be signed in to Scrimba when opening it — the first signed-in user to open the link becomes the owner.
finish_explainer_stream
ChatGPTFinalize an open explainer stream. Call this when you are done pushing — the explainer stops showing as generating and narration becomes playable for viewers.
finish_guide_stream
ChatGPTFinalize an open guide stream. Call this when you are done pushing — the guide stops showing as generating and narration becomes playable for viewers.
push_guide_chunk
ChatGPTAppend a chunk of OPML to an open guide stream. Chunks accumulate into one growing document — never resend earlier content. Viewers see it render live. Returns remaining budget.
start_explainer_stream
ChatGPTOpen an explainer stream where YOU author the content. Scrimba creates an empty explainer and renders whatever explainer markup you push — live, as you push it. Use this for custom narration, mermaid diagrams, LaTeX maths, precisely composed code walkthroughs, and side-by-side before/after diffs of changed code. Returns: a stream_token for pushing and the complete explainer markup contract — no link yet. The explainer's link arrives in your first successful append_explainer_chunk result; make that first append a real playback buffer (an opening section, not just a title slide), then send the link to the user as a non-blocking progress update and keep appending (anyone who opens it can watch, no account needed, and holders of the link can claim ownership from the page). The result is large — more than 50KB — so if your client stores or truncates large tool results, use any storage or file tools you have to extract it in full. Read the contract to the end before pushing.
start_guide_stream
ChatGPTOpen a guide stream where YOU author the content. Scrimba creates an empty guide, returns a link plus the OPML contract, and renders whatever OPML you push — live, as you push it. Use this instead of create_guide when you want full control over the guide's content and structure (custom narration, mermaid diagrams, code walkthroughs composed by you). Returns: the guide url (anyone with the link can view), a claim_url (give this one to the user — opening it signed in makes them the owner), a stream_token for pushing, and the complete OPML authoring contract. Read the contract carefully before pushing.