Example: a deck from your notes

You have rough notes from a meeting and you want a presentation out of them, without leaving your AI client. This is the whole run, call by call.

You need: an AI client connected to Papera.

The ids below are shortened for readability. Real ones are UUIDs, and you never type them yourself: each tool returns the ids the next one needs.

1. Hand over the notes

You: Here are my notes from the pricing review. Turn them into a short deck for the leadership meeting, keep the tone professional. (notes pasted below)

The client starts the run with your text as the source:

// generate_outline
{
  "topic": "Pricing review summary for the leadership meeting",
  "source_text": "Notes: three tiers today ... churn concentrated in the low tier ...",
  "mood": "professional"
}

Because source_text is there, Papera builds the plan from your notes instead of researching the topic from scratch. It comes back with an outline.

2. Shape the outline

Papera: Here is the outline: 1. Where pricing stands. 2. What the data shows. 3. Three options. 4. What we recommend. Approve it?

You want one change:

// approve_outline
{ "decision": "Good, but drop section 3 and fold the options into the recommendation." }

Papera re-plans and shows you the shorter outline. This time you accept:

// approve_outline
{ "decision": "approved" }

Approving writes the content, which comes back for review.

3. Approve the content

Papera: Here is the written content, four sections. Approve it?

// approve_content
{ "decision": "approved" }

As with the outline, you can send feedback instead ("cut the second paragraph of section 2") as many times as you like. Nothing is saved until you approve.

4. Save it

// save_project
{}

Papera saves the content as a project and returns the link, plus the project and workflow ids:

{
  "project_id": "9c1f2a84",
  "workflow_id": "41ac7d20",
  "link": "https://papera.dev/project/9c1f2a84"
}

Open that link, signed in to the same account, and the content is there.

5. Render the deck

The project holds the content; a presentation is one thing you can render from it. Your client asks for the content node and attaches the slides renderer:

// get_workflow
{ "project_id": "9c1f2a84", "workflow_id": "41ac7d20" }

// attach_renderer
{
  "project_id": "9c1f2a84",
  "workflow_id": "41ac7d20",
  "source_node_id": "content-3b71",
  "renderer_id": "slides"
}

attach_renderer returns the rendered deck, so your client can tell you how many slides came out and what is on them. Open the project link to present it, export it, or keep editing.

Variations

Use your own model. Add "use_my_model": true to the generate_outline call. Papera then hands your client a prompt at each step, your client's model runs it, and submit_agent_action returns the result. The run costs no AI credits. See Create content.

No notes, just a topic. Drop source_text and pass topic alone. Papera researches it first.

Add it to an existing project. Pass "project_id": "9c1f2a84" to save_project and the content becomes a new workflow in that project instead of a project of its own. This is also the way out when you are at your project limit.

Another output. Swap renderer_id for document, blog or social-post, or attach several: they all render from the same content.