Style Profiles

Style Profiles are your brand's writing-style fingerprint, stored as structured traits (tone, formality, sentence length, vocabulary, structure). There are two ways to create a profile: paste text or upload a file directly, or point the system at a domain that it crawls and derives the style from. A finished profile is selected during article generation and gets folded into the writing prompt as an instruction — so every article stays consistent in tone, no matter who or what generated it.

What it does

• Profile from text/file — paste text or upload TXT/Markdown/DOCX/PDF; the AI analyzes tone, formality, sentence length, vocabulary, and structure.
• Profile from a website — give a domain, the system crawls the sitemap (with a homepage crawl as fallback) and analyzes several content-rich subpages.
• Review before activation — profiles learned from a URL get a review step before you actually use them.
• Default profile — mark one profile per team as the default (in the UI under /style-profiles).
• Reused across modules — the same profile is selectable for single articles, the article editor, Storylines, and Autopilot.

When to use it

• Multiple authors or AI-generated articles need to read as if they came from the same person.
• You want to adopt an existing website's style without describing it by hand.
• A client or brand has a clearly recognizable tone that has to be held across many articles.
• You're building an automated workflow (Autopilot, bulk generation) where nobody manually checks the style of every article.

Workflow

1. See existing profiles — GET /projects/{project}/style-profiles lists all profiles for the project.
2. Create from text/file — currently runs through the UI at /style-profiles: paste text or upload a file, analysis runs asynchronously in the background (10 credits), then status: ready.
3. Create from a URL — POST /projects/{project}/style-profiles/from-url with {source_url, name?}. It checks upfront whether enough credits are available (otherwise 402); with enough credits the analysis starts asynchronously, response 202 with {id, status: 'pending', source_url, message}.
4. Poll status — GET /style-profiles/{id}/scrape-status returns status, scraped_pages, failure_reason, and pending_review for polling.
5. Review the result — once the URL analysis finishes (pending_review: true), either POST /style-profiles/{id}/accept (keep it) or POST /style-profiles/{id}/discard (discard it — only works while pending_review is still true).
6. Create a profile directly — POST /projects/{project}/style-profiles creates a profile synchronously with style data you already have ({name, description?, style_data?}), e.g. for your own integration that computed the style itself. No credit charge.
7. Use it — when starting an article (single article, article editor, Storylines, Autopilot), pick a profile with status: ready; the style is folded into the writing prompt automatically.

API

Example: starting a URL analysis:

POST /v1/projects/42/style-profiles/from-url
{"source_url": "https://example.com", "name": "Style from example.com"}

Credits & limits

• Learning from a URL costs 25 credits — charged only after the job completes successfully, not at the start. If the analysis fails before enough pages (at least 5) were successfully analyzed, 0 credits are charged.
• Learning from text/file (through the UI) costs 10 credits.
• Creating directly via the API with style data you already have, plus listing, updating, deleting, and accept/discard, are all credit-free.
• The URL analysis runs asynchronously and can take 2–4 minutes; track progress by polling scrape-status.

Related modules

• Storylines — storyline articles can be generated with the same style profile used for single articles.
• Automation — Autopilot also uses a stored style profile when generating articles automatically.
• AI Content Editor — the article editor lets you pick a style profile for rewrites.