# Scraped topic drafts

In Admin → The journal, enter a topic and choose **Submit & build draft**.
Python researches the web, combines scraped facts, and saves an editable research draft.
Use **Review & edit article** to edit, upload a cover if desired, and publish.
Source URLs stay in the database and in the authenticated editor activity view as plain text.
Public article content and navigation do not link to source websites.

## Setup

No Ollama, OpenAI key, or paid API is required.

Research searches CarWale, BikeWale, RushLane and IndiaCarNews for the English topic,
checks robots.txt and extracts relevant article text. It requires at least two distinct
source domains. Blocked or irrelevant pages are skipped. If research is insufficient,
the job fails without creating a post. Search results and website availability can change.
RSS imports remain separate from topic writing. Internet access is needed for research.
Review drafts for factual accuracy and rewrite them in your own editorial voice before publishing.

Generation runs in a background thread with one article at a time. Reloading the page does not
submit the topic again. A server restart interrupts an active generation; after 15 minutes
the stale job becomes failed, and the topic can be resubmitted. No article is saved on failed
research or invalid output. Completed drafts are not automatically published.

Tests mock research responses. Live drafting needs internet access to approved source websites.