Article Workflow

Articles flow through a state machine from initial generation to publication. The admin dashboard provides controls at each stage.

State Machine

                +-------+
                | draft |  (initial AI generation)
                +---+---+
                    |
              admin reviews
                    |
         +----------+----------+
         |          |          |
    +----v---+ +---v----+ +---v-----------+
    |approved| |rejected| |revision_requested|
    +----+---+ +--------+ +---+-----------+
         |                     |
    published             AI revises
    to site                    |
         |               +-----v----+
    +----v------+        |  pending  |
    |unpublished|        +-----+-----+
    +-----------+              |
                          admin reviews
                               |
                    +----------+----------+
                    |          |          |
               approved   rejected   revision_requested

Status Definitions

Status	Description
`draft`	Newly generated article, awaiting first admin review
`pending`	Revised article awaiting re-review
`approved`	Published and visible on the public site
`rejected`	Discarded by admin, not shown publicly
`revision_requested`	Admin has provided feedback; AI will revise
`unpublished`	Previously approved article taken down

Article Types

The pipeline auto-detects article type from source content, or it can be set manually:

Type	Description	Detection
`summary`	Standard news article covering a single topic	Default when no deal keywords found
`deal`	Time-limited offer (bonus, sale, discount)	Triggered by 2+ deal keywords across sources
`roundup`	Multiple related stories grouped together	Set manually or by clustering logic

Two-Column Layout

Published articles display in a two-column layout on the article page:

Main column – article content (HTML)
Sidebar – key details (label/value pairs), relevant links, Scout’s Take, and source attribution

Deduplication

Content deduplication happens at two levels:

Raw article level – each scraped article has a SHA-256 content hash stored in the hash column. Duplicate hashes are rejected at insert time.
Cluster level – the clustering engine groups raw articles about the same topic. The pipeline requires at least 2 unique source sites per cluster to reduce hallucination risk and ensure multi-source verification.

Voting System

Published articles include an upvote/downvote widget. Votes are:

Stored with an IP hash (not the raw IP) for privacy
One vote per IP per article
Monitored for anomalies using configurable thresholds:

Config	Default	Purpose
`votes.alert_threshold`	10	Minimum votes before alerting
`votes.alert_ratio`	0.4	Downvote ratio that triggers alert
`votes.alert_window`	24	Window in hours for ratio calculation

When the downvote ratio exceeds the threshold, the system flags the article for admin review.

Reference Pages