ADR: CI Data Model v1

Date: 2026-06-26 Status: Accepted

Context

The current ledger stores sources, snapshots, signals, and synthesis_runs. That is enough to run daily and weekly reports, but not enough to power a dashboard, source repair queue, report archive, delivery audit, or stakeholder-safe action workflow.

Decision

Extend the CI ledger with four first-class concepts: source health events, report index, action items, and bot deliveries.

Schema

source_health_events

Tracks health history rather than only latest snapshot state.

Fields:

• id
• source_url
• status
• failure_reason
• http_status
• last_success_at
• last_failure_at
• failure_streak
• replacement_recommendation
• created_at

report_index

Dashboard archive index over generated Markdown, HTML, PDF, and delivery-ready summaries.

Fields:

• id
• cadence
• date_start
• date_end
• markdown_path
• html_path
• pdf_path
• quality_score
• top_signal_ids
• top_action_owners
• source_health_summary
• status
• created_at

action_items

Turns recommendations into trackable work.

Fields:

• id
• report_id
• title
• owner
• recommendation
• evidence_signal_ids
• status
• priority
• due_date
• athena_review_status
• created_at
• updated_at

Statuses: proposed, accepted, assigned, completed, rejected.

bot_deliveries

Separates report generation from delivery.

Fields:

• id
• report_id
• bot_profile
• channel
• recipient
• status
• delivered_at
• error
• created_at

Rationale

The dashboard should be built on operational records, not filesystem scraping alone. Filesystem archive is useful, but it cannot answer "what failed?", "what was delivered?", "who owns the action?", or "is this stakeholder-safe?" without a data model.

Consequences

• CI v1 can index historical daily and weekly reports.
• CI bot delivery can be audited separately from Athena oversight.
• Source health becomes a product surface, not just a log line.
• Stakeholder access remains blocked until report and claim quality are explicit.