Database Models

The database is PostgreSQL (Neon recommended). All models are defined in prisma/schema.prisma.

Verdict

One record per agent per debate run.

Field	Type	Description
`id`	UUID	Primary key
`persona`	string	Agent key: `bull`, `bear`, `quant`, `macro_regime`, `narrative`, `contrarian`
`token`	string	Token symbol (uppercase)
`timestamp`	DateTime	When this verdict was produced
`market_regime_context`	string?	Macro regime at analysis time
`stance`	string	`long`, `short`, or `neutral`
`conviction`	float	0–100
`position_direction`	string	`long`, `short`, or `none`
`position_size_pct`	float	% of portfolio
`entry_ref`	float	Reference entry price
`invalidation`	float	Invalidation price level
`target`	float?	Target price
`reasoning_trail`	string	JSON array of reasoning steps
`summary`	string	Plain-language summary
`consensusId`	string?	FK → `Consensus.id`

Aggregated council result — one per debate run.

Field	Type	Description
`id`	UUID	Primary key
`token`	string	Token symbol
`timestamp`	DateTime	When the consensus was generated
`conviction_score`	float	Weighted average conviction
`agreement`	string	`unanimous`, `majority`, or `split`
`majority_stance`	string	`long`, `short`, or `neutral`
`outlier_persona`	string?	The dissenting agent (if any)
`outlier_reason`	string?	Why the outlier dissented
`final_call`	string	`long`, `short`, or `neutral`

Relations: verdicts[] (Verdict), divergenceEvents[] (DivergenceEvent)

Detected shift between two consecutive consensus snapshots.

Field	Type	Description
`id`	UUID	Primary key
`token`	string	Token symbol
`timestamp`	DateTime	When detected
`triggered`	boolean	Whether the shift crossed a notification threshold
`shift_type`	string	`convergence`, `divergence`, `flip`, or `none`
`previous_agreement`	string	Prior agreement level
`current_agreement`	string	New agreement level
`previous_call`	string	Prior final call
`current_call`	string	New final call
`note`	string	Human-readable description
`consensusId`	string	FK → `Consensus.id`

Running accuracy stats per agent per market regime.

Field	Type	Description
`id`	UUID	Primary key
`persona`	string	Agent key
`regime`	string	`risk_on`, `risk_off`, `neutral`, `unclassified`
`debates_scored`	int	Total resolved positions
`correct_calls`	int	Correctly resolved positions
`accuracy_pct`	float	`correct_calls / debates_scored × 100`
`avg_conviction_when_correct`	float	Mean conviction on winning calls
`avg_conviction_when_wrong`	float	Mean conviction on losing calls
`last_updated`	DateTime	Auto-updated

Unique constraint: (persona, regime) — one record per agent per regime.

User-configured alert with schedule and conditions.

Field	Type	Description
`id`	UUID	Primary key
`userId`	string	Clerk user ID
`token`	string	Token to monitor
`gapMinutes`	int	Check frequency: 15, 30, 60, 120, 240, 720, or 1440
`windowStartMin`	int?	Daily window start (minutes from UTC midnight)
`windowEndMin`	int?	Daily window end (minutes from UTC midnight)
`activeDaysMask`	int	Bitmask: bit 0=Mon … bit 6=Sun. 127 = all days
`conditions`	string	JSON array of `AlertCondition[]`
`conditionLogic`	string	`AND` or `OR`
`notifyInApp`	boolean	Record TriggerEvent on fire
`notifyEmail`	boolean	Send email on fire
`emailAddress`	string?	Resolved from Clerk at save time
`notifyMode`	string	`immediate` or `daily_summary`
`enabled`	boolean	Whether the trigger is active
`scheduleVersion`	int	Increments on every schedule change (used to cancel stale Inngest jobs)
`nextRunAt`	DateTime	When the next check is scheduled
`lockedUntil`	DateTime?	Optimistic lock — set during active job
`lastCheckedAt`	DateTime?	When the last check ran
`lastEmailSentAt`	DateTime?	For 1-hour email cooldown enforcement

One record per scheduled check run.

Field	Type	Description
`id`	UUID	Primary key
`triggerId`	string	FK → `AlertTrigger.id` (cascade delete)
`token`	string	Token that was checked
`checkedAt`	DateTime	When the check ran
`fired`	boolean	`true` if ≥1 condition was met
`consensusId`	string?	The Consensus produced by this check

One record per condition that fired within a check.

Field	Type	Description
`id`	UUID	Primary key
`triggerId`	string	FK → `AlertTrigger.id` (cascade delete)
`checkId`	string	The TriggerCheck this belongs to
`token`	string	Token symbol
`firedAt`	DateTime	When the condition fired
`conditionType`	string	`consensus_change`, `council_stance`, `agreement_below`, `conviction_threshold`, `agent_stance`, `divergence`, `market_regime`
`previousState`	string	JSON snapshot of prior state
`newState`	string	JSON snapshot of new state
`note`	string	Human-readable description

Legacy recurring token recheck schedule (per user/token, independent of AlertTrigger).

Unique constraint: (userId, token)