Filament Jev Moderation Laravel Package

Product Decisions This Supports

• Build vs. buy for moderation operations: v0.1.0 provides a ready-made Filament moderation console—review queue, decision history, author history, calibration, volume, and overview metrics—so the team can avoid building internal triage UI before validating the moderation workflow.
• Human-in-the-loop launch strategy: Supports a roadmap where AI flags content, moderators approve/reject/reclassify with notes, and ReviewResolved fires through the core. This makes it viable to launch assisted moderation rather than fully automated enforcement.
• Queue prioritization and staffing: The unified review queue spans moderated models, shows confidence and time waiting, sorts most confident first and then longest waiting, and supports filters by model type, category, confidence, source, and date. The overview widget adds waiting count, oldest wait, automatic blocks since midnight, median time to review, and override rate.
• Threshold and policy tuning: The read-only thresholds page exposes each policy’s review_above, block_above, uncertain_below, version, affected models, and calibration bands. The calibration widget highlights bands with high override rates and suggests threshold changes, but changes still require config updates and targeted jev-moderation:rescan runs—so this supports a PM/engineering decision loop, not autonomous tuning.
• Moderation quality and trust reporting: Decision history includes every Jev result, report, moderator review, rescan, and failed check, with moderator/reporter, notes, Jev version, policy version, confidence, category probabilities, and stored content copies. Author history aggregates decisions across moderated types for repeat-offender and user-risk review.
• Category correction and feedback quality: Reclassify lets moderators remove content under a corrected harmful category via the core’s reject(), improving the quality of moderation outcomes when Jev selects the wrong category.
• Access control and role design: authorize() gates pages and widgets; moderation actions can be checked through authorizeActionsUsing() or a ModerationDecision policy. This supports decisions about who can view moderation data versus who can approve, reject, or reclassify.
• Operational capacity planning: The volume chart shows decisions per day by status over 7/30/90 days, supporting reviewer capacity planning and trend reviews.
• Scoping and rollout choices: Plugin options support limiting to selected moderated models, labeling model types, including automatic blocks in the queue, configuring author relations, naming the navigation group, and setting override warning thresholds/sample sizes.

When to Consider This Package

Consider adopting this package when:

• The product runs on Laravel and moderation operations live in or can live in a Filament admin panel.
• The team is using or planning to use phoenix1331/laravel-jev-moderation 0.2 for AI-assisted content classification.
• There is a real human review workflow across one or more Eloquent models, and the team needs approve/reject/reclassify actions with notes and an audit trail.
• Product and trust-and-safety stakeholders need visibility into queue health, override rates, decision history, author history, and calibration bands without building a separate analytics or review tool.
• Moderation is small-to-mid scale and can be operated from an internal admin panel rather than a dedicated moderation platform.
• The team can accept an early-stage v0.1.0 package and validate maturity, upgrade path, and dependency stability before broad rollout.
• Access control can be expressed through Filament policies, authorizeActionsUsing(), or a custom callback.

Look elsewhere or defer when:

• The application is not Laravel/Filament, or moderation must run as a standalone service outside the admin panel.
• The team uses a different moderation provider or does not need AI classification plus human review.
• Requirements include complex workflow features not present here: assignment/routing, SLAs, escalations, appeals, multi-step approvals, team-specific queues, or advanced moderator productivity tooling.
• The organization needs vendor-supported SLAs, mature enterprise compliance guarantees, or a long-term supported release rather than a v0.1.0 initial release.
• Analytics must go beyond the built-in 7/30/90-day operational views and 90-day calibration window into a data warehouse or custom BI model.
• Moderation is fully automated and no human review queue is needed.
• The team cannot take on the dependency and version alignment with phoenix1331/laravel-jev-moderation 0.2.

How to Pitch It (Stakeholders)

• Executives: This package gives us a moderation operations console inside our existing Laravel/Filament stack: a unified human review queue, approve/reject/reclassify actions, audit history, author history, calibration feedback, and operational metrics. It reduces the build cost of launching AI-assisted moderation and gives us evidence to tune review/block thresholds, staff reviewers, and demonstrate oversight. The main caveat is maturity: it is a v0.1.0 release with a hard dependency on Jev moderation core 0.2, so we should position it as a validated pilot before relying on it for critical enforcement.
• Engineering: This is a Filament plugin layered on phoenix1331/laravel-jev-moderation 0.2. It provides pages, widgets, queue actions, filters, decision/author history, calibration and volume reporting, plugin configuration for model scoping and author relations, and authorization hooks plus ModerationDecision policy support. Integration work is mainly config, policy/authorization setup, model relations, and threshold rescan operations. We should review the v0.1.0 code and upgrade posture before adoption, verify the default authorization behavior—no callback means access only in local—and confirm that config changes plus jev-moderation:rescan fit our deployment process.