FAQ and Troubleshooting

Marketplace Readiness page with proof path and demo reset controls — **First check:** Marketplace Readiness is the fastest place to confirm demo setup, proof path links, and buyer-facing readiness.

System Health page showing operational findings and remediation links — **Troubleshooting:** System Health surfaces storage, scheduler, mail, webhook, and governance monitoring issues before launch.

I see a blank page

Run php artisan optimize:clear, check PHP error logs, confirm storage and bootstrap/cache are writable, and verify the domain points to /public.

Ask Data says OpenAI is disabled

Open Settings -> OpenAI and enter both the API key and model ID. The demo dashboards remain usable while OpenAI is disabled.

External database test fails

Check host, port, username, password, firewall, database grants, and whether the external database allows remote MySQL connections. Use a read-only user.

Generated SQL is blocked

The guard blocks writes, DDL, multiple statements, hidden tables, and non-whitelisted columns. Open Schema Explorer and verify the intended tables and columns are AI-enabled.

Scheduled reports do not send

Test SMTP settings, configure cron, and verify queue/scheduler permissions on the server.

Does InsightPilot copy my whole business database?

No. InsightPilot stores connection settings, schema metadata, query history metadata, reports, and short-lived cached result rows. It is designed to avoid permanently storing full external database result sets.

Is it safe to connect a production database?

Use a read-only database user, hide sensitive tables, disable sensitive columns, and review permissions. InsightPilot also enforces read-only SQL safety rules, but database-level read-only access is still strongly recommended.

Will OpenAI cost money?

Yes, live AI generation uses the buyer's OpenAI account and may create API usage charges. Demo data and fallback screens can be used without enabling OpenAI.

Why is there no default OpenAI model?

The admin must choose the model intentionally. This avoids surprise billing, unsupported model assumptions, or using a model the buyer does not want.

How do I know a document answer is trustworthy?

Use citation-first answers. Review document name, rule number, section snippet, linked documents, confidence, caveats, and evidence report exports before relying on an answer.

Is AI Governance legal advice?

No. AI Governance includes practical EU AI Act-inspired labels and NIST AI RMF-style control mapping, but it is not legal advice. Buyers should adapt rules, notices, controls, and approval templates with qualified legal, privacy, security, and compliance reviewers.

Does AI Governance replace the Risk Center?

No. Risk Center is for document and compliance risk around uploaded or linked documents. AI Governance is for AI use-case inventory, risk classification, approval workflow, vendors/models, evidence packs, notices, incidents, evaluations, controls, and monitoring.

Does AI Governance need OpenAI?

Only the optional AI Approval Summary uses the configured AI provider. If OpenAI is missing, the app creates a deterministic reviewer brief from the use case, answers, risk factors, classification, and reviewer tracks.

Can a requester approve their own high-risk AI use case?

Not for required high-risk or unknown approval tracks unless the requester has override ai reviewer assignments. Compliance, privacy, security, and legal reviewers should be distinct from the requester.

Can AI Governance evidence include files and URLs?

Yes. Evidence can be uploaded files, URLs, or links to existing InsightPilot documents. Uploaded files are hashed and can be included in evidence-pack ZIP exports.

What is the AI Governance Reviewer Inbox?

It is the daily work queue for assigned reviews, unassigned steps, requested changes, approval conditions, overdue controls, alerts, serious incidents, and unread governance notifications.

What are vendor questionnaires for?

They document vendor/model evidence such as DPA, security overview, model/system card, retention, subprocessors, incident contact, privacy controls, and GPAI/copyright notes. Low scores create review alerts.

Are public transparency notice links safe?

Public notice links expose only approved notice text and basic public metadata. They do not expose assessments, comments, evidence files, reviewer identities, or approval internals. Admins can revoke links.

What is the difference between evidence packs and governance reports?

Evidence packs are ZIP exports with a PDF, manifest, and attached evidence where available. Governance reports are PDF summaries for use cases, approvals, vendors, or evidence records with tracked status and checksum.

How do I seed AI Governance demo records?

Run php artisan insightpilot:demo-install for a full buyer preview. It includes fictional AI Governance records. Run php artisan insightpilot:seed-ai-governance-demo --user_id=1 only when refreshing AI Governance records on an existing preview.

Does AI Governance have API access?

Yes. Use the /api/v1/ai-governance/* endpoints with Sanctum abilities such as ai-governance:read, ai-governance:write, ai-governance:review, ai-governance:evidence, and ai-governance:admin.

Why are AI Governance alerts not appearing?

Run php artisan insightpilot:run-ai-governance manually and verify cron runs php artisan schedule:run every minute. Also confirm the current user or organization has approvals, controls, evaluations, evidence, or incidents that match monitoring rules.

Why does System Health show a critical finding?

Open /insight-pilot/system-health, run checks manually, then follow the remediation link. Common causes are stale cron, failed webhook backlog, unwritable private storage, missing mail configuration, failed jobs, or stale AI Governance monitoring.

How do I reopen installer?

Delete storage/app/installed only on a controlled setup environment. Do not reopen installer on a production site without a backup.