FAQ

Detailed FAQ: docs/FAQ.md
Russian / Spanish / French / Chinese: in-repo only — docs/FAQ.ru.md · docs/FAQ.es.md · docs/FAQ.fr.md · docs/FAQ.zh.md · see Languages

What is AI-Factory?

Self-hosted multi-agent pipeline: idea → spec → code → QA/E2E → security → deploy → optional storefront.

Storefront vs Admin?

• Storefront — filtered marketplace-ready products
• Admin → Pipeline — all products, all stages (source of truth)

Default password?

None in repo. Bootstrap via TTY or data/secrets/bootstrap_admin.txt. Demo host magic-ai-factory.com: passwordless admin (admin, Enter admin demo).

How long does a run take?

Type	Typical
marketing_landing	~20–25 min (DeepSeek, no long repair loops)
full_software	~25–45 min simple brief; hours with gate iterations
First COMPLETED (measured)	PulseBoard Mini prod-a25224bcb746: ~91 min wall-clock

LLM cost?

Landing first pass ~$0.30–$2. Tiny full_software can hit first COMPLETED under ~$0.20 (PulseBoard Mini measured $0.181 / ~$0.18); $3–15+ when QA/repair loops pile up. Bring your own keys. See FAQ detail in docs/FAQ.md / root README FAQ.

Why is my product stuck in repair?

Gate failure (demo/TZ, browser crawl, security, methodologist). State: BUG_FOUND → DEV_FIXING. Check Pipeline tab errors and LLM Logs.

AIFACTORY_GATE_FAILING_MODEL?

Optional model id override on the same LLM provider during repair rounds (quality_repair_round ≥ 1). Does not switch providers. Example: DeepSeek-only fleet → set deepseek-reasoner for hard fixes.

AIFACTORY_MAX_QUALITY_LOOPS?

Caps policy audit / remediation cycles before FAILED (default 8).

Discovery not creating products?

Check Director scheduler, AIFACTORY_DISCOVERY_INTERVAL_HOURS, source health at data/discovery/source_health.json, and admin POST /api/admin/discovery/run.

Gallery images missing on GitHub README?

Gallery lives in repo docs/gallery/*.webp. Some mirrors omit binary paths — clone from Gitea or refresh with scripts/capture_gallery_landings.py.

More

Pipeline · Security · Owner-Guide · docs/FAQ.md