Question 1

Is there an interactive demo?

Accepted Answer

Yes — live in the browser with no install: https://insightitsgit.github.io/PrismShine/demo.html. Walk the named resolution_gate (block vs shrug-score checkers), then pip install prismshine for your stack. Smoke after install: prismshine verify --demo.

Question 2

Is PrismShine a firewall?

Accepted Answer

No. PrismShine is an anti-hallucination verdict engine — it verifies answers against a provided preload and can halt broken preloads before generation. For prompt-injection defense, use PrismGuard (https://www.insightits.com/products/prismguard.html).

Question 3

Do I need ChorusGraph?

Accepted Answer

No. The core library works standalone via ShineGate.verify or the CLI. Optional [chorusgraph] extras and require_shine wiring give the richest runtime integration (ledger-aware cause catch), but they are not required for basic verification.

Question 4

Does PASS mean the answer is true?

Accepted Answer

No. PASS means grounded in the preload you provided — not that the preload is world-true. See docs/LIMITS.md on GitHub for the contract.

Question 5

What’s the difference between the two benchmark tables?

Accepted Answer

The HHEM table is PrismShine alone on HaluEval (Shine-only vs HHEM-2.1-Open). The runtime table is ChorusGraph + PrismShine wired against LangGraph-shaped peers — including cause failures only an evidence-aware gate can catch (R1).

Question 6

Why do competitors score 0 on R1?

Accepted Answer

They are content-only and ignore the runtime ledger (saw_evidence=false). That gap is the product — empty retrieval, tool fail, and stale cache show up as cause failures when Shine is wired.

Question 7

Are there LLM costs on the default path?

Accepted Answer

Default path uses zero LLM calls. Optional Tier-4 judge is opt-in for gray-zone cases only.

Question 8

Is Tier-3 ONNX in the wheel?

Accepted Answer

No. Bare pip is Tiers 0–2. For run4-parity span SotA: pip install "prismshine[spans]" then python -m prismshine.tools.ensure_span_onnx --export (~1 GB). Without that, gray zones may show MISSING_CAPABILITY_FLAG — we do not fake span SotA.

Question 9

What’s free vs paid?

Accepted Answer

The library is free Apache-2.0 on PyPI (Tiers 0–2; optional Tier-3 ONNX download; CLI; BYO / LangGraph / ChorusGraph wiring). Pro/Team and Enterprise are support, calibration, staging wiring, and compliance packaging — pre-validation estimates only until finance confirms. Alpha/beta: PASS ≠ world-true; see the wired FULL_REPORT for method.

Question 10

How do I reproduce the HHEM numbers?

Accepted Answer

See benchmarks/progress/2026-07-20_run4_onnx and docs/BENCHMARKS.md. Receipt date 2026-07-20 on Azure ACI with ONNX Tier-3: prismshine-fast B1 QA F1 0.831 vs HHEM-2.1-Open 0.746; B2 numbers F1 1.000 (0 FP); B1 p50 90 ms vs 216 ms. Absolute Bsum is still modest — do not claim “beats all judges.”