# Prism Pack Family

One canonical identity. Human page: https://www.insightits.com/products/prism-pack.html

| Field | Value |
|-------|-------|
| Slug | `prism-pack` |
| Status | published |
| Family | research |
| Infra category | — |
| Record type | product |
| Parent | — |
| JSON | https://www.insightits.com/catalog/prism-pack.json |
| Markdown | https://www.insightits.com/catalog/prism-pack.md |
| GitHub | https://github.com/insightitsGit/PrismGuard/blob/main/docs/scorecard.md |
| PyPI | — |
| Install | `Soft CTA: Guardrail Scorecard. Pins: chorusgraph 1.3.0 · prismguard 0.1.10 · prismshine 0.2.2 · seed 42.` |
| Version | 2026-07-26 |

## What it is

Prism Pack Family board: FinancePackBench job-by-job wins for the wired stack PrismGuard → ChorusGraph → PrismShine. Not a pip package.

## Problem

Buyers ask whether the integrated pack beats a composed LangGraph+LLM Guard stack on a published finance bench — this page is that evidence board, with disclosures.

## Alternatives

- **LangGraph + LLM Guard (FinancePackBench peer)** — Vendor-authored Race E / mid bake-off. Cite eye-to-eye job wins and disclosures only. ([source](https://www.insightits.com/products/prism-pack.html))

## Features

- **Pack task + PI** — Task 100% / PI 100% attributed to Pack (Guard → ChorusGraph → Shine), not a single cell.
- **Embed tax vs re-embed** — PrismAPI 0.70 vs 4.20 mean embeds; multi-worker fleet ~6× — never API vs database.
- **Warm retrieve Race C** — PrismDriver 0.057 ms vs plain SQL 20.0 ms P50 (~353×). Separate peer map from embeds.

## Architecture

Wire: PrismGuard → ChorusGraph → PrismShine; add PrismAPI for multi-worker embeds; add PrismDriver for warm FAQ vs SQL; PrismCortex optional (no quality claim). Pins: chorusgraph 1.3.0 · prismguard 0.1.10 · prismshine 0.2.2 · seed 42.

## Use cases

- Read the pack family evidence before a Guardrail Scorecard

## Benchmarks

Vendor-authored FinancePackBench pack family board. Cite eye-to-eye job wins and disclosures only — not an overall AWS / AgentCore win.

### Disclosures

- Do not claim the pack beat AWS overall.
- Do not claim PrismAPI is faster than DB.
- Do not cite Cortex as a FAQ quality win.
- Do not treat Race E lexical 10% as a grounding win.

### Evidence URLs

- https://www.insightits.com/products/prism-pack.html


## GitHub

https://github.com/insightitsGit/PrismGuard/blob/main/docs/scorecard.md

## PyPI

_No PyPI package for this identity._

## Documentation

- [Guardrail Scorecard](https://github.com/insightitsGit/PrismGuard/blob/main/docs/scorecard.md)
- [https://www.insightits.com/products/prismguard.html](https://www.insightits.com/products/prismguard.html)
- [https://www.insightits.com/products/chorusgraph.html](https://www.insightits.com/products/chorusgraph.html)
- [https://www.insightits.com/products/prismshine.html](https://www.insightits.com/products/prismshine.html)

## Is not

- A claim that the pack beat AWS / AgentCore overall
- PrismAPI faster than DB
- Cortex FAQ quality win
- Race E lexical 10% as a grounding win
- 100% grounded from allow

## Related products

- [prismguard](./prismguard.md)
- [chorusgraph](./chorusgraph.md)
- [prismshine](./prismshine.md)
- [choruscontrol](./choruscontrol.md)