Skin comparison lab

One shared scenario drives all panes. Pick Reactor and Foundry skins independently, then score them.

Shared scenario

Same Quick Estimate → same hybrid mix and monthly cost. Only the renderer changes.

Same numbers

UsersRequests / user / dayAvg input tokensAvg output tokensRetrieved context tokensExpected growth % / yrUse-case typeData sensitivityLatency target (sec)Freshness (hours)Source-level permissions requiredCitations required

Mix 56/0/44$8,051/moUtil 58%Readiness 35/100Ghosts 9

Reactor signature

Token stream / allocation moment across three skins.

Judge for: Editability, loop clarity

Clarity3Credibility3Shareability3

Foundry request-trace

One question traveling through the architecture.

Chat / Copilot surface

SSO + user context

Authorization

Warehouse / SQL stores

Document corpus

Parse · chunk · enrich

Hybrid lexical + vector

Vector store

Embedding pipeline

Structured query (text-to-SQL)

GraphRAG-lite entity hop

Entity-link pass (GraphRAG-lite)

Domain router

Specialist retrieval agents

Intent classifier (small model)

Rerank + compress

Composer — cited assembly

Four-tier memory

Context Graph

OpenAI-compatible endpoint abstraction

Models routed by intent

Citations + groundedness

Golden sets

Tracing · cost · drift

Source-level ACL filter

Incremental ingestion

Model failover chain

DLP / redaction

MCP gateway

Context engineering budget

A2A observability

Routing & answer evals

Retention & privacy policy

Clarity3Credibility3Shareability3

Loading compare lab…