Components · Layer 4 · copy-in

Generative UI

Runtime-composed interfaces. All 16 components, distributed copy-in.

SummaryCard

Tokens processed
1.84B▲ 9%
vs last 24h
Avg latency
412ms▼ 6%
p95, streaming
Tool calls
27.3K▲ 18%
across all agents
SummaryCardnpx shadcn add https://charcoal.nickcoma.io/r/summary-card.json

InsightCard

Retrieval hit rate climbing

Grounded answers rose to 94% after the reranker rollout, up from 81% last week.

Prompt cache nearing budget

Cached-input spend is at 88% of the monthly cap. Consider trimming system prompts.

Hallucination flagged in support flow

Three responses cited a deprecated pricing table. Grounding source needs an update.

InsightCardnpx shadcn add https://charcoal.nickcoma.io/r/insight-card.json

AdaptiveCard

Deploy retrieval model v2.3

The agent assembled this card at runtime from a validated spec.

Target
production
Rollout
10% canary
Rollback
automatic
AdaptiveCardnpx shadcn add https://charcoal.nickcoma.io/r/adaptive-card.json

DynamicForm

DynamicFormnpx shadcn add https://charcoal.nickcoma.io/r/dynamic-form.json

ComparisonView

Claude Opus
Reasoning
Highest
Cost / Mtok
$$$
Latency
Higher
Claude SonnetRecommended
Reasoning
High
Cost / Mtok
$$
Latency
Balanced
Claude Haiku
Reasoning
Good
Cost / Mtok
$
Latency
Lowest
ComparisonViewnpx shadcn add https://charcoal.nickcoma.io/r/comparison-view.json

ConfirmationPreview

The agent will change these settings
Modelclaude-sonnetclaude-opus
Max tokens40968192
Temperature0.70.3
Streamingenabled
ConfirmationPreviewnpx shadcn add https://charcoal.nickcoma.io/r/confirmation-preview.json

DecisionTree

  • Route the incoming request
    Agent classifies intent before answering
    • Is it a factual product question?
      • Retrieve from docs, then answer
        RAG over the knowledge base
      • Cite sources inline
        Grounding required
    • Does it require an action?
      • Draft a tool call
        Pause for confirmation if high stakes
      • Execute and summarize
        Low-risk actions only
    • Otherwise escalate to a human
      Ambiguous or sensitive requests
DecisionTreenpx shadcn add https://charcoal.nickcoma.io/r/decision-tree.json

ActionPanel

Suggested next steps
ActionPanelnpx shadcn add https://charcoal.nickcoma.io/r/action-panel.json

LiveTable

Live
Agent runModelTokensLatencyStatus
run_4f2aopus12,4801.2scomplete
run_4f2bsonnet8,210640mscomplete
run_4f2csonnet5,904590msstreaming
run_4f2dhaiku2,133210msqueued
LiveTablenpx shadcn add https://charcoal.nickcoma.io/r/live-table.json

SchemaDataDisplay

tool
search_knowledge_base
query
refund policy for annual plans
results
3
top_match
doc_id
kb_2291
score
0.92
section
Billing / Refunds
latency_ms
340
grounded
true
SchemaDataDisplaynpx shadcn add https://charcoal.nickcoma.io/r/schema-data-display.json

StreamingChart

tokens/sec
p95 latency (ms)
StreamingChartnpx shadcn add https://charcoal.nickcoma.io/r/streaming-chart.json

ProgressiveDisclosure

ProgressiveDisclosurenpx shadcn add https://charcoal.nickcoma.io/r/progressive-disclosure.json

ContextSwitcher

active context: prod
ContextSwitchernpx shadcn add https://charcoal.nickcoma.io/r/context-switcher.json

FeedbackCollector

Was this answer helpful?
FeedbackCollectornpx shadcn add https://charcoal.nickcoma.io/r/feedback-collector.json

SlotContainer

Draft ready for review

The agent filled this slot once the summary finished generating.

slot: citations
SlotContainernpx shadcn add https://charcoal.nickcoma.io/r/slot-container.json

GenerativeLayout

Weekly digest
128 runs▲ 11%
composed by the agent
MetricValue
Regionus-east
Grounded answers94%
Escalations6
Avg rating4.6 / 5
GenerativeLayoutnpx shadcn add https://charcoal.nickcoma.io/r/generative-layout.json