Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
JavaScript ai-agentsdashboarddeveloper-toolsdigital-twinquantified-selfsecond-brainself-hostedtailscale
import { MessagesSquare } from 'lucide-react';
import * as api from '../../../services/api';
import { MULTI_TURN_STATUS } from '../constants';
import TwinEvaluationSuitePanel, { SuiteModelResponses } from './TwinEvaluationSuitePanel';
/**
* Multi-Turn Conversation testing (M34 P6). Plays out each scenario's user turns
* in order — the twin sees its own prior replies plus the next message — and
* grades whether it stayed *consistent* across the whole conversation rather
* than contradicting itself, caving to pushback, or forgetting a constraint.
* Reuses the provider/model + persona selection from the parent TestTab via
* props so all suites share one configuration. Lifecycle + shared presentation
* live in `TwinEvaluationSuitePanel`; this file only declares the Multi-Turn
* suite descriptor and its expanded-detail renderer (a full transcript).
*/
const SUITE = {
HeaderIcon: MessagesSquare,
title: 'Multi-Turn Conversation Tests',
description: 'Multi-message exchanges scored on whether your twin stays consistent across the whole conversation.',
runLabel: 'Run Conversations',
loadingText: 'Loading conversation suite',
itemLabel: 'Conversation',
emptyState: (
<>
No multi-turn suite found. Add a <span className="text-gray-300">MULTI_TURN_SUITE.md</span> to your digital twin folder to enable these tests.
</>
),
scoreLabel: 'Consistency Score',
countField: 'consistent',
historyTitle: 'Recent Conversation Runs',
statusMap: MULTI_TURN_STATUS,
passResult: 'consistent',
failResult: 'inconsistent',
getTests: api.getMultiTurnTests,
getHistory: api.getMultiTurnTestHistory,
runTests: api.runMultiTurnTests,
successToast: 'Multi-turn conversation tests completed'
};
const renderTranscript = (tr) => (
<div className="bg-port-card p-3 rounded space-y-2">
{(tr.transcript || []).map((msg, i) => (
<div key={i} className={msg.role === 'twin' ? 'pl-3 border-l-2 border-port-accent' : ''}>
<span className={`text-xs uppercase tracking-wide ${msg.role === 'twin' ? 'text-port-accent' : 'text-gray-500'}`}>
{msg.role === 'twin' ? 'Twin' : 'User'}
</span>
<p className="text-white whitespace-pre-wrap">{msg.content}</p>
</div>
))}
{tr.reasoning && (
<p className="text-sm text-gray-400 mt-2 pt-2 border-t border-port-border">
Reasoning: {tr.reasoning}
</p>
)}
</div>
);
const renderDetail = (scenario, ctx) => (
<div className="space-y-4">
<div>
<h4 className="text-sm font-medium text-gray-400 mb-1">User Turns</h4>
<ol className="space-y-2 list-decimal list-inside">
{scenario.turns?.map((turn, i) => (
<li key={i} className="text-white bg-port-card p-3 rounded">{turn}</li>
))}
</ol>
</div>
<div className="grid grid-cols-1 sm:grid-cols-2 gap-3 text-sm">
<div>
<div className="text-port-success text-xs mb-1">Consistent Trajectory</div>
<div className="text-gray-400 bg-port-card p-3 rounded">{scenario.consistentTrajectory}</div>
</div>
<div>
<div className="text-port-error text-xs mb-1">Inconsistent Trajectory</div>
<div className="text-gray-400 bg-port-card p-3 rounded">{scenario.inconsistentTrajectory}</div>
</div>
</div>
<SuiteModelResponses item={scenario} label="Conversation" renderBody={renderTranscript} {...ctx} />
</div>
);
export default function MultiTurnPanel(props) {
return <TwinEvaluationSuitePanel suite={SUITE} renderDetail={renderDetail} {...props} />;
}