File size: 10,285 Bytes
8c1b9fe
 
 
 
 
 
 
 
 
 
656439d
 
 
 
 
 
 
 
 
5a0feef
 
8c1b9fe
 
 
 
 
 
656439d
8c1b9fe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
656439d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8c1b9fe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
656439d
8c1b9fe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
"use client";
import Link from "next/link";
import type { Citation, CorpusSummary, InsufficientReason, ModelFitSnapshot } from "@/lib/api";
import { Citations } from "@/components/Citations";
import { ConfidenceBar } from "@/components/ConfidenceBar";
import { Markdown } from "@/components/Markdown";
import { CopyButton } from "@/components/CopyButton";
import { InsufficientEvidence } from "@/components/InsufficientEvidence";
import { CorpusInventory } from "@/components/CorpusInventory";

export interface AgenticStep {
  phase: "decompose" | "hop" | "check" | "synthesize";
  label: string;
  detail?: string;
  hop?: number;
  query?: string;
  retrieved?: number;
  sub_questions?: string[];
  sufficient?: boolean;
}

export interface Turn {
  role: "user" | "assistant";
  text: string;
  citations?: Citation[];
  route?: string;
  rationale?: string;
  steps?: AgenticStep[];
  voice?: boolean;
  error?: boolean;
  status?: string;
  insufficient?: InsufficientReason | null;
  inventory?: CorpusSummary | null;
  question?: string;
  confidence?: number;
  semanticCoverage?: number;
  model_fit?: ModelFitSnapshot | null;
}

// Fit-level → semantic token (theme-aware across dark / light / comfort).
const FIT_COLOR: Record<string, string> = {
  "Excellent fit": "text-ok",
  "Recommended": "text-brand",
  "Usable with limits": "text-warn",
  "Not recommended": "text-warn",
  "Does not fit": "text-bad",
};

function ModelFitChip({ mf }: { mf: ModelFitSnapshot }) {
  const modelShort = mf.selected_model.replace(/^(ollama:|local:|hf:)/, "");
  const labelColor = FIT_COLOR[mf.fit_level ?? ""] ?? "text-fg3";
  return (
    <Link
      href="/modelfit"
      className="chip mt-2 !text-[11px] hover:border-edge2 transition-colors"
      title={`ModelFit: ${mf.fit_level ?? "scored"} — open ModelFit Index`}
    >
      <span className="text-fg3">model</span>
      <span className="max-w-[120px] truncate font-mono text-fg2">{modelShort}</span>
      {mf.fit_score != null && (
        <>
          <span className="text-fg3">·</span>
          <span className={`font-semibold ${labelColor}`}>{Math.round(mf.fit_score)}/100</span>
        </>
      )}
      {mf.quantization && (
        <>
          <span className="text-fg3">·</span>
          <span className="font-mono text-fg3">{mf.quantization}</span>
        </>
      )}
      {mf.estimated_vram_gb != null && (
        <>
          <span className="text-fg3">·</span>
          <span className="text-fg3">{mf.estimated_vram_gb.toFixed(1)} GB</span>
        </>
      )}
      {mf.estimate_used && <span className="italic text-fg3">est.</span>}
    </Link>
  );
}

function TypingDots() {
  return (
    <span className="inline-flex items-center gap-1 py-2" aria-label="Assistant is typing">
      <span className="typing-dot" />
      <span className="typing-dot" style={{ animationDelay: "0.18s" }} />
      <span className="typing-dot" style={{ animationDelay: "0.36s" }} />
    </span>
  );
}

const ROUTE_META: Record<string, { label: string; color: string; icon: string }> = {
  fast:   { label: "Fast retrieval",   color: "border-brand/40 text-brand",   icon: "⚡" },
  hybrid: { label: "Hybrid retrieval", color: "border-accent/40 text-accent",  icon: "◈" },
  graph:  { label: "Graph traversal",  color: "border-brand2/40 text-brand2", icon: "◎" },
};

function RouteTag({ route, rationale }: { route: string; rationale?: string }) {
  const meta = ROUTE_META[route] ?? { label: route, color: "border-edge text-fg3", icon: "·" };
  return (
    <div className="mb-3 flex flex-wrap items-center gap-2 border-b border-edge/40 pb-3">
      <span className={`tag font-semibold ${meta.color}`}>
        <span aria-hidden>{meta.icon}</span>
        {meta.label}
      </span>
      {rationale && (
        <span className="truncate max-w-xs text-xs text-fg3" title={rationale}>
          {rationale}
        </span>
      )}
    </div>
  );
}

const STEP_GLYPH: Record<AgenticStep["phase"], string> = {
  decompose: "◆",
  hop: "→",
  check: "✓",
  synthesize: "✎",
};

/** Live multi-hop reasoning trace for the agentic strategy. */
function AgenticSteps({ steps, live }: { steps: AgenticStep[]; live?: boolean }) {
  if (!steps.length) return null;
  return (
    <div className="mb-3 rounded-lg border border-edge/60 bg-panel/40 px-3 py-2">
      <div className="mb-1.5 flex items-center gap-1.5 text-[10px] font-semibold uppercase tracking-wider text-fg3">
        Agentic reasoning
        {live && <span className="inline-block h-1.5 w-1.5 rounded-full bg-brand animate-pulse" aria-hidden />}
      </div>
      <ol className="space-y-1">
        {steps.map((s, i) => (
          <li key={i} className="flex items-start gap-2 text-xs leading-relaxed">
            <span
              className={`mt-0.5 shrink-0 font-mono ${
                s.phase === "check" && s.sufficient === false ? "text-warn" : "text-brand"
              }`}
              aria-hidden
            >
              {STEP_GLYPH[s.phase] ?? "·"}
            </span>
            <span className="min-w-0 flex-1 text-fg2">
              {s.phase === "hop" ? (
                <>
                  <span className="text-fg3">Hop {s.hop}:</span>{" "}
                  <span className="font-medium text-fg">{s.query}</span>
                  {s.retrieved != null && <span className="text-fg3"> · {s.retrieved} passages</span>}
                </>
              ) : s.phase === "decompose" ? (
                <>
                  <span className="text-fg">{s.label}</span>
                  {s.sub_questions?.length ? (
                    <span className="text-fg3"> — {s.sub_questions.join(" · ")}</span>
                  ) : null}
                </>
              ) : s.phase === "check" ? (
                <span className="text-fg">
                  {s.sufficient ? "Evidence sufficient" : "Needs more"}
                  {!s.sufficient && s.detail ? <span className="text-fg3"> — {s.detail}</span> : null}
                </span>
              ) : (
                <span className="text-fg">{s.label}</span>
              )}
            </span>
          </li>
        ))}
      </ol>
    </div>
  );
}

function AiDot() {
  return (
    <span
      className="mt-1 flex h-6 w-6 shrink-0 items-center justify-center rounded-full bg-gradient-to-br from-brand/30 to-brand2/30 ring-1 ring-brand/20 text-[10px] font-bold text-brand"
      aria-hidden
    >
      AI
    </span>
  );
}

export function Message({
  turn,
  streaming,
  isLast,
  onRegenerate,
  onAsk,
  onIngest,
  onOpenSource,
}: {
  turn: Turn;
  streaming: boolean;
  isLast: boolean;
  onRegenerate?: () => void;
  onAsk?: (q: string) => void;
  onIngest?: () => void;
  onOpenSource?: (marker: number) => void;
}) {
  if (turn.role === "user") {
    return (
      <div className="flex justify-end">
        <div className="max-w-[85%] whitespace-pre-wrap rounded-2xl rounded-br-sm bg-gradient-to-br from-brand2/25 to-brand/15 px-4 py-3 leading-relaxed shadow-sm ring-1 ring-brand2/20 text-fg">
          {turn.voice && (
            <span className="mr-1.5 text-xs opacity-60" aria-label="voice query">
              🎙
            </span>
          )}
          {turn.text}
        </div>
      </div>
    );
  }

  // Corpus-inventory answer renders its own card.
  if (turn.inventory) {
    return (
      <div className="flex items-start gap-2.5">
        <AiDot />
        <div className="flex-1 min-w-0 rounded-2xl rounded-tl-sm border border-edge bg-panel2 px-4 py-4 shadow-md">
          <CorpusInventory summary={turn.inventory} question={turn.question} />
        </div>
      </div>
    );
  }

  const empty = !turn.text;
  const live = streaming && isLast;
  const hasCitations = (turn.citations?.length ?? 0) > 0;

  return (
    <div className="group flex items-start gap-2.5">
      <AiDot />
      <div className="flex-1 min-w-0 rounded-2xl rounded-tl-sm border border-edge bg-panel2 px-4 py-3.5 shadow-md">
        {turn.route && <RouteTag route={turn.route} rationale={turn.rationale} />}
        {turn.steps?.length ? <AgenticSteps steps={turn.steps} live={live} /> : null}

        {empty && live ? (
          <TypingDots />
        ) : turn.error ? (
          <p className="whitespace-pre-wrap leading-relaxed text-bad">{turn.text}</p>
        ) : (
          <div className="prose-answer">
            <Markdown
              text={turn.text}
              streaming={live}
              citations={turn.citations}
              onOpenCitation={onOpenSource}
            />
            {live && (
              <span className="ml-0.5 inline-block h-4 w-2 animate-pulse bg-brand/70 align-middle" />
            )}
          </div>
        )}

        {turn.insufficient && (
          <InsufficientEvidence reason={turn.insufficient} onAsk={onAsk} onIngest={onIngest} />
        )}

        {hasCitations && <Citations citations={turn.citations!} onOpenSource={onOpenSource} />}

        {!streaming && turn.model_fit && turn.model_fit.fit_score != null && (
          <ModelFitChip mf={turn.model_fit} />
        )}

        {!streaming && !turn.error && turn.confidence != null && turn.confidence > 0 && (
          <ConfidenceBar
            data={{
              overall: turn.confidence,
              semanticCoverage: turn.semanticCoverage,
            }}
          />
        )}

        {!streaming && !empty && !turn.error && (
          <div className="mt-2.5 flex items-center gap-4 opacity-0 transition-opacity duration-150 group-hover:opacity-100 focus-within:opacity-100">
            <CopyButton getText={() => turn.text} label="Copy" />
            {hasCitations && (
              <span className="text-xs text-fg3">
                {turn.citations!.length} source{turn.citations!.length === 1 ? "" : "s"}
              </span>
            )}
            {isLast && onRegenerate && (
              <button
                type="button"
                onClick={onRegenerate}
                aria-label="Regenerate answer"
                className="ml-auto inline-flex items-center gap-1 text-xs text-fg3 transition hover:text-brand"
              >
                ↻ Regenerate
              </button>
            )}
          </div>
        )}
      </div>
    </div>
  );
}