DeepTutor: Lifelong Personalized Tutoring. https://deeptutor.info/.
Python ai-agentsai-tutorclawbotcli-tooldeepresearchinteractive-learninglearningmulti-agent-systemsrag
/**
* Rail model for a Group discussion.
*
* The product-chat rail answers "where did I ask that?". A Group has a second
* kind of turn — an approved Partner-to-Partner question — and the point of
* putting it on the same rail is rhythm: at a glance you can see where you
* drove the conversation and where the Partners took it between themselves.
* Those ticks are toned differently rather than merged into the user's, so
* the two never get confused.
*/
import { plainTextPreview, type ChatOutlineEntry } from "@/lib/chat-outline";
import type { Round } from "./useGroupSession";
/** Longest title in the set defines the full-width tick. */
function normalise(entries: ChatOutlineEntry[]): ChatOutlineEntry[] {
const longest = entries.reduce(
(max, entry) => Math.max(max, entry.title.length),
1,
);
return entries.map((entry) => ({
...entry,
weight: Math.min(1, entry.title.length / longest),
}));
}
/** First seat that actually produced prose, as the hover preview's answer. */
function firstReply(round: Round, skip: number): string {
for (let index = skip; index < round.seats.length; index += 1) {
const seat = round.seats[index];
const body = seat.message?.content ?? seat.streamed;
const preview = plainTextPreview(body, 160);
if (preview) return preview;
}
return "";
}
export function buildGroupOutline(rounds: Round[]): ChatOutlineEntry[] {
const entries: ChatOutlineEntry[] = [];
rounds.forEach((round, index) => {
if (round.followup) {
// The question is the round's own first seat; the answer follows it.
const question = round.seats[0]?.message;
const title = plainTextPreview(question?.content ?? "", 160);
if (!title) return;
const invocation = round.seats.find((seat) => seat.message?.invocation)
?.message?.invocation;
const badge =
invocation && invocation.requester_partner_name
? `${invocation.requester_partner_name} → ${invocation.target_partner_name}`
: question?.author_name || "";
entries.push({
key: round.turnId,
index,
ordinal: entries.length + 1,
title,
reply: firstReply(round, 1),
weight: 0,
tone: "peer",
badge,
});
return;
}
const title = plainTextPreview(round.user?.content ?? "", 160);
if (!title) return;
entries.push({
key: round.turnId,
index,
ordinal: entries.length + 1,
title,
reply: firstReply(round, 0),
weight: 0,
tone: "user",
});
});
return normalise(entries);
}