Files
Socrates/apps/web/lib/llm/seedInterview.ts
dtoro b55425cc68 Pivot to text-first column-stack workspace + merge-with-review across AI artifacts
Workspace
- Pivot from "set of open panes" to a Finder-style miller column stack:
  TopBar / LeftSidebar / [section → entity → entity ...] / pinned text editor.
  openPanesStore is now an ordered Column[] with pushFrom / closeFrom /
  setStack; only one top-level section is rooted at a time.
- New entity column panes: Term, Block, Association, Constraint,
  Requirement, Finding. Click-through navigation truncates deeper
  columns automatically.
- LeftSidebar surfaces a pending-count chip per section (single-glance
  navigation cue) and spins its analyze ↻ via SVG Spinner whenever the
  LLM is working — including server-initiated runs caught by the runs
  poll, not just user-triggered ones.

Analyze pipeline + persistence
- Unified `concepts` pass (taxonomy + glossary in one LLM call) replaces
  the two-pass setup. Server still accepts ?section=taxonomy|glossary
  and normalizes them for back-compat.
- model / requirements / detection (assumptions, risks, inconsistencies)
  + cross-layer validation rules (X1–X4: stale term link, unlinked
  formalism, undefined linked term, prose-only term).
- Persistence: NarrativeDocument, ModelSnapshot, ChangelogEntry,
  TaxonomyTerm, RequirementEntry, Finding, AnalysisRun. Re-runs MERGE
  instead of replace: gentle update on existing items, suggested on new,
  deprecated on missing — same idiom for every artifact kind. User pins
  preserve "kept" decisions across re-analyses.
- Migrations: pivot_text_first, add_requirement_linked_term,
  term_review_state, review_state_for_reqs_and_findings,
  add_term_definition_pinned.

Concept ↔ ontology integration
- linkedTermId on Block / Association / Constraint / Requirement.
  PromoteToolbar lets the user formalize a concept inline: + Block /
  + Association / + Constraint / + Requirement, all routed through
  applyOps so undo/redo and SSE work for free.
- decideElement op for in-canvas keep/discard on review-pending model
  elements.

User-authored definitions
- TermColumn definition is click-to-edit. Save (Cmd-Enter / blur),
  Cancel (Esc), Reset to AI suggestion when pinned.
- definitionPinned flag on TaxonomyTerm: future Analyze runs leave the
  user's text alone. setTermDefinition repo function + POST
  /api/projects/[id]/terms/[termId]/definition endpoint.
- mergeTaxonomySuggestion + applyGlossaryDefinitions both pin-aware.

UX/UI
- StatusChip: single component for all state idioms (suggested,
  deprecated, accepted, dismissed, resolved, severity, validation code,
  confidence, warn). Replaces 5+ ad-hoc badge classes.
- PaneControls (PaneViewTabs + PaneFilterChip): separates view-mode
  toggles from filter chips so toggling Pending no longer flips you off
  the current view.
- PaneEmpty: unified empty-state with title + hint + action.
- PaneDrawer: collapsible groups for Pending / Discarded review; cards
  group as Kept (top) → Pending (bottom drawer) → Discarded (Findings
  only, hidden when empty). Restore action recovers dismissed/resolved
  findings.
- ConceptCard unifies Tree and A–Z views in Concepts; only Tree parents
  carry the chevron (no empty placeholder offset).
- Type + spacing tokens (--text-xs..xl, --space-1..6, --lh-tight/ui/
  prose, --radius-*) replace every ad-hoc value.
- Buttons standardized to body sans 500 (was a mishmash of mono / display).
- Card shells unified across Concepts / Requirements / Findings.

Cleanup
- Removed: LeftRail, FindingsPanel, IssuesPanel, SocratesDock,
  ProposalCard, SlashMenu, SlashExtension, slashSuggestion,
  CanvasHeader, TaxonomyPane, GlossaryPane, TermDetail (popover; now
  TermColumn).
- Section ids in openPanesStore: dropped taxonomy/glossary, added
  concepts. localStorage migration runs on hydrate.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-01 00:12:06 +02:00

151 lines
4.8 KiB
TypeScript

// Live seed interview — per-turn handler.
//
// Stateless on the server: the client passes the running thread + the draft
// extracted so far, plus the user's latest reply. We call the LLM with the
// character + interview prompts and get back { assistant turn, updated draft,
// confidence, ready }.
import "server-only";
import { defaultGateway, chatJSON, type Message } from "./gateway";
import { loadPrompt } from "./prompts";
export interface SeedDraft {
title: string;
problem: string;
targetUser: string;
desiredOutcome: string;
initialHypothesis?: string;
constraints?: string[];
}
export interface InterviewTurn {
role: "socrates" | "user";
text: string;
}
export interface InterviewStepArgs {
/** Running interview history. Empty array on the first call → Socrates opens. */
history: InterviewTurn[];
/** Latest user reply (empty string for the opening turn). */
userText: string;
/** Draft extracted so far. Empty on the first call. */
draft: SeedDraft;
}
export interface InterviewStepResult {
text: string;
draft: SeedDraft;
confidence: number;
ready: boolean;
inputTokens: number;
outputTokens: number;
provider: string;
model: string;
}
const interviewJsonSchema = {
type: "object",
additionalProperties: false,
required: ["text", "draft", "confidence", "ready"],
properties: {
text: { type: "string", minLength: 1 },
draft: {
type: "object",
additionalProperties: false,
required: ["title", "problem", "targetUser", "desiredOutcome"],
properties: {
title: { type: "string" },
problem: { type: "string" },
targetUser: { type: "string" },
desiredOutcome: { type: "string" },
initialHypothesis: { type: "string" },
constraints: { type: "array", items: { type: "string" } },
},
},
confidence: { type: "number", minimum: 0, maximum: 1 },
ready: { type: "boolean" },
},
} as const;
interface RawResponse {
text?: string;
draft?: Partial<SeedDraft>;
confidence?: number;
ready?: boolean;
}
export async function interviewStep(args: InterviewStepArgs): Promise<InterviewStepResult> {
const character = loadPrompt("socrates/character.md");
const interview = loadPrompt("socrates/interview.md");
const userTurnsSoFar = args.history.filter(t => t.role === "user").length;
const remaining = Math.max(0, 5 - userTurnsSoFar - (args.userText.trim().length > 0 ? 1 : 0));
const messages: Message[] = [
{
role: "system",
content: [
character,
"---",
interview,
"---",
`Draft so far (your previous extraction):\n\`\`\`json\n${JSON.stringify(args.draft, null, 2)}\n\`\`\``,
`Turns remaining before ready signal becomes mandatory: ${remaining}`,
].join("\n\n"),
},
];
// Replay history so the LLM has full context.
for (const t of args.history) {
messages.push({ role: t.role === "socrates" ? "assistant" : "user", content: t.text });
}
if (args.userText.trim().length > 0) {
messages.push({ role: "user", content: args.userText });
} else if (args.history.length === 0) {
messages.push({ role: "user", content: "Begin the interview." });
}
const gateway = defaultGateway();
const { value, result } = await chatJSON<RawResponse>(gateway, messages, {
temperature: 0.4,
maxTokens: 768,
jsonSchema: { name: "interview_step", schema: interviewJsonSchema as Record<string, unknown> },
jsonObjectMode: true,
maxRepairs: 2,
});
// Normalize / merge — the LLM may emit a partial draft; we union with the
// prior draft so a user backtracking doesn't wipe a previously-confirmed field.
const incoming: Partial<SeedDraft> = value?.draft ?? {};
const draft: SeedDraft = {
title: nonEmpty(incoming.title) ?? args.draft.title ?? "",
problem: nonEmpty(incoming.problem) ?? args.draft.problem ?? "",
targetUser: nonEmpty(incoming.targetUser) ?? args.draft.targetUser ?? "",
desiredOutcome: nonEmpty(incoming.desiredOutcome) ?? args.draft.desiredOutcome ?? "",
initialHypothesis: nonEmpty(incoming.initialHypothesis) ?? args.draft.initialHypothesis,
constraints: Array.isArray(incoming.constraints) ? incoming.constraints : args.draft.constraints,
};
return {
text: typeof value?.text === "string" && value.text.length > 0 ? value.text : "[empty response]",
draft,
confidence: clamp01(value?.confidence ?? 0),
ready: !!value?.ready,
inputTokens: result.inputTokens,
outputTokens: result.outputTokens,
provider: gateway.provider,
model: gateway.model,
};
}
function nonEmpty(s: unknown): string | undefined {
if (typeof s !== "string") return undefined;
const t = s.trim();
return t.length > 0 ? t : undefined;
}
function clamp01(n: number): number {
if (Number.isNaN(n)) return 0;
return Math.max(0, Math.min(1, n));
}