SRCmail/app/api/ai/opencode/chat/route.ts

import { NextRequest, NextResponse } from 'next/server';
import { getStalwartCredentials } from '@/lib/stalwart/credentials';
import { configManager } from '@/lib/admin/config-manager';
import { findOpencodeServer, parseModelRef, opencodePrompt } from '@/lib/ai/opencode';
import { logger } from '@/lib/logger';

export const runtime = 'nodejs';

const MAX_BODY_BYTES = 200 * 1024;

interface ChatMessage {
  role: 'system' | 'user' | 'assistant';
  content: string;
}

/**
 * POST /api/ai/opencode/chat — one-shot chat against a locally-running
 * `opencode serve`.
 *
 * Deliberately NOT entitlement-metered, unlike /api/ai/server/chat: this runs
 * on the user's own machine against provider credentials opencode itself
 * holds, so there is no centrally-borne cost for this app to bill — the same
 * reasoning that leaves `local` unmetered (lib/ai/entitlement.ts's header).
 */
export async function POST(request: NextRequest) {
  const auth = await getStalwartCredentials(request);
  if (!auth) {
    return NextResponse.json({ error: 'not authenticated' }, { status: 401 });
  }

  await configManager.ensureLoaded();
  if (configManager.getAiConsoleConfig().classesEnabled.opencode === false) {
    return NextResponse.json({ error: 'the OpenCode class is disabled by admin policy' }, { status: 403 });
  }

  const rawBody = await request.text();
  if (rawBody.length > MAX_BODY_BYTES) {
    return NextResponse.json({ error: 'request too large' }, { status: 413 });
  }

  let body: { model?: unknown; messages?: unknown };
  try {
    body = JSON.parse(rawBody);
  } catch {
    return NextResponse.json({ error: 'invalid JSON body' }, { status: 400 });
  }

  const model = typeof body.model === 'string' ? body.model : '';
  const messages = Array.isArray(body.messages) ? (body.messages as ChatMessage[]) : null;
  if (!model || !messages || messages.length === 0) {
    return NextResponse.json({ error: 'model and messages are required' }, { status: 400 });
  }

  const found = await findOpencodeServer();
  if (!found) {
    return NextResponse.json(
      { error: 'No local OpenCode server is running. The desktop app starts one automatically when the opencode CLI is installed \u2014 install it from opencode.ai, then restart VNCmail+.' },
      { status: 503 },
    );
  }
  if (!found.models.some((m) => m.ref === model)) {
    // The picker is populated from this same list, so a mismatch means the
    // saved model was removed/renamed in opencode since it was chosen -
    // clearer to say so than to forward it and surface opencode's own error.
    return NextResponse.json(
      { error: `OpenCode no longer offers the model "${model}" \u2014 pick another in Settings.` },
      { status: 400 },
    );
  }
  const parsed = parseModelRef(model);
  if (!parsed) {
    return NextResponse.json({ error: `Malformed model reference "${model}"` }, { status: 400 });
  }

  // Flatten our chat-messages shape onto opencode's (system field + text
  // parts). Every non-system message is already just the built prompt.
  const system = messages.filter((m) => m.role === 'system').map((m) => m.content).join('\n\n') || undefined;
  const userText = messages.filter((m) => m.role !== 'system').map((m) => m.content).join('\n\n');
  if (!userText.trim()) {
    return NextResponse.json({ error: 'no user content to send' }, { status: 400 });
  }

  try {
    const result = await opencodePrompt(found.baseUrl, parsed, system, userText);
    if (!result.ok) {
      logger.error('opencode prompt failed', { error: result.error });
      return NextResponse.json({ error: result.error }, { status: 502 });
    }
    return NextResponse.json({ answer: result.answer });
  } catch (cause) {
    logger.error('opencode chat failed', { error: cause instanceof Error ? cause.message : String(cause) });
    return NextResponse.json({ error: 'OpenCode server unreachable' }, { status: 502 });
  }
}