Function handler

src/handlers/chat.js:3133–4044

(res)

      'X-Accel-Buffering': 'no',
    },
    async handler(res) {
      const abortController = new AbortController();
      let unregisterSse = () => {};
      res.on('close', () => {
        if (!res.writableEnded) {
          log.info('Client disconnected mid-stream, aborting upstream');
          abortController.abort();
        }
      });
      const send = (data) => {
        if (!res.writableEnded) res.write(`data: ${JSON.stringify(data)}\n\n`);
      };
      unregisterSse = registerSseController({
        abort(reason) {
          send(chatStreamError(reason || 'server shutting down', 'server_error', 'server_shutdown'));
          if (!res.writableEnded) {
            res.write('data: [DONE]\n\n');
            res.end();
          }
          abortController.abort(reason);
        },
      });

      // SSE heartbeat: keep the TCP/HTTP connection alive through any silent
      // period (LS warmup, Cascade "thinking", queue wait). `:` prefix is a
      // comment line per the SSE spec — clients ignore it, intermediaries see
      // bytes flowing, idle timers get reset.
      const heartbeat = setInterval(() => {
        if (!res.writableEnded) res.write(': ping\n\n');
      }, HEARTBEAT_MS);
      const stopHeartbeat = () => clearInterval(heartbeat);
      res.on('close', stopHeartbeat);

      // ── Cache hit: replay stored response as a fake stream ──
      const cached = cacheGet(ckey);
      if (cached) {
        log.info(`Chat: cache HIT model=${model} flow=stream`);
        recordRequest(model, true, 0, null);
        try {
          send({ id, object: 'chat.completion.chunk', created, model,
            choices: [{ index: 0, delta: { role: 'assistant', content: '' }, finish_reason: null }] });
          if (cached.thinking) {
            send({ id, object: 'chat.completion.chunk', created, model,
              choices: [{ index: 0, delta: { reasoning_content: cached.thinking }, finish_reason: null }] });
          }
          if (cached.text) {
            send({ id, object: 'chat.completion.chunk', created, model,
              choices: [{ index: 0, delta: { content: cached.text }, finish_reason: null }] });
          }
          send({ id, object: 'chat.completion.chunk', created, model,
            choices: [{ index: 0, delta: {}, finish_reason: 'stop' }] });
          send({ id, object: 'chat.completion.chunk', created, model,
            choices: [], usage: cachedUsage(messages, cached.text) });
          if (!res.writableEnded) { res.write('data: [DONE]\n\n'); res.end(); }
        } finally {
          unregisterSse();
          stopHeartbeat();
        }