Bug: Gemini SSE streaming response not cleaned up on client disconnect, leaks server resources

[anshul23102]

Bug Report: Abandoned SSE Streams Keep Gemini API Call Running After Client Disconnects

Description

The ApexAIPage uses Server-Sent Events (SSE) to stream Gemini AI responses.
When the user navigates away or closes the page mid-stream, the browser
closes the SSE connection, but the Express server continues calling
stream.next() on the Gemini generateContentStream until the response
completes. On a busy server, many abandoned streams accumulate as live Gemini
API calls, consuming tokens and server memory unnecessarily.

Steps to Reproduce

1. Navigate to the AI Roadmap page and start generating a roadmap.
2. Immediately navigate away before the response finishes.
3. Monitor the Express server: Gemini API streaming calls continue for
   the duration of the expected response (30-60 seconds).
4. Repeat 10 times: observe 10 live Gemini calls consuming tokens concurrently.

Root Cause

The SSE route handler does not listen for the req.on('close', ...) event
to abort the Gemini streaming call when the client disconnects.

Impact

Abandoned streams consume Gemini API quota and server memory proportional to
the number of concurrent navigations-away. In production this can exhaust
the Gemini API rate limit for legitimate users.

Proposed Fix

Listen for client disconnect and abort the Gemini stream:

router.get("/api/ai/stream", async (req, res) => {
  res.setHeader("Content-Type", "text/event-stream");
  const controller = new AbortController();
  req.on("close", () => controller.abort());

  try {
    const stream = await model.generateContentStream(prompt, { signal: controller.signal });
    for await (const chunk of stream) {
      if (controller.signal.aborted) break;
      res.write(`data: ${JSON.stringify({ text: chunk.text() })}

`);
    }
  } catch (err) {
    if (err.name !== "AbortError") throw err;
  } finally {
    res.end();
  }
});

[anshul23102]

@kunalverma2512 Hi! I came across this issue and I would love to work on it as part of GSSoC '26.

I have reviewed the problem description carefully and am confident I can deliver a well-tested, production-ready fix within the contribution timeline. Please let me know if you need any additional context or have specific implementation preferences before I get started.

Could you please assign this issue to me?

/assign
/claim

[anshul23102]

@kunalverma2512

I'd like to work on fixing Gemini SSE streaming response cleanup on client disconnect as part of GSSoC 2026.

Implementation Details:

• Client disconnect detection
• Stream cleanup on disconnect
• Resource leak prevention
• Connection timeout handling
• Proper SSE event stream closure

Please assign this issue to me. For GSSoC 2026 tracking, could you also add the gssoc-approved label?

/assign
/claim