fix: correct use pending_requests instance in stdio transport

[hashemix]

📌 Summary

This PR fixes a bug where a wrong instance of pending_requests was used in stdio server transport, causing server initiated messages to timeout.

[NickNick]

Hey, I'm still getting a timeout on messages after updating to the version with this fix.

I have the following from the vscode logs (omitted the actual content for brevity):

2025-09-12 11:32:39.846 [debug] [server -> editor] {"id":0,"jsonrpc":"2.0","method":"sampling/createMessage","params":{"maxTokens":158,"messages":[{"content":{"text":"...","type":"text"},"role":"user"}],"systemPrompt":"..."}}
2025-09-12 11:32:41.023 [debug] [editor -> server] {"jsonrpc":"2.0","id":0,"result":{"model":"copilot/gpt-4.1","content":{"type":"text","text":"..."},"role":"assistant"}}

But this last message doesn't seem to 'arrive' in my tool, so to say. It's just stuck after the .await.

[hashemix]

Hey @NickNick , thanks for reporting this issue 👍

Yes, I noticed there's a scenario where a deadlock could occur, which causes that .await to hang. The issue has been fixed but hasn't been pushed or released yet.

I'm planning a major release this weekend, and I am pretty sure that update will resolve the issue.
I'll tag you here once it's out 🙂

[hashemix]

Hey @NickNick v0.7.0 has been released , it should resolve the issue you reported.

See the Upgrade to v0.7 section in the changelog for steps to update your project.