sasipi247
https://developers.openai.com/api/docs/guides/websocket-mode
I have been building on it over the past month holding WebSocket sessions on workers warm, and command routing using NATS JetStream. With this, it has made using sidecar threads for a main thread very simple, as the worker treats them similar.
Reasoning deltas add additional traffic, especially if running many subagents etc. So on large scale, those deltas maybe are just dropped somewhere.
Saying that, sometimes the GPT reasoning summary is funny to read, in particular when it's working through a large task.
Also, the summaries can reveal real issues with logic in prompts and tool descriptions+configuration, so it allowing debugging.
i.e. "User asked me to do X, system instructions say do Y, tool says Z which is different to what everyone else wants. I am rather confused here! Lets just assume..."
It has previously allowed me to adjust prompts, etc.
But I think this is good thing the model checks the code, when adding new packages etc. Especially given that thousands of lines of code aren't even being read anymore.
Its like a microservices architecture with NATS JetStream coordinating stuff. I want to keep the worker core as clean as possible, just managing open sockets, threads and continuation.
Document querying is something I am interested in also. This system allows me to pin a document to a socket as a subagent, which is then called upon.
I have hit alot of slip ups along the way, such as infinite loops trying to call OpenAI API, etc ...
Example usage: 10 documents on warm sockets on GPT 5.4 nano. Then the main thread can call out to those other sockets to query the documents in parallel. It allows alot of possibilities : cheaper models for cheaper tasks, input caching and lower latency.
There is also a frontend
Alot of information is in here, just thoughts, designs etc: https://github.com/SamSam12121212/ExplorerPRO/tree/main/docs