Thats is amazing, Thank You for update!
Hi - thread_z7sqfhtWNwAOdmr8SAViz56d had produced the issue until interrupted, but then I couldnāt reproduce it anymore at all.
Thats is amazing!!Thank You for update! Iād using assistants api since launched. I went from zero to tier 4 only using assistants api and the streaming feature will be fantastic to my goals <3
- Token usage.
this one was a pain for me and I am still working on a memory system for this. I currently have an ai system with long term memory, short and persistent memory too which works like its a base intelligent in that it recalls information from a data storage in away that I can control all context lengths to keep the whole system with in a token limit.
How I achieved this is though a memory cruncher ai that runs in the background updating my persistants, and the rest is all through crafted threads. I do not use the api threading to track context but instead build my own contextual understanding to pass allowing me to control the full data size passed to allow me to use maximum tokens. I also use dynamix token adjustments which have a base value that uses a len to calculate based on input crafted prompt up to a max value to help minimize and scale the requirements to allow simpler response to be smaller and faster. well not realtime because of processing delay each message from voice in though with actual understanding and voice back out with emotions is 10seconds. if vision is required it adds additional time for processing depending on which vision system its using and how much detail you want from the vision system. which can up it another 10+ seconds.
Wanted to share this as some of this may spark ideas 
16k is the model size I use because its the cheapest cost to run this ai work companion every day with true two way voice. id kill for a gpt 3 32k access haha. that would give me so much more. GPT4 makes the ai come of super smart but at a cost. opening up the ai system memory to use a full 128k of data with my system gives it years of insites because of the design. but costs for 128k on gpt4 running my ai for a day is out of what I call personal usage in that its the most cost effective to run gpt3 for 24/7 ai companion.
Dynamic memory building and scaling is key with summarization with understanding
Glad to see the update on streaming. Hopefully it rolls out soon!
1 Like
Yes, very much looking forward to using streaming with assistants! Letās go!!
What is the estimated launch date for the Assistants APIās transition from beta to the first official release?
4 Likes
Hi all, weāre excited to share that streaming is now available in the Assistants API! You can now build real-time experiences with tools like Code Interpreter, retrieval, and function calling. 
We also have helpers in the Python and Node SDKs that make it really easy to integrate. We would love for you to try them out and share feedback by creating an issue on the SDK repos.
Learn more in our updated quickstart: https://platform.openai.com/docs/assistants/overview/step-4-create-a-run?context=with-streaming.
23 Likes