AnswerDotAI / AnswerDotAI/fastllm
Empty markdown cell during time to first token in adisplay_stream
- Dominant language
- Jupyter Notebook
- Stars
- 15
- Forks
- 4
- Avg merge
- 1h 23m
- Merged PRs (30d)
- 12
Description
Hi,
in
```python
async def adisplay_stream(rs, chat=None, mx=2000):
"Use IPython.display to markdown display the response stream, refreshing from `chat.full()` on `Refresh`."
try: from IPython.display import display, Markdown
except ModuleNotFoundError: raise ModuleNotFoundError("This function requires ipython. Please run `pip install ipython` to use.")
acc,h = StreamAccum(chat, mx=mx), display(Markdown(' '), display_id=True)
async for o in rs:
if acc(o) and acc.txt: h.update(Markdown(acc.txt))
return acc.txt
```
h is initialized before the stream starts, leading to an empty Markdown cell being displayed while waiting for the first token.
If desired, this could be circumvented for example by initializing `h=None` at first, and then only creating h in the first loop iteration.
Contributor guide
No contributing guide indexed for this repository
Research direction
Start by locating the adisplay_stream entry point and reproduce a stream with a delayed first token. Adjust the initialization so no empty Markdown cell appears before content arrives, while streamed content still refreshes normally; verify the first visible output contains the response text.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- jupyter, python
- Domain
- frontend
- Issue type
- Bug
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 72/100