-
- [ ] User Interface
- [ ] Display a cursor on x-y coordinate plane
- [ ] Playback file browsing
- [ ] Settings activity
- [ ] Signal Processing
- [x] Access microphone / headset / …
-
---
_edited by agilgur5_: the symptoms of this issue are similar to issues in v3.5.3-v3.5.10, but are different. If you have similar symptoms in v3.5.x, please update to v3.5.11+. You likely had http…
-
> In this issue we are talking about looped network topologies rather than HOCs like `map/filter/reduce`
# Usecase
In "99 bottles" example we have to use explicit lock instead of deferred conne…
-
### Use Case
* Downloading a file from a bucket and saving it directly to disk
* Downloading a file from an HTTP response, and performing processing on its output
* Reading a file from disk as a …
-
### Your current environment
Below is my current Docker Compose configuration:
```yaml
services:
vllm:
image: vllm/vllm-openai:v0.6.4
deploy:
resources:
reservation…
-
Only workaround is to delete the data in the processor state table, presumably problem lies in mismatched sequence number in the recreated stream
-
It is within the realm of possibilities that the adapter solution we provide will be used by the community in the compute environment of Open Targets (Google cloud). Thus, we may benefit from an adapt…
-
**Describe the bug**
We have a service that proxies large streams to different backends. It's compiled with GraalVM native-image and is running in docker with a quite small heap ~80Mb. It has worked …
-
The goal would be to replace the current approach using multiprocessing and internal mechanisms by an Xarray interface. This would enable the possibility to use Dask, stream the processing, but also k…
-
Right now it looks like updates will pull many documents into memory at once. A heap dump shows 4000. Because updates pull in full documents, if these documents are large, it can quickly use up more t…