Flexible indexing management

[simulogreed]

Hello, I'm using Kilocode and wanted to index the project. The paid online versions work fine, but I'm having problems with the local versions. Indexing doesn't work at all for Qwen's models (qwen3-embedding:0.6b).

In log docker i see that

[GIN] 2025/11/18 - 13:19:39 | 500 | 59.992218622s |      172.17.0.1 | POST     "/api/embed"

[GIN] 2025/11/18 - 13:19:39 | 500 |  59.99983229s |      172.17.0.1 | POST     "/api/embed"

time=2025-11-18T13:19:39.638Z level=INFO source=server.go:1640 msg="aborting embedding request due to client closing the connection"

time=2025-11-18T13:19:39.638Z level=INFO source=server.go:1640 msg="aborting embedding request due to client closing the connection"

time=2025-11-18T13:19:39.638Z level=INFO source=server.go:1640 msg="aborting embedding request due to client closing the connection"

time=2025-11-18T13:19:39.638Z level=INFO source=server.go:1640 msg="aborting embedding request due to client closing the connection"

time=2025-11-18T13:19:39.638Z level=INFO source=server.go:1640 msg="aborting embedding request due to client closing the connection"

time=2025-11-18T13:19:39.638Z level=INFO source=server.go:1640 msg="aborting embedding request due to client closing the connection"

time=2025-11-18T13:19:39.638Z level=INFO source=server.go:1640 msg="aborting embedding request due to client closing the connection"

time=2025-11-18T13:19:39.638Z level=INFO source=server.go:1640 msg="aborting embedding request due to client closing the connection"

for nomic-embed-text it is work for 6000 from 24000 and then aborting again. The GPT chat suggested that this might be related to response time. Perhaps my hardware is too slow. Can you improve the plugin and add settings so I can index offline with slow hardware?

[kajiev]

@simulogreed
It works for 1024 as "model dimension". (which is default and max for 0.6b model)
Also search open issues on indexing - there are three or more issues with usefull discussion in comments.