Add support for queue prioritization or per-index queues #662
westdavidr
started this conversation in
2. Feature requests
Replies: 1 comment
-
You should be able to do this today with some configuration, without code changes. During the ingestion you can provide a list of steps, which is a list of strings. Each string maps to a handler and to a queue. Via configuration, you should try registering the same handler with multiple names, so to have different queues. For instance, this is the default config: "extract": { "Assembly": "Microsoft.KernelMemory.Core.dll", "Class": "Microsoft.KernelMemory.Handlers.TextExtractionHandler" },
"partition": { "Assembly": "Microsoft.KernelMemory.Core.dll", "Class": "Microsoft.KernelMemory.Handlers.TextPartitioningHandler" },
"gen_embeddings": { "Assembly": "Microsoft.KernelMemory.Core.dll", "Class":"Microsoft.KernelMemory.Handlers.GenerateEmbeddingsHandler" },
"save_records": {"Assembly": "Microsoft.KernelMemory.Core.dll","Class": "Microsoft.KernelMemory.Handlers.SaveRecordsHandler"},
"summarize": {"Assembly": "Microsoft.KernelMemory.Core.dll","Class": "Microsoft.KernelMemory.Handlers.SummarizationHandler"}, and you could change it to: "chat-extract": { "Assembly": "Microsoft.KernelMemory.Core.dll", "Class": "Microsoft.KernelMemory.Handlers.TextExtractionHandler" },
"doc-extract": { "Assembly": "Microsoft.KernelMemory.Core.dll", "Class": "Microsoft.KernelMemory.Handlers.TextExtractionHandler" },
"email-extract": { "Assembly": "Microsoft.KernelMemory.Core.dll", "Class": "Microsoft.KernelMemory.Handlers.TextExtractionHandler" },
etc. then set |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
In our implementation of Kernel Memory, we generate numerous indexes from various sources such as SharePoint, Azure DevOps, web crawling, and in-chat documents.
We are facing challenges in sharing the queues for all these indexes, especially when they are under heavy load during synchronization tasks.
To address this, I propose adding a "PrefixConfiguration" to the queue options. This would allow us to set specific prefixes for different categories. For example, we could set a prefix for the chat documents index (chatdocs) to prepend the queue names with
chat-
, resulting in queue names likechat-extract
,chat-gen_embeddings
, etc.What are your thoughts on this approach?
Beta Was this translation helpful? Give feedback.
All reactions