Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Browse files
Browse the repository at this point in the history
# Backport This will backport the following commits from `main` to `8.12`: - [[Obs AI Assistant] ES|QL improvements (#173303)](#173303) <!--- Backport version: 8.9.7 --> ### Questions ? Please refer to the [Backport tool documentation](https://github.com/sqren/backport) <!--BACKPORT [{"author":{"name":"Dario Gieselaar","email":"[email protected]"},"sourceCommit":{"committedDate":"2023-12-17T21:23:50Z","message":"[Obs AI Assistant] ES|QL improvements (#173303)\n\nUse a combination of system prompt + recall for ES|QL to allow for:\r\n\r\n- faster responses\r\n- wider support of functions (e.g. any documented function is now\r\nsupported, in theory)\r\n\r\nPerformance for the current ES|QL scenarios is similar.\r\n\r\nSome things I've noticed:\r\n- ~for whatever reason it is much harder to get the LLM to not make\r\ncertain mistakes. I've tried mentioning some mistakes close to the end\r\nof the conversation, but will explore other options.~ I am again opting\r\nfor repeating common mistakes at the start of the message, but this time\r\nI'm allowing the LLM to pick them.\r\n- if we can prevent it from making \"known\" mistakes, we can figure out a\r\nway to have the LLM store known mistakes. This is also useful for other\r\nfunctions\r\n- another thing I want to figure out is if more examples improves for\r\ncommands and custom documentation improves performance. but later.\r\n\r\n---------\r\n\r\nCo-authored-by: kibanamachine <[email protected]>","sha":"d54109d8f8243cae00f6871c73d3bdc19f4f9905","branchLabelMapping":{"^v8.13.0$":"main","^v(\\d+).(\\d+).\\d+$":"$1.$2"}},"sourcePullRequest":{"labels":["release_note:fix","v8.12.0","v8.12.1","v8.13.0"],"number":173303,"url":"https://github.com/elastic/kibana/pull/173303","mergeCommit":{"message":"[Obs AI Assistant] ES|QL improvements (#173303)\n\nUse a combination of system prompt + recall for ES|QL to allow for:\r\n\r\n- faster responses\r\n- wider support of functions (e.g. any documented function is now\r\nsupported, in theory)\r\n\r\nPerformance for the current ES|QL scenarios is similar.\r\n\r\nSome things I've noticed:\r\n- ~for whatever reason it is much harder to get the LLM to not make\r\ncertain mistakes. I've tried mentioning some mistakes close to the end\r\nof the conversation, but will explore other options.~ I am again opting\r\nfor repeating common mistakes at the start of the message, but this time\r\nI'm allowing the LLM to pick them.\r\n- if we can prevent it from making \"known\" mistakes, we can figure out a\r\nway to have the LLM store known mistakes. This is also useful for other\r\nfunctions\r\n- another thing I want to figure out is if more examples improves for\r\ncommands and custom documentation improves performance. but later.\r\n\r\n---------\r\n\r\nCo-authored-by: kibanamachine <[email protected]>","sha":"d54109d8f8243cae00f6871c73d3bdc19f4f9905"}},"sourceBranch":"main","suggestedTargetBranches":["8.12"],"targetPullRequestStates":[{"branch":"8.12","label":"v8.12.0","labelRegex":"^v(\\d+).(\\d+).\\d+$","isSourceBranch":false,"state":"NOT_CREATED"},{"branch":"main","label":"v8.13.0","labelRegex":"^v8.13.0$","isSourceBranch":true,"state":"MERGED","url":"https://github.com/elastic/kibana/pull/173303","number":173303,"mergeCommit":{"message":"[Obs AI Assistant] ES|QL improvements (#173303)\n\nUse a combination of system prompt + recall for ES|QL to allow for:\r\n\r\n- faster responses\r\n- wider support of functions (e.g. any documented function is now\r\nsupported, in theory)\r\n\r\nPerformance for the current ES|QL scenarios is similar.\r\n\r\nSome things I've noticed:\r\n- ~for whatever reason it is much harder to get the LLM to not make\r\ncertain mistakes. I've tried mentioning some mistakes close to the end\r\nof the conversation, but will explore other options.~ I am again opting\r\nfor repeating common mistakes at the start of the message, but this time\r\nI'm allowing the LLM to pick them.\r\n- if we can prevent it from making \"known\" mistakes, we can figure out a\r\nway to have the LLM store known mistakes. This is also useful for other\r\nfunctions\r\n- another thing I want to figure out is if more examples improves for\r\ncommands and custom documentation improves performance. but later.\r\n\r\n---------\r\n\r\nCo-authored-by: kibanamachine <[email protected]>","sha":"d54109d8f8243cae00f6871c73d3bdc19f4f9905"}}]}] BACKPORT--> Co-authored-by: Dario Gieselaar <[email protected]>
- Loading branch information