Open
Conversation
…ll-resource paths
…round-trip in repo conversion test
…onversations.context blobs
…lly needs rewriting
…ages are included
…ning and continuing
…rs cannot strand legacy rows
…FSET to survive concurrent row churn
…ad of assuming non-null
… the migration module
…duce it (behaviour preserved)
…on_window and max_tokens
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR adds an immutable message history and deterministic summarization with sliding window eviction.
The immutable history is prepared for the request via "projection" concept which maps the history to the form that LLM should be seeing.
Currently, this is done deterministically: walk the history and perform summarization when compaction criteria are met.
Last N summarizations (by default, 2) stay in the chat, the rest are skipped: this prevents context bloat which is observed with the current "insert summaries endlessly" approach.