Week 4 Day 5: add token-aware context compaction - #170
Draft
skyzh wants to merge 1 commit into
Draft
Conversation
Assisted-by: Codex
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Day 4 sessions preserve the full transcript, but an unbounded replay eventually exceeds the model context and can feed oversized observations back into generation.
This checkpoint adds exact rendered-token admission, tool-specific observation bounds, immutable structured summaries, durable compaction markers, and safe cache reconciliation. Summary generation uses a separate cache, has one schema-checked attempt, and falls back deterministically without rewriting source events. Replayed markers are validated for chronology, complete-turn coverage, and structured provenance before they can hide history.
The CLI also rejects generation budgets that exceed its reserved context and records context/cache metrics without treating model completion as task correctness.
AI-assisted: Codex.