Repository navigation
Publish indexed evidence memory and current LLVM measurements - #5
Merged
Merged
Conversation
Separate recorded-data freshness from current-work applicability. Preserve native tool results and their certificates across closure, and publish actual changed-file contents as independent, resumable background observations. Unavailable paths cannot authorize absence or block native-result saving. Keep source refresh notices file-specific, rank graph context against the current request, bound neighborhood growth, and respect the hook deadline. Prove that freshness is accepted exactly for valid matching dependencies; cover changed files, restored bytes, interrupted scans and immutable history. Validation: Lean build, 57 native/handoff test groups, and an actual LLVM source edit/search/restore through production hooks. The edit preflight uses no model calls and leaves the LLVM source unchanged.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
New chats now reuse selected original evidence from indexed local memory without reconstructing the full graph. This integrates the page database, persistent dependency and ranking reuse, source freshness checks that preserve immutable observations, and question-focused graph delivery with duplicate suppression.
The README documents the current architecture and the native GPT-6 Luna LLVM comparison: three fresh references and two reuse laps with existing seed/topic history. The reuse laps consumed 124,529 and 206,265 input-plus-output tokens versus 4,387,133 fresh tokens (97.2% and 95.3% reduction). History creation is separate; answer review and hook timings are scoped in the public measurement record.
Keep version 0.1.0. The manual packaging workflow builds and tests all four platforms before replacing standalone archives, checksums, and the plugin ZIP. Content-addressed runtime archives remain pinned, and the published runtime.json identifies the exact source commit. Package descriptions match indexed memory and the independent harness adapters.
Linux DB creation now uses atomic no-replace directory publication, matching the existing macOS boundary. Clean-runner workflows prepare the pinned numerical runtime/model before engine tests. Lifecycle and adapter checks inspect committed DB roots and actual delivered evidence rather than the removed single-file representation or a mandatory repeated tool denial.
Validation: native Lean build; core regressions, hook lifecycle and setup/package checks; workflow YAML and shell syntax; embedded/checked-in manifest equality. GitHub CI additionally checks Linux/macOS engine and adapter contracts before merge. No new LLVM runtime execution or broader accuracy claim is implied.