Alan-Jowett / LexonGraph / 27075521704

81%

push

github

Add streaming indexer crate (#60)

* Add streaming indexer crate

Requirements: add a parallel streaming indexer line alongside the existing batch-oriented indexer while keeping the indexing protocol strategy-neutral.

Specs: add the rust-streaming-indexer-crate package and clarify docs/protocol/indexing.md so replay-based multi-pass planning is conformant.

Implementation: add lexongraph-streaming-indexer with replay-driven pass training, final materialization replay, shared streaming clustering integration, and executable validation/conformance tests.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Address PR review feedback

Compact the replay baseline into deterministic hashes, avoid cloning the baseline during finalization, and tighten the validation tests for repository path resolution and duplicate child detection.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Refine streaming indexer review fixes

Clarify factory timing documentation, emit full pass status transitions for finish_pass clustering work, correct the deterministic tie-break comment, and use a genuinely distinct resolver implementation in the VAL-STREAM-INDEXER-005 test.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Address additional streaming indexer review feedback

Clarify the streaming clustering factory fallback docs, make phase elapsed timing consistent, enforce store integrity checks for leaf and branch persistence, narrow the block-size validation to branch nodes, and add regression coverage for the new guarantees.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Tighten streaming indexer review follow-ups

Handle large derived cluster counts without silent truncation, add regression coverage for the overflow path, and correct the singleton-group tie-break documentation. Keep the classifier trait import because it is required for method resolution.

Co-authored-by: Copilot... (continued)

1191 of 1560 new or added lines in 2 files covered. (76.35%)

7522 of 8986 relevant lines covered (83.71%)

229.05 hits per line