Skip to content

Optimize lineage discovery and document loading #45

Description

@iam-brain

Question

Reduce the remaining cold lineage scan floor without weakening mutation detection or changing accounting totals.

PR #44 profiles the retained corpus at roughly 74 seconds for pass-one discovery and 74 seconds for pass-two document loading, while reconciliation is down to about 15 seconds. The parser still performs separate summary and evidence passes and rebuilds overlapping identity/model evidence.

Acceptance

  • profile signature hashing, JSONL parsing, metadata extraction, model tracking, and observation construction separately
  • preserve exact file-change rejection before/during pass one and pass two
  • avoid retaining full observations during discovery
  • demonstrate unchanged frozen UTC rows/totals
  • materially reduce discovery or load time on the retained corpus
  • keep cancellation responsive and memory bounded

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions