v0.0.1cerebra2026-06-04

Phase 1 — Source Memory Foundation

Implements the full source ingestion pipeline — file discovery, type detection, markdown and text parsing, heading-based chunking with five explicit strategies, batch SQLite persistence, and idempotent re-ingest.

v0.0.12026-06-04

·WAL mode + foreign keys on every SQLite connection via central connect() factory
·Migration002: sources, documents, chunks, memory_records tables (one migration per phase)
·Idempotency key: (canonical_path, content_hash, parser_version, chunker_version)
·Markdown adapter: stdlib-only, single-pass, code-block-aware, frontmatter, out-of-order warnings
·Chunker: 5 strategies (heading, sliding_window, code_block, code_block_oversized, mixed_overflow); stable chunk IDs
·Write-then-rename artifact storage; batch executemany() inserts per document
·cerebra ingest CLI: --dry-run, --exclude, --extensions, --vault, --json