v0.6.0

Breaking

• Make SearchQuery fields private and expose search options through SearchQuery::opts()
• Move search-domain types under the public sift_core::search module and reduce root re-exports to SearchError, SearchOutcome, and SearchQuery
• Change SearchExecution to borrow candidates instead of taking ownership
• Replace boolean grep/index modes with GrepMode, UnindexedStrategy, and IndexWalkConfig domain types
• Change CLI dispatch to consume Cli during routing

Performance

• Cache candidate file size and directory depth from walk/index metadata, reducing repeated filesystem metadata calls
• Skip total file byte accounting unless stats are requested

Fixes

• Use bitflags::set() in PatternCompiler so disabled flags are applied correctly
• Replace unchecked storage reads with explicit little-endian read helpers

Docs

• Revamp README installation and usage documentation
• Add benchmark chart generation and benchmark result images
• Add implementation and review plans for the architecture cleanup

Deps

• Bump memmap2 from 0.9.10 to 0.9.11
• Bump actions/checkout from 6 to 7

v0.5.0

Breaking

• Replace Daemon::send with Daemon::index; remove public DaemonOp export
• Daemon::serve takes ServeConfig instead of loose arguments

Features

• Ship sift-daemon in GitHub release assets and install.sh
• Collapse daemon into index/daemon/mod.rs with inline Daemon::serve event loop

Refactor

• Tighten daemon API: Daemon::bootstrap, Daemon::executable, domain methods on grouped serve types (#100)

v0.4.0

Breaking

• meta.json layout changed (corpus / walk / filters nesting). Re-run sift index build after upgrading; old stores are not migrated automatically.
• Daemon IPC types moved from sift_core::DaemonOp to sift_grep::index::daemon::{Daemon, DaemonOp, DaemonError}; core no longer exports daemon types.

Features

• Modular index daemon with async index update, --lazy build, and search-triggered background indexing
• MR-SW snapshot index store with unified reconcile and partial-path updates
• Walk unindexed corpus paths during daemon-enabled search for files not yet in the snapshot

Refactor

• Migrate grep/filter helpers to domain-type impls (PatternConfig::search_options, FilterConfig::candidate_config, OutputConfig::separators, ByteSize)
• Migrate core search helpers to domain types (CandidateFilter::collect, WalkOptions::discover_files, IgnoreConfig::matcher)
• Consolidate daemon into CLI-only index/daemon.rs (Daemon::send, Daemon::serve, Daemon::ensure_running); remove Serve, coordinator, and core IPC types
• Remove no-op DaemonOp::Watch IPC opcode; reload store metadata before daemon reconciles; fail serve on startup reconcile errors; rebind filesystem watcher when corpus root changes
• Unify TrigramIndex::build(config, dir, paths) API (empty paths = full corpus)

Documentation

• Add LICENSE-MIT, LICENSE-APACHE-2.0, CONTRIBUTING.md, SECURITY.md
• Add crate metadata and README release-scope section
• Expand rg compatibility matrix and integration test coverage

v0.3.0

Bug Fixes

• reconcile on startup to catch changes made while daemon was down (#71)
• configure WalkBuilder from VisibilityConfig in no-index search (#59)

Features

• add idle timeout to daemon + redesign coordinator state machine (#72)
• concurrent daemon and multi-reader-single-writer index store (#68)
• add QueryPlanner, clean up index API, split regression tests (#60)

Miscellaneous

• bump sift-core path dep during release

Performance

• 2-pass trigram-only radix sort for posting assembly (#83)
• use incremental update in CLI when index exists (#80)
• use thread-local bitset for trigram dedup in from_bytes (#81)
• stream-decode posting lists during intersection (#82)
• defer content-level validation on index open (#78)
• drop redundant result sort in scan workers (#65)
• skip path work when no ignore rules apply (#66)
• read files for trigram extraction instead of mmap (#67)

Refactor

• reorganize index and trigram modules by domain (#64)

v0.2.0

Bug Fixes

• prune gitignored directories during index build (#58)
• redesign daemon event loop with RefreshState, remove is_relevant_event (#57)
• optimize index build with varint postings and unified visibility (#54)
• eliminate #[allow], split benchmarks per module, add profiling (#22)
• rustfmt import formatting
• remove unused rel_match_context helper, use string literals
• Windows clippy, context prefix formatting, and expanded tests

Documentation

• rewrite READMEs and AGENTS.md for clarity and index generality (#32)
• mark no-op flags (line-buffered, block-buffered, mmap, no-mmap) explicitly
• rewrite READMEs and add AGENTS.md to all projects and modules (#18)
• update Linux benchsuite snapshot with fresh results and chart
• tighten agent notes — scannable layout, no policy essays
• branch-per-phase workflow before roadmap slices

Features

• redesign daemon architecture with explicit config, spawn lock, and --once mode (#56)
• unified Index trait, auto-init, incremental updates with fingerprint-based change detection (#31)
• modular public-API-only benchmarks (#26)
• comprehensive unit and integration test coverage for sift-core (#25)
• add --no-config, --unicode/--no-unicode, --colors, --regex-size-limit, --dfa-size-limit, -M/--max-columns, --max-columns-preview flags
• add --no-config, --unicode/--no-unicode, --colors, --regex-size-limit, --dfa-size-limit, -M/--max-columns, --max-columns-preview flags
• add -j/--threads, --line-buffered, --block-buffered, --path-separator, --one-file-system, -U/--multiline, --multiline-dotall, --crlf, --mmap/--no-mmap flags
• add -j/--threads, --line-buffered, --block-buffered, --path-separator, --one-file-system, -U/--multiline, --multiline-dotall, --crlf, --mmap/--no-mmap flags
• add -r/--replace, --trim, -b/--byte-offset, --passthru, --include-zero flags
• add -r/--replace, --trim, -b/--byte-offset, --passthru, --include-zero flags
• add --no-ignore-parent, --no-ignore-global, --no-ignore-exclude, --no-messages, --no-ignore-messages, --no-ignore-files, --ignore-file flags
• add --no-ignore-parent, --no-ignore-global, --no-ignore-exclude, --no-messages, --no-ignore-messages, --no-ignore-files, --ignore-file flags
• add -a/--text and --binary flags for binary file handling
• add -a/--text and --binary flags for binary file handling
• add filter flags for max-depth, max-filesize, types, iglob, files, sort
• add --max-depth, --max-filesize, --iglob, --ignore-file, --files, -t/--type, -T/--type-not, --type-list, --type-add, --type-clear, --sort/--sortr filter flags
• add --context-separator, --no-context-separator, --field-match-separator, --field-context-separator flags (#10)
• add --column, --vimgrep, --pretty, -N/--no-line-number, --version flags (#9)
• implement scope-based path display resolution
• --json JSON Lines output (ripgrep-compatible) (#6)
• bytes searched in SearchStats and --stats (#5)
• elapsed time in SearchStats and --stats output (#4)
• --stats and SearchStats counters (#3)
• --color, --null, grouped output structs (#2)
• context lines (-A/-B/-C) for standard search
• search parity — paths, ignores, follow, filter pipeline

Miscellaneous

• bump softprops/action-gh-release from 2 to 3 (#34)
• bump actions/upload-artifact from 4 to 7 (#35)
• bump actions/download-artifact from 4 to 8 (#36)
• add release infrastructure — changelog, release script, ARM64, checksums, dependabot (#33)
• remove sift-profile binary from core crate (#23)
• fix pre-existing clippy lints (map_or, is_ok_and, byte str literals) (#17)
• remove fff.nvim
• remove useless scripts
• remove unused #[allow(dead_code)] from rel_match_context

Performance

• optimized trigram index build with packed sort and codec removal (#55)
• parallel corpus walk with WalkBuilder::build_parallel() (#47)
• extract trigrams from raw bytes instead of lossy UTF-8 (#48)
• reduce PathBuf allocations in resolve_candidates (#49)
• parallelize save_to_dir index file writes (#50)
• avoid materializing Vec in all_file_ids (#51)
• sift-profile revamp, matcher/searcher caches, parallel and index tuning

Refactor

• split search and grep modules, add index intersection planning (#30)
• remove parallel threshold, always use Rayon (#28)
• split grep module into domain folders (#27)
• restructure core into index/, grep/, and query/ modules (#24)
• harden integration test suite with TestProject helper (#21)
• organize CLI into domain-oriented modules (#19)
• replace Option with ColumnAction enum for max_columns
• avoid needless String allocation and double trim_start()
• add doc comments to ignore-granular structs
• rename parse_filesize to parse_size_suffix for consistency

Testing

• add 200 inline unit tests, convert CLI to lib+bin layout (#20)

Deps

• bump clap from 4.6.0 to 4.6.1 (#37)
• bump serde_json from 1.0.149 to 1.0.150 (#38)
• bump rayon from 1.11.0 to 1.12.0 (#39)
• bump bitflags from 2.11.0 to 2.11.1 (#40)

v0.1.2

Full Changelog: v0.1.1...v0.1.2