Feat/v050 phase 1 retrieval#5
Merged
Merged
Conversation
Adds two new functions to db.rs: - auto_repair(path, timestamp): dumps all data tables from a corrupted DB via SELECT *, rebuilds a fresh DB with initialize_schema(), repopulates FTS, verifies integrity, then atomically renames old -> .corrupt.<ts> and new -> original path. Never deletes the corrupted original. - quick_check(conn): fast PRAGMA quick_check wrapper for runtime B-tree checks. Includes unit tests for both: test_quick_check_clean_db and test_auto_repair_recovers_data (writes then corrupts a temp DB file).
state.rs: - Adds db_corrupted: Arc<AtomicBool> field to RuntimeState. - initialize(): runs PRAGMA integrity_check before accepting requests, logs "[cortex] DB integrity: OK" on clean, calls auto_repair() on corruption. On repair success, reopens and continues normal startup. On repair failure, continues in degraded mode with db_corrupted=true. main.rs: - Adds 30-minute background task that runs quick_check() on db_read. Sets db_corrupted=true on failure with prominent stderr warning. Clears the flag if a subsequent check passes (post-manual repair). handlers/health.rs: - Exposes db_corrupted in /health response. - Sets status="degraded" and degraded=true when db_corrupted is set.
…rolling backups Phase 5C: Crash-Safe WAL Handling - WAL checkpoint interval reduced from 60s to 10s (limits data loss to <10s on crash) - WAL recovery on startup via PRAGMA wal_checkpoint(TRUNCATE) - Added documentation comment explaining why synchronous=NORMAL is safe with WAL Phase 5B: Rolling Backups - Daily automatic backup in WAL checkpoint loop (checks >24h since last backup) - Backup rotation: keep max 7 daily backups, remove oldest - \cortex backup\ CLI command for manual backup with forced WAL checkpoint - \cortex restore <file>\ CLI command with: - Pre-restore backup (preserves current DB as backup) - Integrity verification before accepting restore - Automatic rollback if integrity check fails - Safety checks for running daemon All changes tested: cargo test passes (67 tests), cargo clippy passes. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Implements rrf_fuse(lists, k) -- Reciprocal Rank Fusion per Cormack et al. Score = sum(1/(k+rank+1)) across lists, k=60.0. Returns descending by fused score. Five unit tests covering: single list, two-list agreement, middle promotion, empty input, empty list. Wire-up deferred to Task 1.5 (pipeline integration).
…1.4) Phase 1 Task 1.4: Compound Scoring Function Added three helper functions to calculate compound scores combining RRF rank, importance, and recency: - days_since(created_at): Calculate elapsed days from ISO 8601 timestamp, handling invalid timestamps gracefully (returns MAX) - normalize(importance): Normalize DB score field (typically 0-100) to 0.0-1.0 range with clamping - compound_score(rrf, importance, created_at): Compound formula = rrf * 0.6 + importance_norm * 0.2 + recency * 0.2 Recency uses 21-day half-life: exp(-days/30) All functions marked with #[allow(dead_code)] -- wire-up into pipeline is Task 1.5's responsibility. Added 3 unit tests: - test_days_since: Validates day calculations for today, yesterday, and invalid dates - test_normalize: Verifies normalization and clamping for 0-100 range and edge cases - test_compound_score: Tests compound score calculations with high/low RRF, importance, and recency All 81 tests pass. Zero clippy warnings. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
…ound scoring (task 1.5)
AdityaVG13
added a commit
that referenced
this pull request
Apr 24, 2026
Scoped low-risk deletion of two modules with zero call sites. Confirmed unreferenced across the full workspace (grep on crate::logging, crate::mcp_stdio, logging::, mcp_stdio::, log_line, log_path — each returned only self-references inside the files themselves). Per scope-lock.md Question #5 rules: - Cap ~300 LOC (this pass: 166 LOC removed, well under) - Per-file risk annotation (see table below) - No high-risk deletes (both classified low-risk) - Separate commit (this one) Per-file risk table: daemon-rs/src/logging.rs deleted LOW risk 27 LOC. Two helpers (log_path, log_line), both behind #[allow(dead_code)]. Never called. Superseded by tracing/eprintln diagnostic paths used throughout the daemon. Git history preserves if a file-based audit log is ever reintroduced. daemon-rs/src/mcp_stdio.rs deleted LOW risk 139 LOC. Stdio-transport MCP runner. Module was declared with #[allow(dead_code)] guarding the whole module and no external import. Superseded by the axum HTTP MCP path (handlers/mcp.rs::handle_mcp_message_with_caller) called from server.rs. If a future release wants a pure-stdio MCP binary, recover from git and wire it as a separate bin target. daemon-rs/src/main.rs edited LOW risk Removed `mod logging;`, `mod mcp_stdio;`, and the `#[allow(dead_code)]` attr that was masking mcp_stdio's dead state. Intentional #[allow(dead_code)] left in place (false positives per Repowise classification — kept for discoverability or future use): - auth.rs:31,383 — path-resolver internals - state.rs:42,236,252,283 — event payload, mcp_sessions cache, db_path, write_buffer_path (each with inline comment explaining why retained) - Struct-field annotations in conflict.rs, crystallize.rs, db.rs, diary.rs, feedback.rs, indexer.rs, recall.rs, setup.rs — all correspond to schema fields currently read through generic queries Validation: cargo check --manifest-path daemon-rs/Cargo.toml # clean cargo clippy --manifest-path daemon-rs/Cargo.toml --all-targets -- -D warnings # clean cargo test --manifest-path daemon-rs/Cargo.toml # 457/457 (434 unit + 5 doc + 11 integration + 7 recall-benchmark)
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.
No description provided.