chore: release v0.14.1 to main

.cargo/mutants.toml

@@ -0,0 +1,42 @@
+# cargo-mutants configuration for uteke
+#
+# Usage (from workspace root):
+#   cargo mutants --file crates/uteke-core/src/jaccard.rs        # single file
+#   cargo mutants --output mutants-run.out                      # custom output dir
+#   cargo mutants -j 4 --timeout 60                              # parallel jobs + timeout
+#
+# This is a LOCAL-ONLY tool — not in default CI.
+# Pre-release validation: run manually before tagging a new version.
+# CI integration (if added): only on develop→main PRs as a pre-release gate.
+#
+# Exclude modules that require external services (SQLite DB, embedding API, etc.)
+# Focus mutation testing on pure logic — the code most prone to silent bugs.
+
+# Files to EXCLUDE from mutation testing.
+# These are excluded because they require I/O, external services, or are
+# thin wrappers around stdlib/third-party calls — not pure logic worth mutating.
+exclude_globs = [
+    # Embedding backends — require external API/network
+    "crates/uteke-core/src/embed/*.rs",
+    # SQLite-backed modules — require on-disk database
+    "crates/uteke-core/src/memory/*.rs",
+    # Import/export — I/O heavy
+    "crates/uteke-core/src/import_export.rs",
+    # Update check — network fetch
+    "crates/uteke-core/src/update_check.rs",
+    "crates/uteke-cli/src/commands/update_check.rs",
+]
+
+# Equivalent mutants — PROVEN equivalent (behavior-identical, unkillable by any
+# test). Do not count against the mutation score. Reviewed & verified:
+#   chunker.rs:282:24 (< → <=) — is_char_boundary(text.len()) is always true,
+#     so the loop-guard condition is identical either way.
+#   chunker.rs:425:25 (&& → ||) — invariant: in_block == !current_lines.is_empty()
+#     always holds when reached (set together at ~line 435), so both
+#     operands always have the same truth value.
+#   chunker.rs:443:17 (&& → ||) — same invariant as above, at the final flush.
+exclude_re = [
+    "chunker.rs:282:24: replace < with <=",
+    "chunker.rs:425:25: replace && with \\|\\|",
+    "chunker.rs:443:17: replace && with \\|\\|",
+]

.github/workflows/deploy-website.yml

@@ -26,18 +26,14 @@ jobs:
     steps:
       - uses: actions/checkout@v7
 
-      - uses: actions/setup-node@v7
-        with:
-          node-version: 24
-          cache: npm
-          cache-dependency-path: docs/package-lock.json
+      - uses: oven-sh/setup-bun@v2
 
       - name: Install dependencies
-        run: npm ci
+        run: bun install --frozen-lockfile
         working-directory: docs
 
       - name: Build
-        run: npm run build
+        run: bun run build
         working-directory: docs
 
       - name: Verify build output

.github/workflows/mutation-testing.yml

@@ -0,0 +1,58 @@
+# Mutation testing — pre-release validation only.
+#
+# This workflow runs ONLY on PRs from develop → main, serving as a pre-release
+# quality gate. It is NOT part of the default CI pipeline.
+#
+# Mutation testing validates test quality: cargo-mutants modifies source code
+# (operator mutations, boundary changes, etc.) and checks if existing tests
+# catch the mutation. "Missed mutants" = tests that pass on buggy code.
+#
+# Usage:
+#   - Runs automatically on develop → main PRs
+#   - Can be triggered manually via workflow_dispatch
+#   - Results are informational — they don't block merge by default
+#
+# To fix missed mutants, see the pattern in:
+#   crates/uteke-core/src/salience_recency.rs (mutation-killing test comments)
+#   Reference: nginjen project (78% missed-mutant reduction)
+
+name: Mutation Testing
+
+on:
+  pull_request:
+    branches: [main]
+  workflow_dispatch:
+
+# Cancel in-progress runs for the same PR
+concurrency:
+  group: mutation-testing-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  mutation-test:
+    name: Cargo Mutants
+    runs-on: ubuntu-latest
+    # Only run on PRs from develop (pre-release validation)
+    if: |
+      github.event_name == 'workflow_dispatch' ||
+      (github.event_name == 'pull_request' && startsWith(github.head_ref, 'develop'))
+    steps:
+      - uses: actions/checkout@v7
+
+      - uses: dtolnay/rust-toolchain@stable
+
+      - name: Install cargo-mutants
+        run: cargo install cargo-mutants --locked
+
+      - name: Run cargo-mutants on core logic
+        working-directory: crates/uteke-core
+        run: cargo mutants --timeout 120 -j 2
+        continue-on-error: true
+
+      - name: Upload mutant results
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: mutant-results
+          path: crates/uteke-core/mutants.out/
+          retention-days: 30

.github/workflows/release.yml

@@ -269,14 +269,23 @@ jobs:
         run: cargo publish --token ${{ secrets.CARGO_REGISTRY_TOKEN }} -p uteke-core --allow-dirty
         continue-on-error: true
 
+      - name: Wait for crates.io index propagation
+        run: sleep 60
+
       - name: Publish uteke-mcp
         run: cargo publish --token ${{ secrets.CARGO_REGISTRY_TOKEN }} -p uteke-mcp --allow-dirty
         continue-on-error: true
 
+      - name: Wait for crates.io index propagation
+        run: sleep 30
+
       - name: Publish uteke-cli
         run: cargo publish --token ${{ secrets.CARGO_REGISTRY_TOKEN }} -p uteke-cli --allow-dirty
         continue-on-error: true
 
+      - name: Wait for crates.io index propagation
+        run: sleep 30
+
       - name: Publish uteke-server
         run: cargo publish --token ${{ secrets.CARGO_REGISTRY_TOKEN }} -p uteke-server --allow-dirty
         continue-on-error: true

.gitignore

@@ -1,27 +1,9 @@
-/target/
-*.swp
-*.swo
-*~
-.DS_Store
-.idea/
-.vscode/
-*.pdb
-__pycache__/
-milestone*.md
-
-# Runtime index files
-uteke_index.keys
-uteke_index.usearch
-.fusion/
-progress.md
-.cora/
-scripts/demo-record.sh
-
-# Internal docs — never commit to public repo
-internal/
-
 # Benchmark results (large, ephemeral — commit RESULTS.md summaries, not raw data)
 benchmarks/longmemeval/results_*/
 
+# cargo-mutants output (local only — run before release)
+mutants*.out/
+mutants*.out.old/
+
 # VitePress build artifacts (keep config.ts, not lock files)
 docs/package-lock.json

CHANGELOG.md

@@ -1,3 +1,21 @@
+# Changelog
+
+## [0.14.1] — 2026-08-14
+
+Patch release: two chunker bug fixes found by mutation testing, plus test-suite hardening. No API or behavior changes beyond the fixes.
+
+### Fixed
+
+- **Markdown heading duplicated in oversized sections** — When a markdown section exceeded `max_chunk_chars`, its heading appeared twice in the first sub-chunk: once from `split_by_headings`, then again from the sub-chunking loop. Every oversized markdown document was quietly producing corrupted chunk content, which flows directly into embedding input and degrades retrieval quality. The dead re-prepend path is removed entirely (#1024).
+
+- **Infinite loop on multi-byte text with small chunk sizes** — The zero-progress guard in `split_long_text` advanced by raw byte offsets that could land inside multi-byte UTF-8 characters (CJK, emoji), flooring back to the start position forever. `chunk_markdown("日本語", 2)` would hang. The guard now advances to the next valid character boundary (#1024).
+
+### Changed
+
+- **cargo-mutants configuration moved to `.cargo/mutants.toml`** — The previous config at `crates/uteke-core/mutants.toml` used the key `exclude_files`, which is not a valid field in cargo-mutants v27 and was silently never applied. Migration to `.cargo/mutants.toml` (the location cargo-mutants actually reads) with the correct v27 schema (`exclude_globs`, `exclude_re` as arrays) and workspace-root-relative glob paths. This release is the first where mutant exclusions actually work.
+
+- **Chunker test suite grew from 20 to 60 tests** — Mutation score improved from 50% to 97% (149/164 mutants caught, 0 missed non-equivalent). 3 proven-equivalent mutants are excluded with documented reasoning.
+
 ## [0.14.0] — 2026-08-14
 
 Minor release: hybrid recall as default strategy, scene-segmented extraction, memory tools guide, lifecycle introspection, and source provenance. One behavior change (recall default).

Cargo.toml

@@ -9,7 +9,7 @@ members = [
 ]
 
 [workspace.package]
-version = "0.14.0"
+version = "0.14.1"
 edition = "2024"
 license = "Apache-2.0"
 repository = "https://github.com/codecoradev/uteke"

crates/uteke-cli/Cargo.toml

@@ -14,7 +14,7 @@ name = "uteke"
 path = "src/main.rs"
 
 [dependencies]
-uteke-core = { path = "../uteke-core", version = "0.14.0", features = ["onnx"] }
+uteke-core = { path = "../uteke-core", version = "0.14.1", features = ["onnx"] }
 clap = { version = "4", features = ["derive"] }
 serde_json = "1"
 tracing = "0.1"