v0.2.0

Query/index upgrade highlights

• Cost-aware candidate planning retains usable extension/prefix/trigram predicates for combined short-word and path queries.
• Early pruning of common postings reduces peak construction memory; long prefix candidates are verified without false negatives.
• Deterministic DSL results, MFT/USN and permission boundaries remain intact. No new Rust dependency or default full-text/semantic service.
• Local million-record probe: build time 11.93s -> 8.91s; peak working set 1,005,248,512 -> 258,981,888 bytes; eight complete ordered differential checks passed. Actual CLI growth: 3,072 bytes (0.13%). Synthetic measurements are not universal performance guarantees.
• Windows workspace, desktop and CodeForge Regression Gate passed before merge (#19); release packaging reruns format/clippy/tests/build.

Raw measurements and limitations: https://github.com/styayur/local-context-engine/blob/v0.2.0/docs/UPGRADE-2026-10.md

What's Changed

• test: stabilize named pipe handshake startup wait by @styayur in #12
• chore: onboard CodeForge v0.2 verification by @styayur in #13
• chore: add CodeForge regression-only CI by @styayur in #14
• Clarify README evidence and pending real desktop capture by @styayur in #15
• Reduce planner fallbacks and bound index construction memory by @styayur in #19

New Contributors

• @styayur made their first contribution in #12

Full Changelog: v0.1.0...v0.2.0