fix(ci): make sccache optional in Docker builds to survive GHA cache …

…outages

Release 0.14.0

This release introduces a unified search architecture alongside powerful AST-based structural code search and rewriting capabilities (1c33c77, 07fbd5f, 8abed69). Key improvements include branch-aware delta indexing, automatic repository indexing for MCP, and expanded support for high-performance embedding models (7c51599, a969090, 0955895). System reliability is further enhanced through strict structured LLM outputs and optimized batch processing with automated retries (fb0196a, d40f641, 5739cb1).

⚠️ **Important**: This release contains breaking changes that may require code updates.

- **search**: unify search and add tests `1c33c779`

- **embeddings**: add mpnet and jinabert qk embedding support `09558952`
- **server**: enhance MCP tool schemas and add new tools `6b7d4e65`
- **clear**: add commits and graphrag clear modes `e793d8d5`
- **grep**: add code rewrite with AST templates `07fbd5f2`
- **grep**: add AST structural code search `8abed697`
- **indexer**: add last segment matching for imports `ca7fcd10`
- **grep**: add AST-based structural code search `c15dcddf`
- **indexer**: add AST-based function call extraction `c698904d`
- **graphrag**: add branch-aware filtering and merging `09e6337c`
- **llm**: enable strict mode for structured output `fb0196a3`
- **graphrag**: validate description model structured output `e4344d71`
- **benchmark**: add retrieval quality benchmark infrastructure `e1095d2d`
- **indexer**: add retries and improve batching `d40f641c`
- **mcp**: add automatic repository indexing `a9690903`
- **git**: add branch-aware delta indexing `7c515998`
- **branding**: add project logo to readme `fb536f29`
- **assets**: add svg icon to readme header `7449dece`

- **mcp**: replace handlers module with rmcp transport `bb5f929b`
- **docs**: standardize LLM config and model format `0c25e456`
- **llm**: consolidate JSON completion methods `4f2a756d`
- **indexer**: move LLM structured output validation to index start `d5bce043`
- **graphrag**: use structured JSON output for LLM calls `25d596d9`
- **benchmark**: clean up config comments `ff24fb59`
- **mcp**: use is_some_and for task check `d1469db7`

- **graphrag**: detect partial indexing when relationships missing `e51fe5fb`
- **graphrag**: clarify architectural analysis JSON response format `d3f9fb96`
- **graphrag**: clarify JSON response format for file descriptions `872b8ca6`
- **clippy**: resolve lint warnings in grep and tests `c6f80587`
- **graphrag**: ensure deleted files don't break processing `0c105ffa`
- **docs**: swap code and docs search mode labels `ab4fbc2b`
- **graphrag**: add error context to batch retry logs `b1771ba4`
- **graphrag**: add retries for batch AI operations `5739cb1b`

- update recommended text embedding model to MiniLM `5d5e7cfb`
- add AI agent workflow docs and MCP setup `50448b99`
- **grep**: document structural rewrite capability `d62fb98b`
- fix anchor links in navigation bar `c521f047`
- **readme**: clarify structural code analysis approach `d48033ab`

5 maintenance, dependency, and tooling updates not listed individually.

fix(crate): prepare dependency tree for publication

- Reconcile version mismatches in dependency tree
- Remove obsolete pin-utils package
- Add required utf8_iter for icu components
- Reduce keyword count to meet crates.io requirements
- Bump indexmap dependency version
- Stabilize build environment for crate publication

Release 0.12.2

This release improves indexing reliability and overall system performance. Enhanced metadata handling ensures more consistent code analysis results, while dependency updates provide better stability and compatibility.

- **indexer**: atomic metadata storage after batch `d35ebfb2`

- **deps**: bump octolib to 0.11.0 and update related dependencies `61fb0cd7`
- **deps**: bump dependency versions in Cargo.lock `ec53cb04`
- bump Rust toolchain from 1.92.0 to 1.94.0 `2d40e7e6`

**Total commits**: 4 across 2 categories

🔧 **1** improvement - *Better performance & code quality*
🔄 **3** other changes - *Maintenance & tooling*

Release 0.12.1

This release fixes a search issue where queries could return no results when matching zero-vector embeddings.

- **graphrag**: resolve zero-vector search returning empty results `72ebbc7e`

**Total commits**: 1 across 1 categories

🐛 **1** bug fix - *Improved stability*

ci(release): publish crates without default features

ci(release): fix macOS runner build for x86_64 targets

Release 0.10.0

This release adds Java language support and enhances tokenization for improved semantic search (eae4107, 4f463f0). Several bug fixes improve system stability, including better metadata handling, dependency updates, and LSP initialization (1574ac9, 48978d2, 76e1410, 11fb216). Documentation and configuration guidance have also been updated for clearer usage and setup.

- **indexer**: add Java language support `eae4107f`
- **huggingface**: add BPE tokenizer support for RoBERTa-style models `4f463f07`

- **java**: merge single-line nodes and deduplicate symbols `94d9e139`
- **embedding, indexer**: fix code formatting and whitespace issues `a7823895`

- **indexer**: clarify git metadata storage condition message `1574ac91`
- **lancedb**: update dependency to resolve aws_credential error `48978d2a`
- **lsp**: resolve initialization and update issues with --with-lsp `76e1410e`
- **embedding**: update fastembed to 5.0.2 and ensure thread safety `11fb216b`

- **config**: update embedding models and vector index guidance `7a6264d6`
- **mcp**: clarify semantic search tool description for queries `2078ef8a`

- **deps**: upgrade dependencies to latest versions `deb779a7`

**Total commits**: 11 across 5 categories

✨ **2** new features - *Enhanced functionality*
🔧 **2** improvements - *Better performance & code quality*
🐛 **4** bug fixes - *Improved stability*
📚 **2** documentation updates - *Better developer experience*
🔄 **1** other change - *Maintenance & tooling*

Release 0.9.1

This release includes several bug fixes that enhance code indexing accuracy and stability, such as improved parsing of export statements, serialized indexing to prevent conflicts, refined method-level indexing for PHP, and better handling of code fences in markdown files (bd5bc26, ae71a02, 19acf3d, 9df2c44). Additionally, the initial release process now correctly uses the current version without requiring a manual bump (693e2fd).

- **svelte**: improve parsing of export statements in indexer `bd5bc26e`
- **lock**: serialize indexing with file-based lock `ae71a02e`
- **php**: index classes by methods instead of whole class `19acf3d4`
- **markdown**: prevent unrelated chunks from code fence splits `9df2c44d`
- **release**: use current version for initial release without bump `693e2fd7`

**Total commits**: 5 across 1 categories

🐛 **5** bug fixes - *Improved stability*

Release 0.9.0

This release adds Lua language support and enhances code indexing with enriched context for improved search accuracy (60eeadc, 8e741cd). Performance and responsiveness are improved through faster file counting and optimized memory handling (c891595, f9ca076). Several bug fixes address search threshold accuracy, code parsing, rendering issues, and stability across multiple components (9461b83, 0d04a53, 72f3141, b8e4eb4, fdd9e5d, bae5806, 0b927b2).

- **indexer**: enrich code block embeddings with file context `8e741cd4`
- **parser**: add Lua language support with tree-sitter `60eeadca`

- **indexer**: add fast file counting to optimize indexing performance `c891595a`
- **store**: unify line indexing to zero-based internally `1c282523`
- **cpp**: extract variable names from declarations alongside functions `bb941c5e`

- **graphrag**: load nodes from DB before relationship discovery `9461b83a`
- **search**: correct threshold handling for -t parameter `0d04a53c`
- **huggingface**: await async model init to prevent panic `72f31414`
- **cpp**: correct import and export extraction logic `b8e4eb43`
- **render**: correct line number display offset in outputs `fdd9e5d8`
- **cpp**: handle function declarations in indexing and signatures `bae5806d`
- **memory**: improve MCP responsiveness and indexing efficiency `f9ca0766`
- **diff_chunker**: prevent panic on UTF-8 boundary slicing `0b927b2a`

**Total commits**: 13 across 3 categories

✨ **2** new features - *Enhanced functionality*
🔧 **3** improvements - *Better performance & code quality*
🐛 **8** bug fixes - *Improved stability*