v0.0.8

🚀 New Features & Capabilities

Knowledge Compilation Module (Beta) (#160)

• Hierarchical Indexing: Introduced an offline document processing module that structures data into hierarchical tree indices and knowledge clusters to boost retrieval precision.
• New CLI Entry Point: Added a dedicated command-line interface to trigger the compilation process.
• Search Pipeline Integration: Seamlessly integrated the generated index artifacts and clusters into the existing search workflow.

LLM Upgrades (#167)

• MiniMax-M3 Default: Upgraded the suggested default model to MiniMax-M3 (featuring a 512K context window, 128K max output, and multimodal support).
• Model Lineup Cleanup: Deprecated and removed references to MiniMax-M2.5 while keeping MiniMax-M2.7 variants as available alternatives.

---

🛠️ Improvements & Optimizations

• System Configuration & Linting: Added a system health check utility and removed hardcoded model names/processing limits to allow for dynamic runtime configuration. (#160)
• I/O Performance: Implemented file hash reuse throughout the data pipeline to eliminate redundant I/O operations. (#160)
• Documentation & Versioning: Bumped Docker image tags to version 0.0.7 across all English/Chinese READMEs, Docker docs, and history tables, alongside minor markdown formatting fixes. (#162, #167)

---

🐛 Bug Fixes

• Cross-Reference Tuning: Fixed performance bottlenecks encountered during the cross-reference building phase. (#160)

What's Changed

• update readme for docker version by @wangxingjun778 in #162
• [Feature] Introduce Knowledge Compiling Module by @wangxingjun778 in #160
• feat: upgrade MiniMax default model to M3 by @octo-patch in #167

Full Changelog: v0.0.7post1...v0.0.8