WikiLoop v0.4.5

What's Changed

🐛 Bug Fixes

• fix(mcp): wire api_key auth in Start(), add x-api-key to CORS headers by @jasen215 in #29

📖 Documentation

• docs: add VitePress site for jasen215.github.io/wikiloop by @jasen215 in #30
• docs: add ecosystem pages and Google Analytics by @jasen215 in #31
• docs: add key technologies page for RAG and LLM Wiki paradigms by @jasen215 in #32
• docs: add OKF section to LLM Wiki technologies by @jasen215 in #33
• docs: add industry use cases and WikiLoop in industry pages by @jasen215 in #34
• docs: add kb_add tool to README and site documentation by @jasen215 in #35
• docs: add kb_add to all i18n README files by @jasen215 in #36
• docs: add kb_add to site index, what-is, and llm-wiki-technologies pages by @jasen215 in #37

⚙️ Maintenance

• ci: auto-label PRs by title prefix, categorize release notes by @jasen215 in #28

Full Changelog: v0.4.4...v0.4.5

WikiLoop v0.4.4

What's Changed

• fix(ci): auto-generate release notes from PR titles by @jasen215 in #19
• feat(convert): extract embedded Excel from docx/pptx and inject as markdown table by @jasen215 in #20
• fix(distill): preserve tables in technical spec docs via table-density detection by @jasen215 in #21
• docs: add HarmonyOS PC community build instructions by @jasen215 in #22
• docs: remove Claude Code config from hosted env section by @jasen215 in #23
• fix(distill): delete queue row on completion instead of marking done by @jasen215 in #24
• chore: rename release artifact to macos-arm64, drop Intel macOS build by @jasen215 in #25
• feat(i18n): add 8-language support with external JSON files by @jasen215 in #26
• chore: stop tracking eval/ directory by @jasen215 in #27

Full Changelog: v0.4.3...v0.4.4

WikiLoop v0.4.3

Features

• Red-links detection: kb_lint now detects and cleans broken related_to/supports/contradicts links from the DB. Concept-name gaps are written to wiki/index/red_links.json and shown as a knowledge-gap panel in WebUI — click "忽略" to dismiss individual entries.
• Pure Go SQLite (modernc.org/sqlite): replaces mattn/go-sqlite3 (CGO). Fixes Windows sqlite at runtime (CGO_ENABLED=0 now fully works). Binary size reduced 27%: 19.8MB → 14.4MB.

Fixes

• Distill prompt now constrains related_to/supports/contradicts fields to existing wiki paths only, preventing broken links from re-accumulating after reindex.
• WebUI files table: wider Modified column, right-aligned Size column.
• Build script: removed obsolete libtokenizers and libonnxruntime downloads (vector module was removed in v0.4.x).

WikiLoop v0.4.2

What's New

• Knowledge search engine for agents — kb_search returns layered results with related field for graph navigation
• kb_page tool — deep-read full document content by ID
• Removed vector search — pure FTS (SQLite FTS5 + BM25), no embedding model required; binary ~20MB
• Windows support — pure Go build, no CGO dependencies
• Lark/Feishu import — wikiloop import-lark <URL> imports Wiki pages and embedded tables
• synthesized pages _draft/ threshold — single-source pages go to _draft/, graduate when ≥2 sources
• synthesizedBoost — multiplicative 1.3× boost for concept/comparison/decision pages

Installation

macOS Apple Silicon

1. Download WikiLoop-0.4.2-darwin-arm64.dmg
2. Open the DMG and drag WikiLoop to Applications

Linux (x86_64 / ARM64)

tar -xzf wikiloop-0.4.2-linux-<arch>.tar.gz
sudo ln -sf $(pwd)/wikiloop /usr/local/bin/wikiloop

Windows (x86_64)

Expand-Archive wikiloop-0.4.2-windows-amd64.zip
# Add wikiloop.exe to PATH

See README for full setup instructions.

WikiLoop v0.4.1

What's New

• Knowledge search engine for agents — kb_search returns layered results with related field for graph navigation
• kb_page tool — deep-read full document content by ID
• Removed vector search — pure FTS (SQLite FTS5 + BM25), no embedding model required; binary ~20MB
• Windows support — pure Go build, no CGO dependencies
• Lark/Feishu import — wikiloop import-lark <URL> imports Wiki pages and embedded tables
• synthesized pages _draft/ threshold — single-source pages go to _draft/, graduate when ≥2 sources
• synthesizedBoost — multiplicative 1.3× boost for concept/comparison/decision pages

Installation

macOS Apple Silicon

1. Download WikiLoop-0.4.1-darwin-arm64.dmg
2. Open the DMG and drag WikiLoop to Applications

Linux (x86_64 / ARM64)

tar -xzf wikiloop-0.4.1-linux-<arch>.tar.gz
sudo ln -sf $(pwd)/wikiloop /usr/local/bin/wikiloop

Windows (x86_64)

Expand-Archive wikiloop-0.4.1-windows-amd64.zip
# Add wikiloop.exe to PATH

See README for full setup instructions.

WikiLoop v0.4.0

What's New

• Knowledge search engine for agents — kb_search returns layered results with related field for graph navigation
• kb_page tool — deep-read full document content by ID
• Removed vector search — pure FTS (SQLite FTS5 + BM25), no embedding model required; binary ~20MB
• Windows support — pure Go build, no CGO dependencies
• Lark/Feishu import — wikiloop import-lark <URL> imports Wiki pages and embedded tables
• synthesized pages _draft/ threshold — single-source pages go to _draft/, graduate when ≥2 sources
• synthesizedBoost — multiplicative 1.3× boost for concept/comparison/decision pages

Installation

macOS Apple Silicon

1. Download WikiLoop-0.4.0-darwin-arm64.dmg
2. Open the DMG and drag WikiLoop to Applications

Linux (x86_64 / ARM64)

tar -xzf wikiloop-0.4.0-linux-<arch>.tar.gz
sudo ln -sf $(pwd)/wikiloop /usr/local/bin/wikiloop

Windows (x86_64)

Expand-Archive wikiloop-0.4.0-windows-amd64.zip
# Add wikiloop.exe to PATH

See README for full setup instructions.

WikiLoop v0.3.2

Installation

macOS Apple Silicon

1. Download WikiLoop-0.3.2-darwin-arm64.dmg
2. Open the DMG and drag WikiLoop to Applications

Linux (x86_64 / ARM64)

tar -xzf wikiloop-0.3.2-linux-<arch>.tar.gz
sudo ln -sf $(pwd)/wikiloop /usr/local/bin/wikiloop

Embedding Model

Download separately and place in <WIKILOOP_KB>/models/bge-small-zh/:

• model.onnx
• tokenizer.json
• meta.json

See README for full setup instructions.

WikiLoop v0.1.0

Installation

macOS (Apple Silicon)

1. Download WikiLoop-0.1.0-darwin-arm64.dmg
2. Open the DMG and drag WikiLoop to Applications
3. Download the embedding model (see below) and place in ~/wikiloop-kb/models/bge-small-zh/

Linux (x86_64 / ARM64)

tar -xzf wikiloop-0.1.0-linux-<arch>.tar.gz
sudo ln -sf $(pwd)/wikiloop /usr/local/bin/wikiloop

Embedding Model

Download separately and place in <WIKILOOP_KB>/models/bge-small-zh/:

• model.onnx
• tokenizer.json
• meta.json

See README for full setup instructions.

Embedding Models v1

Embedding Model: bge-small-zh

Pre-packaged ONNX embedding model for WikiLoop vector search.

Installation

Extract to your KB's models directory:

tar -xzf bge-small-zh.tar.gz -C $WIKILOOP_KB/models/

This creates $WIKILOOP_KB/models/bge-small-zh/ with the required files.

Model Info

• Model: BAAI/bge-small-zh
• Format: ONNX
• Dimensions: 384
• License: MIT