Summary

KiwiFS addresses the need for persistent, agent-native knowledge filesystems by treating plain Markdown files as the single source of truth. It targets AI agent builders and teams seeking a self-hosted, versioned wiki alternative to proprietary SaaS solutions, offering agents a filesystem interface and humans a rich web UI.

How It Works

KiwiFS centers on the principle that files are the ultimate truth, deriving all functionality from them. It leverages Git for robust versioning, audit trails, and crash recovery, while SQLite FTS5 and pluggable vector stores provide tiered search capabilities (full-text and semantic). Structured queries are enabled via frontmatter, accessible through a DataView Query Language (DQL). This approach ensures agent-native access via standard filesystem operations (cat, grep) and a human-friendly web UI, all without vendor lock-in.

Quick Start & Requirements

Installation is streamlined via a one-line curl script or Docker. Building from source requires Go 1.25+ and Node.js 20+. Initialization involves kiwifs init followed by kiwifs serve. Official documentation, FAQs, and a roadmap are available.

Highlighted Details

• Unified Interface: Agents interact via filesystem mounts, REST API, or MCP; humans use an embedded web UI featuring Obsidian-like wiki links, backlinks, graph views, and a Notion-style editor.
• Multi-Protocol Access: Supports REST API, NFS, S3, WebDAV, and FUSE mounts, allowing diverse integration scenarios.
• Tiered Search: Offers grep (basic), SQLite FTS5 (BM25 ranked, default), and pluggable vector search for comprehensive querying.
• Git Versioning: Every write is an atomic Git commit, providing crash recovery, immutable audit trails, blame, and point-in-time restore.
• Structured Data & Querying: Frontmatter is mirrored to SQLite for structured queries via DQL, supporting complex data manipulation and computed fields.
• Data Management: Bulk import from 18 sources (databases, CSV, Notion, etc.) and export to JSONL/CSV, including embeddings.
• Knowledge Health: Built-in analytics and janitor scans for stale pages, orphans, broken links, and contradictions.
• Provenance Tracking: X-Actor and X-Provenance headers enable lineage tracking for agent-generated content.

Maintenance & Community

The README does not explicitly detail community channels (e.g., Discord, Slack) or notable contributors/sponsorships.

Licensing & Compatibility

KiwiFS is licensed under the Business Source License 1.1 (BSL 1.1), permitting free use, self-hosting, and modification. Offering KiwiFS as a commercial hosted service is restricted; a commercial license is required. The license converts to Apache 2.0 after four years.

Limitations & Caveats

Intentionally unsupported POSIX features include hard links, chmod/chown, POSIX ACLs, and extended attributes, as these conflict with the Git-centric model and API-level access control. Concurrency relies on optimistic locking via ETags and serialized writes; distributed locking is not provided.