Roadmap

Seek.js is in an experimental phase: APIs and on-disk formats may change. The items below mirror the engineering roadmap in the project README—prioritized research, not a shipping guarantee calendar.

Near-term experiments

1. Vector sharding — Split very large .msp files into smaller fragments so mobile clients can hydrate incrementally instead of downloading a monolith.
2. LLM citation drift — Measure hallucination rates when the edge model paraphrases retrieved chunks; if citations wander, fall back to structured citation maps.
3. IndexedDB cache versioning — Ensure schema upgrades cleanly invalidate stale caches without bricking offline search.
4. Parser stress tests — Push the WASM pipeline against huge file counts in Bun/Node to find practical limits (file descriptor pressure, memory).

Product direction

• Harden bundler plugins for Next.js, Astro, and Vite so indexing runs in the default build with minimal config.
• Expand MCP-friendly exports so documentation indexed by Seek.js is easy for agents to consume.

Business context (README)

The open-source framework targets $0 self-hosted workflows; a future managed tier (placeholder names in the README: Vaan / Vantage) would automate embedding, CDN hosting, and edge inference for teams that want zero ops.

For how Seek.js positions against alternatives at a glance, see Comparison.