github atomicstrata/llm-wiki-compiler v1.5.0-rc.1
1.5.0-rc.1

pre-release3 hours ago

A cleaner foundation for your existing wiki

1.5.0-rc.1 is the first release candidate for llmwiki 1.5. It is available on npm's next channel for early testing; 1.4.2 remains the stable release on latest.

The main change separates wiki behavior from local filesystem operations. Compilation, retrieval, review and project state work through focused storage interfaces, with the existing disk layout as their implementation. Existing wikis need no migration or new setting. This creates a foundation for alternative storage without shipping a remote backend, hosted service, role-based permissions or a public storage-plugin API yet.

Also in this candidate

  • Converse gateways: use bearer-authenticated endpoints speaking the Converse protocols.
  • Source-driven extraction: compile --no-pages-limit removes the requested concept count so the model can follow the substance of a source. Generated pages still need review.
  • Viewer improvements: refresh wiki snapshots without restarting the viewer and configure its port with LLMWIKI_VIEW_PORT.
  • Correctness fixes: normalize extracted wikilink titles, support GPT-6 Luna request parameters, keep MCP stdout clean, and avoid blocking freshness checks on nonregular source files.

Try it

npm install --global llm-wiki-compiler@1.5.0-rc.1

The matching core and local-workflow packages install automatically. Back up important wikis before trying a preview, and exercise your existing CLI, SDK or MCP workflows. Please report regressions before the stable release.

Every newly generated page records prompt version v7, including default runs. Upgrading alone does not recompile unchanged sources; extraction snapshots from the earlier prompt version are not reused when extraction is needed again. The MCP Registry entry is not updated for this candidate.

Thank you

Thanks to @LorenzoGentile for the Converse provider (#282), source-driven extraction (#277, integrated as #285), wikilink normalization (#278), and GPT-6 Luna support (#279); @Erick0412-dev for viewer refresh (#275) and port configuration (#274); and @vishalhabib99 for quiet MCP wiki tools (#270). Their contributions make this release more useful and reliable across everyday workflows.

See the full changelog, upgrade notes, and architecture.

Don't miss a new llm-wiki-compiler release

NewReleases is sending notifications on new releases.