/oss · 3 merged

xerj-org/xerj

A search engine that works with existing Elasticsearch tools, built for AI agents to search code, docs and logs.

the project on GitHub (opens in a new tab)
  1. 20 Oct 2026

    Date range searches now respect the time zone

    What was broken
    A date range search ignored the time zone you gave it and read every date as UTC. Documents inside the real window were left out.
    What I changed
    I made the search read dates in the given zone, including daylight saving changes, and round dates there too, the way Elasticsearch does.
    Why it matters
    Range searches with a time zone return the same documents as Elasticsearch.

    see the code, #1744 (opens in a new tab)

  2. 17 Oct 2026

    Turning off the disk check lifts the full-disk block

    What was broken
    When a disk got too full, writes were blocked. The standard setting to turn that check off was accepted but ignored, so the block stayed on.
    What I changed
    I made the server read that setting, so turning the check off also lifts the block.
    Why it matters
    Users can unblock writes the same way they would in Elasticsearch.

    see the code, #1596 (opens in a new tab)

  3. 17 Oct 2026

    Changing one cluster setting keeps the others

    What was broken
    Saving one cluster-wide setting wiped every other setting saved before it. Settings sent in the nested style were stored but never used.
    What I changed
    I made it merge each update into the saved settings, the way Elasticsearch does, and store every setting under one name form.
    Why it matters
    Earlier settings, like a disk or memory limit, stay in force when someone changes something else.

    see the code, #1595 (opens in a new tab)

← every project