Skip to content

Graphicle 0.1.18

Choose a tag to compare

@github-actions github-actions released this 16 Nov 09:16
· 88 commits to main since this release

Every crate is at 0.1.18, and graphicle --version prints
graphicle 0.1.18 (core). The Python and TypeScript SDKs are at 0.1.18 too.
This release adds OpenCode, makes updates of large graphs faster and reports
incomplete dependency invalidation. No migration runs, and graphs are not
rebuilt: an update writes the same graph and report as 0.1.17.

Added

  • OpenCode: graphicle setup finds OpenCode (its ~/.config/opencode
    folder or the opencode program), and graphicle connect opencode and
    graphicle disconnect opencode add and remove the project's MCP server in
    OpenCode's global opencode.json (or opencode.jsonc, comments kept). The
    entry takes OpenCode 2's shape (mcp.servers) when the file already uses
    it or opencode --version reports 2 or later, else OpenCode 1.x's (mcp,
    which 2.x still reads); disconnect removes it from both. The graph
    engine's OpenCode skill and plugin stay a separate install:
    graphicle graph install --platform opencode (#43, requested by
    @rfrezino).
  • GRAPHICLE_PROFILE=1 also times an update's read of the stored graph,
    its manifest load and its split of changed files.

Fixes

  • Invalidation: a walk that reaches the 100,000-artifact limit used to stop
    without a sign, leaving the rest of the dependents fresh. The change still
    goes through, and the source now carries an invalidation_incomplete
    marker that freshness and every repository's freshness count, as they
    count an outdated answer. The marker stays until a walk from that source
    finishes, or until the document or project is removed, so a source with
    more than 100,000 dependents keeps it (#32, by @dajiaohuang).

Performance

  • Updates of large graphs: matching each new community to the previous
    run's, which keeps community ids stable, counts overlaps from each node's
    communities instead of comparing every pair of communities, with the same
    result. On Kubernetes v1.37.1 (437,234 nodes, about 22,000 communities,
    4 vCPUs) the clustering step of an update went from about 22 s to 10 s.
  • An update no longer loads the whole graph to learn that its HTML view
    cannot render (more communities than the view's node limit); it decides
    from the counts it already holds, with the same message. On the same
    Kubernetes graph that step went from 6.4 s to 1.5 s.

Documentation

  • The README leads with what Graphicle does, a quick start, and Graphicle
    0.1.17 against Graphify 0.9.108 on Kubernetes; docs/benchmarks.md has
    the method and every measurement.

Install

brew install graphicle-ai/tap/graphicle
# or
curl -fsSL https://github.com/Graphicle-AI/graphicle/releases/download/v0.1.18/install.sh | sh -s -- --version 0.1.18

On Windows (x86-64, preview), in PowerShell:

$env:GRAPHICLE_VERSION = '0.1.18'; irm https://github.com/Graphicle-AI/graphicle/releases/download/v0.1.18/install.ps1 | iex

Each archive's SHA-256 sum is in SHA256SUMS and beside it in a .sha256 file.

graphicle.exe and graphicle-graph.exe in the Windows archive carry a signature from a SignPath test certificate. Windows does not trust that certificate, so it treats them as unsigned programs and may warn before it runs them. Code signing policy