Code graph and memory for your coding agents, in one Rust program. For Claude Code, Codex, Cursor and any MCP client.
Rust agent-memoryai-agentsclaude-codecode-analysiscode-graphcode-intelligencecode-searchcodexcoding-agentscontext-engineeringcursordeveloper-toolsgraphragknowledge-graphlocal-firstmcpmcp-servermodel-context-protocolrusttree-sitter
# Graphicle review: one pull-request review from Muse Spark, with the
# repository's context from Graphicle (docs/review.md).
#
# Security (docs/review.md#security):
# - runs on the pull_request event only; refuses pull_request_target, and skips
# pull requests from forks, which get no secrets;
# - checks out the pull request's head without keeping the token in .git;
# - the model key comes from the input the workflow fills from the
# MODEL_API_KEY secret, and only the review step sees it; only the posting
# step sees the GitHub token;
# - the actions it uses are pinned by full commit SHA, and the Graphicle
# release by version and SHA-256;
# - one model call per run, refused before it is sent when its worst case
# costs more than max-cost-usd at the configured price;
# - network: the release download, the model API and the GitHub API that
# posts the review.
name: Graphicle review
description: Review a pull request with Muse Spark, with the changed code's callers, callees, tests and architecture from Graphicle, and post one review.
author: Graphicle
branding:
icon: eye
color: blue
inputs:
graphicle-version:
description: The Graphicle release to install, for example 0.1.9. Pin one; there is no "latest".
required: true
graphicle-sha256:
description: The SHA-256 of that release's archive for the runner (graphicle-VERSION-TARGET.tar.gz in the release's SHA256SUMS).
required: true
model-api-key:
description: The Meta Model API key. Pass secrets.MODEL_API_KEY; without it the review is skipped.
required: false
default: ""
model:
description: The Meta model, muse-spark-1.3 or muse-wonder-1.0.
required: false
default: muse-spark-1.3
price:
description: The model's price in US dollars per million tokens, as JSON; the cost cap needs it. The default is muse-spark-1.3's list price on 26 Oct 2026.
required: false
default: '{"input":1.25,"output":4.25,"cached_input":0.15}'
max-cost-usd:
description: The most one review may cost, in US dollars at that price. The call is refused before it is sent when its worst case (the whole input uncached plus the output limit) costs more.
required: false
default: "0.50"
budget:
description: The pack's budget in tokens.
required: false
default: "200000"
max-comments:
description: Inline comments at most, the most severe first.
required: false
default: "10"
post-without-comments:
description: Post the review's summary when it has no inline comments ("true" or "false").
required: false
default: "true"
github-token:
description: The token that posts the review; it needs pull-requests write.
required: false
default: ${{ github.token }}
outputs:
comments:
description: The number of inline comments posted.
value: ${{ steps.review.outputs.comments }}
runs:
using: composite
steps:
- name: Check the event
id: gate
shell: bash
env:
EVENT: ${{ github.event_name }}
HEAD_REPO: ${{ github.event.pull_request.head.repo.full_name }}
BASE_REPO: ${{ github.repository }}
HAS_KEY: ${{ inputs.model-api-key != '' }}
run: bash "$GITHUB_ACTION_PATH/gate.sh"
- name: Check out the pull request's head
if: steps.gate.outputs.run == 'true'
uses: actions/checkout@d23441a48e516b6c34aea4fa41551a30e30af803 # v6.1.0
with:
ref: ${{ github.event.pull_request.head.sha }}
fetch-depth: 0
persist-credentials: false
- name: Install Graphicle ${{ inputs.graphicle-version }}
if: steps.gate.outputs.run == 'true'
shell: bash
env:
VERSION: ${{ inputs.graphicle-version }}
SHA256: ${{ inputs.graphicle-sha256 }}
run: bash "$GITHUB_ACTION_PATH/install.sh"
- name: Restore the index
if: steps.gate.outputs.run == 'true'
uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
with:
path: ${{ runner.temp }}/graphicle-data
key: graphicle-index-${{ runner.os }}-${{ inputs.graphicle-version }}-${{ github.event.pull_request.base.sha }}
restore-keys: |
graphicle-index-${{ runner.os }}-${{ inputs.graphicle-version }}-
- name: Review
id: review
if: steps.gate.outputs.run == 'true'
shell: bash
env:
MODEL_API_KEY: ${{ inputs.model-api-key }}
MODEL: ${{ inputs.model }}
PRICE: ${{ inputs.price }}
MAX_COST: ${{ inputs.max-cost-usd }}
BUDGET: ${{ inputs.budget }}
MAX_COMMENTS: ${{ inputs.max-comments }}
BASE_SHA: ${{ github.event.pull_request.base.sha }}
HEAD_SHA: ${{ github.event.pull_request.head.sha }}
run: bash "$GITHUB_ACTION_PATH/review.sh"
- name: Post the review
if: steps.gate.outputs.run == 'true'
shell: bash
env:
GH_TOKEN: ${{ inputs.github-token }}
REPOSITORY: ${{ github.repository }}
PULL: ${{ github.event.pull_request.number }}
POST_EMPTY: ${{ inputs.post-without-comments }}
run: bash "$GITHUB_ACTION_PATH/post.sh"