Organization-wide defaults and shared automation for every repository in the FalkorDB GitHub organization.
Looking for FalkorDB itself? Head to FalkorDB/FalkorDB, the docs, or www.falkordb.com.
| Path | Purpose |
|---|---|
profile/README.md |
The page shown on the organization profile. |
CODE_OF_CONDUCT.md |
Default Code of Conduct (Contributor Covenant 2.1). |
CONTRIBUTING.md |
Default contribution guidelines: licensing, issues vs. Discussions, PR flow. |
SECURITY.md |
Default security policy and vulnerability reporting process. |
.github/workflows/report-scheduled-failure.yml |
Reusable workflow that reports failing nightly runs against falkordb/falkordb:edge. |
.github/workflows/release-notes.yml |
Reusable workflow that rewrites a release's auto-generated notes into user-facing ones. Its scripts live in .github/scripts/release-notes/. |
GitHub uses the CODE_OF_CONDUCT.md, CONTRIBUTING.md and SECURITY.md in this
repository for any FalkorDB repository that doesn't have its own copy (in its root,
docs/ or .github/ folder). A repository's own file always takes precedence.
Because these files apply to every repository at once, keep them generic: anything specific to one project belongs in that project's own repository.
Call a shared workflow at @main:
jobs:
<job-id>:
uses: FalkorDB/.github/.github/workflows/<name>.yml@mainThat way a fix merged here reaches every project on its next run, with no change in each repository. It also means every change here runs with the callers' secrets and tokens, so it must go through a reviewed pull request, never a direct push.
Pinning a full commit SHA instead is still fine, for a repository that wants to take
changes on its own schedule. Dependabot's github-actions ecosystem keeps such pins
current where it is enabled; the client libraries call report-scheduled-failure
this way.
A cron-triggered workflow has nobody watching it, so a failing nightly run can go
unnoticed. This workflow posts the failure to the drivers team's Google Chat, or,
if no webhook is configured, opens (and then comments on) a single tracking issue
labelled nightly-edge-failure in the calling repository.
Call it from the workflow that runs against falkordb/falkordb:edge:
report-scheduled-failure:
name: Report scheduled failure
needs: [build] # every job the run should be considered failed for
if: ${{ always() && github.event_name == 'schedule' && contains(needs.*.result, 'failure') }}
permissions:
issues: write # required: a called workflow can only downgrade permissions
uses: FalkorDB/.github/.github/workflows/report-scheduled-failure.yml@main
secrets:
DRIVERS_GOOGLE_CHAT_WEBHOOK_URL: ${{ secrets.DRIVERS_GOOGLE_CHAT_WEBHOOK_URL }}| Name | Kind | Required | Description |
|---|---|---|---|
issue-title |
input | no | Title of the tracking issue. Defaults to CI is failing against falkordb/falkordb:edge; keep it stable per repository. |
DRIVERS_GOOGLE_CHAT_WEBHOOK_URL |
secret | no | Google Chat incoming webhook. When unset, the tracking-issue fallback is used. |
GitHub's auto-generated release notes list PR titles, most of which mean little to users. When a release is published, this workflow rewrites them with an Azure OpenAI model:
- It drops changes users don't notice, such as dependency bumps, CI and refactors.
- It writes one plain sentence per remaining change, under features, improvements, fixes, security and deployment.
- It keeps GitHub's original notes collapsed at the bottom.
Reruns start again from those originals, so running it twice never rewrites the rewrite.
Call it from a workflow of its own, so the repository also gets a manual Run
workflow button for one tag, a range of tags, or a dry run that only prints the
result. The header of release-notes.yml has
the full caller to copy, including those manual inputs; its core is:
jobs:
polish:
permissions:
contents: write # edit the release
pull-requests: read # read the PRs it lists
id-token: write # find this workflow's commit, to run its scripts
uses: FalkorDB/.github/.github/workflows/release-notes.yml@main
with:
product: FalkorDB Python client
product-description: the Python library for FalkorDB graph databases
tag: ${{ inputs.tag }}
from-tag: ${{ inputs.from_tag }}
dry-run: ${{ inputs.dry_run || false }}
secrets:
AZURE_OPENAI_API_KEY: ${{ secrets.AZURE_OPENAI_API_KEY }}| Name | Kind | Required | Description |
|---|---|---|---|
product |
input | yes | The product's name as the notes should call it. |
product-description |
input | no | One phrase on what the product is, for the model's context. |
whats-new |
input | no | Lead the notes with a short ## What's New section. Only for a product that shows it to users after they upgrade, as FalkorDB Browser does. Defaults to false. |
tag |
input | no | Release tag to rewrite, or the newest one of a range. Empty on a release event. |
from-tag |
input | no | Oldest tag of a range; every published release from it through tag is rewritten. |
dry-run |
input | no | Print the new notes without updating the release. |
AZURE_OPENAI_API_KEY |
secret | yes | Key for the Azure OpenAI resource. |
AZURE_OPENAI_ENDPOINT |
variable | yes | The resource URL, e.g. https://my-resource.openai.azure.com. Set on the organization or the calling repository. |
AZURE_OPENAI_DEPLOYMENT |
variable | no | The model deployment name. Defaults to gpt-6-luna. |
A leftover ## Download section between <!-- release-downloads:start --> and
<!-- release-downloads:end --> lines, from FalkorDB Browser's older notes, is never shown
to the model and is dropped.
Before relying on it, run it by hand with dry-run on to check the output.
Changes to this repository affect the whole organization, so please open a pull
request rather than pushing to main. See CONTRIBUTING.md for
the general process, and report security issues privately as described in
SECURITY.md.