Show HN: Local automation runner with built-in LLM steps – YAML pipelines

YAML pipelines. Python plugins. Runs on your machine - no server to set up, no cloud account needed.

Stepyard is an automation runner for developers. Pipelines live as YAML files in your repo. You extend them with plain Python functions. Everything runs from the CLI or a local daemon - on your machine or a server you own.

pip install stepyard
stepyard init my-project && cd my-project
stepyard run hello

A deploy pipeline: build and push a container, run a smoke test, and post the result to Slack. One YAML file, no supporting scripts.

name: deploy
description: Build, push, and verify the production container.

steps:
  - id: build
    uses: shell.run
    with:
      command: docker build -t myapp:${{ env.GIT_SHA }} .

  - id: push
    uses: shell.run
    with:
      command: docker push myapp:${{ env.GIT_SHA }}

  - id: smoke_test
    uses: http.request
    with:
      url: https://staging.myapp.com/healthz
      method: GET

  - id: notify
    uses: llm.generate               # built-in - no plugin needed
    with:
      model: gpt-4o-mini
      prompt: |
        Summarise this deploy result in one sentence for Slack:
        HTTP status: ${{ steps.smoke_test.output.status }}
        SHA: ${{ env.GIT_SHA }}

  - id: post_to_slack
    uses: http.request
    with:
      url: ${{ env.SLACK_WEBHOOK }}
      method: POST
      json_body:
        text: ${{ steps.notify.output.output }}

Run it:

GIT_SHA=$(git rev-parse --short HEAD) stepyard run deploy
✓  build          12.4 s
✓  push            4.1 s
✓  smoke_test      0.3 s
✓  notify          0.9 s
✓  post_to_slack   0.2 s

Flow completed in 18.0 s

Why Stepyard?¶ #

Flows are files in your repo Steps, conditions, loops, and retries are plain YAML keys - no proprietary DSL. Version-control them alongside your code and validate them with

stepyard validate

. - Extend with plain Python One

@node

decorator turns any function into a reusable step. Inputs are type-validated automatically; plugin dependencies are isolated so they never conflict with Stepyard's own. - Nothing leaves your machine State is stored in a local SQLite database. Data goes out only if a step in your flow explicitly sends it.

Scheduled and on-demand executioncron

,interval

, andstartup

triggers. Runstepyard service start

and your flows execute on schedule - no external service or cloud account needed.

Real-world examples¶ #

Daily database backup¶

name: pg_backup
trigger:
  uses: cron
  with:
    schedule: "0 3 * * *"   # every day at 03:00

steps:
  - id: dump
    uses: shell.run
    with:
      command: pg_dump ${{ env.DATABASE_URL }} | gzip > /tmp/backup.sql.gz

  - id: upload
    uses: shell.run
    with:
      command: |
        aws s3 cp /tmp/backup.sql.gz \
          s3://${{ env.BACKUP_BUCKET }}/db/$(date +%Y-%m-%d).sql.gz

  - id: cleanup
    uses: shell.run
    with:
      command: rm -f /tmp/backup.sql.gz

Automated code review on every PR¶

Fetch the diff via the GitHub API, review it with an LLM, and post a comment back - all with built-in nodes, no additional service.

name: pr_review
description: Review a PR diff with an LLM and post a comment if issues are found.

steps:
  - id: diff
    uses: http.request
    with:
      url: https://api.github.com/repos/${{ env.GITHUB_REPO }}/pulls/${{ env.PR }}/files
      headers:
        Authorization: Bearer ${{ env.GITHUB_TOKEN }}
        Accept: application/vnd.github+json

  - id: review
    uses: llm.generate
    with:
      model: gpt-4o
      system_prompt: |
        You are a senior engineer doing a code review.
        Be concise. If everything looks good reply with exactly: LGTM
      prompt: |
        Review this pull request for bugs, security issues, and obvious mistakes:

        ${{ steps.diff.output.body }}

  - id: post_comment
    if: ${{ steps.review.output.output != "LGTM" }}
    uses: http.request
    with:
      url: https://api.github.com/repos/${{ env.GITHUB_REPO }}/issues/${{ env.PR }}/comments
      method: POST
      headers:
        Authorization: Bearer ${{ env.GITHUB_TOKEN }}
        Accept: application/vnd.github+json
      json_body:
        body: ${{ steps.review.output.output }}

Multi-environment deployment with rollback¶

name: deploy_multi

steps:
  - id: deploy_staging
    uses: shell.run
    with:
      command: kubectl apply -f k8s/staging/

  - id: integration_tests
    uses: shell.run
    continue_on_error: true
    with:
      command: pytest tests/integration/ -q

  - id: deploy_production
    if: ${{ steps.integration_tests.output.code == 0 }}
    uses: shell.run
    with:
      command: kubectl apply -f k8s/production/

  - id: rollback
    if: ${{ steps.integration_tests.output.code != 0 }}
    uses: shell.run
    with:
      command: kubectl rollout undo deployment/myapp -n staging

Documentation overview¶ #

Core ConceptsHow-to GuidesBuilt-in NodesPlugin DevelopmentCLI Reference## Compatibility¶

| Python | 3.10 · 3.11 · 3.12 · 3.13 | | OS | macOS · Linux · Windows (WSL) | | Storage | SQLite (default) | | License | MIT |