GitHub Workflows for Ollama and pgvector

Configure GitHub Actions workflows for testing RAG pipelines, vector embeddings, and Ollama-based AI applications.

When to Apply

Use this skill when:

• Setting up CI/CD for RAG applications
• Testing pgvector and embedding functionality in CI
• Configuring Ollama service containers
• Running integration tests that need LLM inference
• Building pipelines for AI-powered applications

Complete Workflow Template

name: CI tests
on:
  pull_request:
    branches:
      - main
  push:
    branches:
      - main
  workflow_dispatch:

concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}-tests
  cancel-in-progress: true

env:
  PGPM_VERSION: '2.7.9'

jobs:
  test:
    runs-on: ubuntu-latest
    continue-on-error: true
    strategy:
      fail-fast: false
      matrix:
        package:
          - my-rag-package

    env:
      PGHOST: localhost
      PGPORT: 5432
      PGUSER: postgres
      PGPASSWORD: password
      OLLAMA_HOST: http://localhost:11434

    services:
      pg_db:
        image: pyramation/postgres:17
        env:
          POSTGRES_USER: postgres
          POSTGRES_PASSWORD: password
        options: >-
          --health-cmd pg_isready
          --health-interval 10s
          --health-timeout 5s
          --health-retries 5
        ports:
          - 5432:5432

      ollama:
        image: ollama/ollama:latest
        ports:
          - 11434:11434

    steps:
      - name: Configure Git (for tests)
        run: |
          git config --global user.name "CI Test User"
          git config --global user.email "ci@example.com"

      - name: Checkout
        uses: actions/checkout@v4

      - name: Setup pnpm
        uses: pnpm/action-setup@v2
        with:
          version: 10

      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
          node-version: '20'
          cache: 'pnpm'

      - name: Install dependencies
        run: pnpm install

      - name: Cache pgpm CLI
        uses: actions/cache@v4
        with:
          path: ~/.npm
          key: pgpm-${{ runner.os }}-${{ env.PGPM_VERSION }}

      - name: Install pgpm CLI globally
        run: npm install -g pgpm@${{ env.PGPM_VERSION }}

      - name: Build
        run: pnpm -r build

      - name: Seed pg and app_user
        run: |
          pgpm admin-users bootstrap --yes
          pgpm admin-users add --test --yes

      - name: Wait for Ollama and pull models
        run: |
          echo "Waiting for Ollama to be ready..."
          for i in $(seq 1 30); do
            if curl -s http://localhost:11434/api/tags > /dev/null 2>&1; then
              echo "Ollama is ready!"
              break
            fi
            echo "Waiting for Ollama... ($i/30)"
            sleep 2
          done
          echo "Pulling nomic-embed-text model (for embeddings)..."
          curl -s -X POST http://localhost:11434/api/pull -d '{"name": "nomic-embed-text"}'
          echo ""
          echo "Pulling mistral model (for RAG response generation)..."
          curl -s -X POST http://localhost:11434/api/pull -d '{"name": "mistral"}'

      - name: Test ${{ matrix.package }}
        run: cd ./packages/${{ matrix.package }} && pnpm test

Service Containers

PostgreSQL with pgvector

Use the Constructive PostgreSQL image with pgvector and other extensions:

services:
  pg_db:
    image: pyramation/postgres:17
    env:
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: password
    options: >-
      --health-cmd pg_isready
      --health-interval 10s
      --health-timeout 5s
      --health-retries 5
    ports:
      - 5432:5432

For projects requiring additional extensions:

services:
  pg_db:
    image: ghcr.io/constructive-io/docker/postgres-plus:17
    env:
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: password
    options: >-
      --health-cmd pg_isready
      --health-interval 10s
      --health-timeout 5s
      --health-retries 5
    ports:
      - 5432:5432

Available images:

• pyramation/postgres:17 - PostgreSQL 17 with pgvector (recommended)
• ghcr.io/constructive-io/docker/postgres-plus:17 - PostgreSQL 17 with pgvector and additional extensions

Ollama Service

services:
  ollama:
    image: ollama/ollama:latest
    ports:
      - 11434:11434

Note: Ollama doesn't have a built-in health check, so we wait for it in a step.

Environment Variables

env:
  # PostgreSQL connection
  PGHOST: localhost
  PGPORT: 5432
  PGUSER: postgres
  PGPASSWORD: password

  # Ollama connection
  OLLAMA_HOST: http://localhost:11434

Waiting for Ollama

Ollama takes time to start. Use this pattern to wait:

- name: Wait for Ollama and pull models
  run: |
    echo "Waiting for Ollama to be ready..."
    for i in $(seq 1 30); do
      if curl -s http://localhost:11434/api/tags > /dev/null 2>&1; then
        echo "Ollama is ready!"
        break
      fi
      echo "Waiting for Ollama... ($i/30)"
      sleep 2
    done

Pulling Models

Models must be pulled before tests run:

- name: Pull embedding model
  run: |
    curl -s -X POST http://localhost:11434/api/pull \
      -d '{"name": "nomic-embed-text"}'

- name: Pull generation model
  run: |
    curl -s -X POST http://localhost:11434/api/pull \
      -d '{"name": "mistral"}'

Model Pull Times

Model	Size	Approximate Pull Time
nomic-embed-text	~275MB	30-60s
mistral	~4GB	2-5 min
llama2	~4GB	2-5 min
all-minilm	~45MB	10-20s

Consider using smaller models in CI for faster runs.

Test Timeout Configuration

LLM operations can be slow. Configure Jest timeout:

// In test file
jest.setTimeout(300000); // 5 minutes

// Or in jest.config.js
module.exports = {
  testTimeout: 300000,
};

Caching Strategies

Cache pnpm dependencies

- name: Setup pnpm
  uses: pnpm/action-setup@v2
  with:
    version: 10

- name: Setup Node.js
  uses: actions/setup-node@v4
  with:
    node-version: '20'
    cache: 'pnpm'

Cache pgpm CLI

env:
  PGPM_VERSION: '2.7.9'

steps:
  - name: Cache pgpm CLI
    uses: actions/cache@v4
    with:
      path: ~/.npm
      key: pgpm-${{ runner.os }}-${{ env.PGPM_VERSION }}

  - name: Install pgpm CLI globally
    run: npm install -g pgpm@${{ env.PGPM_VERSION }}

Matrix Testing

Test multiple packages in parallel:

strategy:
  fail-fast: false
  matrix:
    package:
      - packages/embeddings
      - packages/rag-service
      - packages/vector-search

steps:
  - name: Test ${{ matrix.package }}
    run: cd ./${{ matrix.package }} && pnpm test

Minimal Workflow (Embeddings Only)

For projects that only need embeddings (no LLM generation):

name: Embedding Tests
on: [push, pull_request]

jobs:
  test:
    runs-on: ubuntu-latest

    env:
      PGHOST: localhost
      PGPORT: 5432
      PGUSER: postgres
      PGPASSWORD: password
      OLLAMA_HOST: http://localhost:11434

    services:
      postgres:
        image: pyramation/postgres:17
        env:
          POSTGRES_USER: postgres
          POSTGRES_PASSWORD: password
        options: >-
          --health-cmd pg_isready
          --health-interval 10s
          --health-timeout 5s
          --health-retries 5
        ports:
          - 5432:5432

    steps:
      - uses: actions/checkout@v4

      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
          node-version: '20'

      - name: Install Ollama
        run: curl -fsSL https://ollama.com/install.sh | sh

      - name: Start Ollama and pull model
        run: |
          ollama serve &
          sleep 5
          ollama pull nomic-embed-text

      - name: Install dependencies
        run: npm install

      - name: Run tests
        run: npm test

Debugging Failed Tests

View Ollama logs

- name: Debug Ollama
  if: failure()
  run: |
    echo "Checking Ollama status..."
    curl -s http://localhost:11434/api/tags || echo "Ollama not responding"

Check PostgreSQL

- name: Debug PostgreSQL
  if: failure()
  run: |
    psql -h localhost -U postgres -c "SELECT version();"
    psql -h localhost -U postgres -c "SELECT * FROM pg_extension WHERE extname = 'vector';"

Best Practices

1. Use fail-fast: false - Let all tests complete even if some fail
2. Set generous timeouts - LLM operations are slow
3. Pull models early - Do it before running tests
4. Use smaller models in CI - all-minilm instead of nomic-embed-text for speed
5. Cache dependencies - pnpm and pgpm caching speeds up runs
6. Health checks - Always use health checks for PostgreSQL
7. Wait for Ollama - It doesn't have built-in health checks

Troubleshooting

Issue	Solution
Ollama not responding	Increase wait time, check service logs
Model pull timeout	Use smaller model or increase timeout
pgvector not found	Ensure using pgvector-enabled image
Tests timeout	Increase Jest timeout, use streaming
Out of memory	Use smaller models or reduce parallelism

References

• Related skill: github-workflows-pgpm for general PGPM CI/CD
• Related skill: pgpm (references/testing.md) for database testing
• Related skill: rag-pipeline for RAG implementation
• Related skill: ollama-integration for Ollama client
• Ollama Docker documentation