2. Getting started

Two ways to run SPEAR: a container that carries everything, and a native installation. Start with the quick start. A colleague who only wants to use SPEAR needs nothing past Running the public image; the later chapters are for whoever installs it natively or builds the images.

• 2.1. Getting started
  • 2.1.1. Prerequisites
  • 2.1.2. Install
  • 2.1.3. Configure a model backend
  • 2.1.4. Use a project
  • 2.1.5. Make a change
  • 2.1.6. Bind a standard
  • 2.1.7. Make a change the standard governs
  • 2.1.8. Read the result
  • 2.1.9. Retrieval is not a detail
  • 2.1.10. Where next
• 2.2. Running the public image
  • 2.2.1. What you need
  • 2.2.2. Getting the image
  • 2.2.3. Laying out your checkouts
  • 2.2.4. Running it
  • 2.2.5. The first session: indexing
  • 2.2.6. When something is wrong
• 2.3. Installation
  • 2.3.1. Prerequisites
  • 2.3.2. Native installation
  • 2.3.3. Container installation
  • 2.3.4. The inference runtime
  • 2.3.5. What ends up where
  • 2.3.6. Checking the installation
• 2.4. Container
  • 2.4.1. Managing images
  • 2.4.2. Daily use
  • 2.4.3. What is baked, and what is not
  • 2.4.4. Two build contexts
  • 2.4.5. The optional inputs
  • 2.4.6. Two profiles
  • 2.4.7. Baking the trees
  • 2.4.8. Common and local state
  • 2.4.9. A deployment’s own content, and how it gets in
  • 2.4.10. Publishing, and not publishing
  • 2.4.11. Relative corpus paths
  • 2.4.12. Mounts are derived, not listed
  • 2.4.13. The three flags that are not optional
  • 2.4.14. Refreshing the index
  • 2.4.15. Why retrieval is not served from the GPU host