Skip to content

Easy8 AI on-premises

Install Easy8 AI next to your on-premises Easy8, with cloud models or with models on your own GPU servers. This section is for system administrators.

What the package contains

The standard Easy8 package contains:

  • the main Easy8 docker-compose.yml
  • docker-compose_easy_ai.yml, which adds the AI services (chain, chroma, celery)
  • the environment variables you add to the main Easy8 .env file

The package defines the actual Docker image names and tags. The guides in this section use example values.

Choose a scenario

Pick the guide by product and by where the models run.

Product Services you need Models you need
Easy8 AI Basic chain, celery LLM
Easy8 AI Knowledge assistant PRO chain, celery, chroma LLM, embeddings
Easy8 AI Helpdesk chain, celery, chroma LLM, embeddings

Easy8 AI Knowledge assistant PRO and Easy8 AI Helpdesk have the same infrastructure requirements.

  • Cloud AI: the LLM and embeddings come from a cloud provider (OpenAI or Azure OpenAI).
  • Self-hosted AI: you also run the LLM and, for PRO products, the embeddings on your own GPU servers.
Scenario Minimum additional variables Guide
Basic, cloud AI OPENAI_API_KEY, or AZURE_OPENAI_API_KEY and AZURE_OPENAI_ENDPOINT Basic with cloud AI
Basic, self-hosted AI LLM_URL Basic with self-hosted AI
PRO or Helpdesk, cloud AI OPENAI_API_KEY, or AZURE_OPENAI_API_KEY and AZURE_OPENAI_ENDPOINT PRO and Helpdesk with cloud AI
PRO or Helpdesk, self-hosted AI LLM_URL, EMBEDDINGS_URL PRO and Helpdesk with self-hosted AI

Every scenario also needs the variables from Common configuration and a reverse proxy rule for /ai/chain. Start with Prerequisites.

How the services connect

Browsers reach Chain through /ai/chain on your Easy8 host. Easy8 itself calls Chain at http://chain:8000 on the internal Docker network. Easy8 and Chain must share the same EASY_JWT_SECRET.

Terms

  • Chain: the Easy8 AI backend service. Easy8 and the browser call its API.
  • Chroma: the vector database that stores indexed knowledge for retrieval.
  • Celery: the background worker for indexing and other asynchronous tasks. It uses Redis as broker.
  • LLM: the language model that generates answers.
  • Embeddings: vector representations of your content, used for retrieval in Chroma.

After installation

Reference