Table of Contents

1. Configuration
   • AI Assistant
   • API
   • RAG

---

Configuration

The setup of AI-OPS Agent and API can be customized with environment variables loaded from a .env file; there are three classes of configuration settings: Assistant, API and RAG (in-development).

To override the default configuration values, create a .env file in the root of your project directory. This file can include any of the settings listed above. The .env file should look something like this:

# example .env
MODEL=mistral:7b-instruct-v0.3-q8_0
ENDPOINT=http://your-ollama-endpoint

AI Assistant

Variable	Description
MODEL	The model used by the AI assistant. This should be set to the exact name of the model as recognized by Ollama. For example, if you are using mistral:7b-instruct-v0.3-q8_0, set MODEL=mistral:7b-instruct-v0.3-q8_0. Default value: mistral
ENDPOINT	The endpoint where the Ollama service is running. This is used for both the AI agent model and the embedding model (for RAG). In most local deployments, this will point to http://localhost:11434, but you should change it if Ollama is deployed elsewhere. Default value: http://localhost:11434

Additional variables will be used in the future such as:

• USE_RAG: If set to True, this enables the Retrieval-Augmented Generation (RAG) functionality, which allows the agent to fetch information from a vector database.
• PROVIDER and PROVIDER_KEY: The model provider and an eventual required api key. Currently only ollama is supported, but this parameter is in place for future support of additional providers.

API

Variable	Description
PROFILE	If set to True, enables profiling of system performance, such as memory usage. This setting can be useful for debugging and development. Default value: False

RAG

Note: RAG functionality is in active development and may not be fully stable. It is recommended that users prefer persistent Qdrant instances when using RAG.

Setting	Description
RAG_URL	The URL of the Qdrant vector database (local or remote). Default Value: http://localhost:6333
IN_MEMORY	If set to True, the vector database will be kept in memory, which is suitable for lightweight testing but not recommended for large datasets. Default Value: True
EMBEDDING_MODEL	The model used for generating embeddings, should be an Ollama-compatible model. Default Value: nomic-embed-text
EMBEDDING_URL	The URL of the embedding service, typically the same as the ENDPOINT URL. Default Value: http://localhost:11434