Configure LLM

AilaFlow requires access to large language models (LLMs) to execute work and assist users. Administrators can configure LLM providers in the Configuration section of the AilaFlow Portal.

AilaFlow lets you select a different model for each of these use cases:

• Admin chat
  Requires a strong coding and reasoning model that can reliably configure and build processes.

• User chat
  Requires a capable general-purpose model that can reliably use available processes and tools.

• Agent step
  Requires a strong reasoning model that can reliably use tools and complete assigned tasks independently.

AilaFlow currently supports these providers:

• OpenAI-compatible API
• OpenAI
• Codex app-server
• Anthropic

After selecting a provider and entering the required configuration, select Fetch models to test the connection. If the test succeeds, save the configuration and set your preferred models in the Use cases section.

OpenAI-compatible API

Use this provider for services that expose an API compatible with the OpenAI API. Enter the API key and the provider’s API URL.

For example, the Groq API URL is:

https://api.groq.com/openai/v1

OpenAI

Select OpenAI and enter your OpenAI API key. An API URL is not required.

Codex app-server

Use Codex app-server to connect AilaFlow through your Codex subscription instead of an API key.

To start the Codex WebSocket server, run Codex app-server on the machine hosting AilaFlow:

codex app-server --listen ws://127.0.0.1:4500

Keep the command running and set the WebSocket URL in AilaFlow to:

ws://127.0.0.1:4500

Anthropic

Select Anthropic and enter your Anthropic API key. Then choose an available Claude model.