Running agentify locally
Everything you need in order to run agentify locally
Overview
All agentify instances are built on Houdart's pipelines / build agents. The built image then gets pushed to a container registry you own (Azure Container Registry / GitHub Container Registry / ...), from which you can deploy the instance in any way you would like.
Set-up
Pre-requisites
- Have docker installed.
Connecting to your container registry
Connecting to your container registry may differ based on where it is hosted and your organisations policies. Here's an example with Azure Container Registry:
Run the following commands in your terminal
# Login to azure (you may have to select a subscription if your org has multiple)
az login
# After logging in to azure, login to the container registry
az acr login --name <registry>At this point you should be logged in to the azure container. If you get an error, the most common culprits are:
- Missing permissions
- Selecting the wrong subscription
- Docker engine is not running
Starting the agentify instance
Dashboard
The dashboard is already built with all your environment variables, so you only have to run it:
docker run -p 3000:3000 path-to-dashboard-image:latestAgentify API
The Agentify API reads the environment variables at runtime, meaning you need to provide them.
docker run -p 3001:3001 \
-e APPLICATION_ID= \
-e APPLICATION_KEY= \
-e APPLICATION_SECRET= \
-e LLM_API_KEY= \
-e LLM_ENDPOINT= \
-e LLM_MODEL_NAME= \
-e LLM_PROVIDER= \
-e LLM_VERSION= \ # optional
-e POSTHOG_API_KEY= \ # optional
-e TELEMETRY_LEVEL= \ # number for how much telemetry data can be sent, from 0-5.
path-to-api-image:latestIn a self-hosted version you can also pass environment variables for your tools here, such as: TOOL_MYTOOLNAME_API_KEY, alternatively you can set them in the database, similarly to non-self-hosted instances.