agent/

Runme Agent

This is a quickstart to get you up and running to work on Runme Agent.

./pkg/agent contains the golang server The client side web application intended to be served by the golang server lives here https://github.com/runmedev/web and distributed via npm.

Quickstart Setup

Configure OpenAI

Create a minimal configuration file ~/.runme-agent/config.yaml

apiVersion: ""
kind: ""
logging:
  level: debug
  sinks:
    - path: stderr
openai:
  apiKeyFile: /Users/${USER}/.runme-agent/openai_key_file
cloudAssistant:
  vectorStores:
    - ${VSID} # e.g. vs_67e5xxxxcabxfakexxxe13b2fcd7e612, get your own from OpenAI
assistantServer:
  port: 8080
  httpMaxReadTimeout: 0s
  httpMaxWriteTimeout: 0s
  staticAssets: /workspaces/runme-web/packages/react-components/dist/app # bundled version of https://github.com/runmedev/web
  agentService: true
  runnerService: true
  corsOrigins:
    - "http://localhost:5173"
    - "http://localhost:3000"
    - "http://localhost:8080"

• set apiKeyFile to the path of your OpenAI API key
• set vectoreStores to contain the ID of your OpenAI API vector store
• Change the path to the static assets to the location where you checked out the repository

runme agent config set assistantServer.staticAssets=$(PWD)/web/dist

Build the static assets

git clone http://github.com/runmedev/web runme-web
cd runme-web
runme run setup clean build

Start the server

runme agent serve

Open up https://localhost:8443.

Development Mode

If you make changes to the UI you need to rerun npm run build to recompile the static assets. However, you don't need to restart the GoLang server; it is sufficient to refresh the page to pick up the latest static assets.

Local Tracing

It's handy to have local tracing for debugging. Make sure to configure the OTLP endpoint in the config.yaml file.

telemetry:
  otlpHTTPEndpoint: localhost:4318

Run Jaeger locally

docker run --rm --name jaeger \
  -p 16686:16686 \
  -p 4317:4317 \
  -p 4318:4318 \
  -p 5778:5778 \
  -p 9411:9411 \
  jaegertracing/jaeger:2.6.0