feat: make Serper base URL configurable via AGENT_SERPER_DEV_BASE_URL

[ivashhchenko]

The Serper endpoint is hardcoded in the web-browsing agent plugin:

const { response, error } = await fetch(
  "https://google.serper.dev/search",

Anyone who needs that request to go elsewhere has to patch the file and carry the patch across upgrades. It comes up in a few ordinary situations: an egress proxy on a locked-down network, a regional endpoint for latency or data residency, and pointing the agent at a local mock during development so tests do not make live calls.

Approach

This follows the convention already in the codebase. updateENV.js registers AgentSearXNGApiUrl → AGENT_SEARXNG_API_URL, a provider URL sitting next to its key — so Serper gets the same treatment:

AgentSerperApiKey: {
  envKey: "AGENT_SERPER_DEV_KEY",
  checks: [],
},
AgentSerperApiUrl: {
  envKey: "AGENT_SERPER_DEV_BASE_URL",
  checks: [],
},

and the plugin reads it with the current URL as the default:

const serperBaseUrl = (
  process.env.AGENT_SERPER_DEV_BASE_URL || "https://google.serper.dev"
).replace(/\/+$/, "");

Files

• server/utils/agents/aibitat/plugins/web-browsing.js — use the variable
• server/utils/helpers/updateENV.js — register it
• server/.env.example, docker/.env.example — document it next to the key

Ten added lines, no new dependency. Trailing slashes are stripped so https://host, https://host/ and https://host/// all behave the same, and an unset or empty value resolves to the current endpoint — behaviour is unchanged unless the variable is set.

Happy to add a field to the Web Search admin screen in a follow-up if you would like it configurable from the UI rather than only through the environment.