Add GitHub Copilot embedding provider

## What changed OpenClaw added a GitHub Copilot embedding provider on 19 May 2026. The update targets memory search and lets plugins reuse the existing transport layer. It respects remote overrides and handles token refresh automatically.

The change ships as part of the core agent runtime. No separate install or new CLI flag is required once the provider key is set in the YAML file.

## Why it matters Vibe Builders gain another low-cost option for storing and retrieving long-term context without managing a separate vector database. The integration reduces the number of moving parts in a self-hosted stack.

It also pressures other embedding services by making Copilot's transport interchangeable. Builders who already pay for Copilot seats can now route memory calls through the same account instead of opening a second bill.

## How to use it Edit the config.yaml file and add the copilot section under embeddings. Supply your GitHub token and set the model name to the desired embedding endpoint.

Restart the agent or trigger a config reload from the CLI. Memory search calls will then route through Copilot while still honoring any per-plugin override you define.

## Watch for Confirm the bet if token usage stays flat after switching and search recall improves on older conversations. The bet breaks if Copilot rate limits appear during peak agent activity. Expect a follow-up move that adds the same provider to the browser-control and skill-writing paths.