Speed up first copilot mode calls by pre-warming session

[hillary-mutisya]

This pull request introduces several enhancements to the Copilot integration, focusing on improved connection options, diagnostics, and performance. The main updates include support for connecting to an external Copilot CLI server (eliminating local CLI startup latency), detailed timing diagnostics for requests, and pre-warming of the Copilot client and session to reduce initial request latency. There are also related configuration and API changes to enable these features.

Copilot CLI Connection Enhancements:

• Added support for connecting to an already-running Copilot CLI server via cliUrl, as an alternative to spawning a local CLI process. This is configurable via the new cliUrl option in both code and config files, and is mutually exclusive with cliPath. [1] [2] [3] [4] [5] [6] [7] [8]

Performance and Diagnostics:

• Introduced detailed timing diagnostics for Copilot client operations (startup, session creation, request handling, etc.), with logging controlled via debug namespaces and environment variables. This includes breakdowns of model time versus SDK/session overhead and logs of the final prompt sent to the model. [1] [2] [3] [4] [5] [6] [7] [8]

Pre-warming and Startup Optimization:

• Added logic to pre-warm the Copilot client and session at host startup when Copilot is the active provider, significantly reducing latency for the first user-visible request. This is triggered automatically from runtime configuration setup. [1] [2] [3] [4]

API and Configuration Changes:

• Updated the Copilot API settings and related types to include the new cliUrl option, and exported new pre-warming functions from the main API surface.

Configuration Handling:

• Updated configuration parsing and sample files to support the new cliUrl option, ensuring it can be set via environment variables or YAML config. [1] [2] [3] [4]

These changes collectively improve Copilot performance, flexibility, and observability, especially in environments where process startup latency is a concern.