LlmConnector Development for Copilot and Codex - C# App (incl source available)

Project: LlmConnector - extend from LM Studio to Copilot and Codex Context (current state) - We already have a working LM Studio integration, including chat and model listing via /v1/models. - The UI is a WPF desktop app (.NET, C#) with a provider dropdown, base URL, model, API key, system prompt, and buttons for "Load models" and "Test connection". - The app uses an ILlmClient interface and ILlmClientFactory in LlmClients.cs. LM Studio implementation is done and should remain unchanged. - The GitHub Copilot and Codex client classes are present but currently stubs or incomplete. Tech stack (short) - C# / .NET (WPF desktop app) - REST API calls via HttpClient - Configuration stored in JSON settings file Goal / priority We want the app to support more than LM Studio. Highest priority is GitHub Copilot. Codex support is a secondary goal, only if feasible given available endpoints and auth. Scope of work Implement production-ready clients for: 1) GitHub Copilot (priority 1) 2) Codex (priority 2, if possible) Deliverables - Fully working Copilot client implementation for chat + model listing. - Codex client implementation for chat + model listing (if endpoint and auth are available). - Minimal, clear error handling that surfaces failures in the UI/logs. - Short code comments indicating where to configure provider-specific details (endpoint, headers, auth token source). - Keep LM Studio client unchanged. Constraints / notes - Use the existing interfaces in [login to view URL] (ILlmClient and ILlmClientFactory). - The UI already calls Load Models (GET /v1/models equivalent) and Send Chat. - The client must return a list of model IDs and a single string response for the assistant. - Avoid breaking changes to existing UI flow. Test cases (manual or automated) 1) GitHub Copilot models list: With Copilot configured and reachable, clicking "Load models" returns at least one model ID and populates the model dropdown; no 4xx/5xx errors. 2) GitHub Copilot chat: With a valid configuration, sending "Hello" returns a non-empty assistant response and logs it without exceptions. 3) Codex models + chat: With a valid OpenAI-compatible endpoint and API key, "Load models" returns model IDs and a chat request returns a non-empty response without errors. Delivery: 1. A short video how you show me you could connect with Copilot and list models and chat with models with all 3 sample messages. 2. Updated source including description of the changes

Project ID: 40281139