copilot readme update.

Author: sipsorcery

examples/GetStarted/README.md

@@ -1,55 +1,105 @@
-# OpenAI Realtime WebRTC Get Started Example
+# OpenAI Realtime SIP Get Started Example
 
-This is a minimal WebRTC application demonstrating interaction with OpenAI's [Realtime API](https://platform.openai.com/docs/guides/realtime-webrtc). It sets up a peer connection and streams audio from a Windows audio device. Once connected, it sends a message to start the conversation and prints transcription results from both the user and the assistant.
+This example demonstrates placing a SIP call to OpenAI's Realtime SIP endpoint and then upgrading that call to a realtime WebSocket session after an incoming webhook. Audio is captured from the default Windows input/output devices (via SIPSorcery + Windows audio endpoint) and sent using PCM (Opus is not currently negotiated successfully as of 05 Sep 2025).
 
-> ⚠️ **Note**: As of 10 May 2025, this example successfully establishes an audio stream and receives data channel messages. However, echo cancellation is not implemented—use a headset or ensure your audio device supports echo cancellation to avoid feedback loops.
+> ⚠️ Note (05 Sep 2025): The example successfully places a SIP call, receives the webhook, accepts the call, and establishes a realtime WebSocket. Echo cancellation is NOT implemented. Use a headset or a device with hardware echo cancellation to avoid the assistant talking to itself.
 
-## Features
+## What This Sample Does
 
-- Establishes a WebRTC connection with OpenAI's realtime endpoint
-- Streams audio directly from Windows devices
-- Sends a response prompt to trigger conversation
-- Handles and logs transcription deltas and completions for both user and assistant
+1. Starts a minimal ASP.NET Core web server to receive OpenAI webhook callbacks at `/webhook`.
+2. Places an outbound SIP TLS call to `sip.api.openai.com` using your OpenAI Project ID as the user part: `<PROJECT_ID>@sip.api.openai.com`.
+3. Waits for OpenAI to POST a webhook containing the `call_id`.
+4. Accepts the call via `POST /v1/realtime/calls/{call_id}/accept`.
+5. Opens a realtime WebSocket: `wss://api.openai.com/v1/realtime?call_id=...`.
+6. Sends an initial `response.create` instruction ("Say Hi.").
+7. Logs all incoming WebSocket text messages (JSON events from OpenAI).
+8. Streams audio between your local microphone/speakers and OpenAI (PCM).
 
 ## Requirements
 
-- Windows OS with audio devices
-- [.NET 8.0 SDK](https://dotnet.microsoft.com/en-us/download/dotnet/8.0)
-- OpenAI API key with access to the Realtime API
+- Windows OS (for WindowsAudioEndPoint in this demo)
+- .NET 8.0 SDK
+- OpenAI API key with Realtime + SIP access
+- OpenAI Project ID (e.g. `proj_...`)
+- A publicly accessible HTTPS endpoint for webhooks (ngrok recommended)
 
-## Getting Started
+## Environment Variables
 
-1. **Set your OpenAI API key as an environment variable**:
+Set these before running:
 
-```bash
+Windows (cmd.exe):
+```
 set OPENAI_API_KEY=your_openai_key
+set OPENAI_PROJECT_ID=your_openai_project_id
 ```
 
-2. **Run the application**:
+PowerShell:
+```
+$env:OPENAI_API_KEY="your_openai_key"
+$env:OPENAI_PROJECT_ID="your_openai_project_id"
+```
 
-```bash
-dotnet run
+## Exposing the Webhook (ngrok)
+
+1. Reserve / configure a domain in the ngrok dashboard (recommended) or use a temporary forwarding URL.
+2. In the OpenAI dashboard: Settings -> Webhooks -> Add webhook
+   - URL: `https://<your-ngrok-domain>/webhook`
+3. Start ngrok to forward to the Kestrel HTTPS port from `launchSettings.json` (default shown there is `https://localhost:53742`):
+```
+ngrok http --url=<your-ngrok-domain> https://localhost:53742
 ```
 
-3. **Interact**:
+## Run
+
+```
+dotnet run
+```
 
-Speak into your microphone and observe the transcription logs for both your voice and the assistant’s responses.
+You should see logs indicating:
+- Web server started
+- SIP call attempt to `<PROJECT_ID>@sip.api.openai.com;transport=tls`
+- Incoming webhook with `call_id`
+- Accept POST success
+- WebSocket connected and subsequent JSON event logs
 
 ## File Overview
 
 ### Program.cs
+Core sample logic:
+- Configures Serilog logging.
+- Validates `OPENAI_API_KEY` and `OPENAI_PROJECT_ID` env vars.
+- Registers an HTTP POST `/webhook` endpoint to receive call events.
+- On webhook: extracts `call_id`, sends accept request, starts WebSocket task.
+- Initiates SIP call using SIPSorcery (`SIPUserAgent`, `VoIPMediaSession`).
+- Opens WebSocket and sends an initial `response.create` instruction.
+- Streams and logs incoming WebSocket messages.
+
+### launchSettings.json
+Specifies the local HTTPS port (used for your ngrok forwarding target).
+
+## Audio Notes
+
+- Example uses `WindowsAudioEndPoint` with default input/output devices.
+- Opus was attempted (commented line) but PCM only negotiates successfully at time of writing.
+- No echo cancellation; prefer headset.
+
+## Customizing
+
+- Change initial instruction: edit the anonymous object `responseCreate` in `StartWebSocketConnection`.
+- Provide different model/instructions for acceptance by altering `call_accept` record fields.
+- Add parsing of WebSocket JSON events to handle partial transcripts, tool calls, etc.
+
+## Troubleshooting
 
-Contains the core application logic:
-- Initializes the OpenAI WebRTC endpoint
-- Connects audio from the default Windows input device
-- Sends session updates and creates a response to initiate conversation
-- Logs transcription updates and completions
+- No webhook received: verify ngrok is running and the correct HTTPS URL is registered in OpenAI settings.
+- 401 / auth errors: confirm `OPENAI_API_KEY` environment variable is set in the same shell you run `dotnet run`.
+- SIP call fails immediately: check outbound TLS (firewall/proxy) and that your Project ID is correct.
+- WebSocket closes: inspect logged close status and ensure accept POST succeeded.
 
-## Notes
+## Security
 
-- Echo cancellation is not handled in this demo. If you're using speakers, OpenAI may end up responding to itself. Use a headset for clean operation.
-- Transcription is enabled using the `Whisper1` model.
-- This demo is part of the `SIPSorcery.OpenAI.WebRTC` library.
+- Do NOT hardcode your API key. Use environment variables or a secure secrets store.
+- Restrict exposure of your webhook endpoint. ngrok URLs are public; rotate as needed.
 
 ## License