v0.1.0

v0.1.0 — 2026-04-23
Initial release.
Core

• GitHub Copilot as a conversation agent in Home Assistant's Assist voice pipeline
• Zero pip dependencies — all API calls use raw aiohttp
• OpenAI-compatible streaming SSE responses with progressive text display in the HA UI
  Authentication
• GitHub Device Flow OAuth (same flow as VS Code) — no API keys to copy/paste
• Automatic Copilot token exchange with proactive refresh (60-second buffer before expiry)
• Retry-on-401 with transparent token renewal
• Re-authentication flow when credentials become invalid
  Models
• Dynamic model discovery from the Copilot /models API (filtered to chat-capable model families)
• Model selection during initial setup and in the options flow
• Default model: gpt-4.1
  Conversation
• Streaming chat completions via chat_log.async_add_delta_content_stream
• Tool calling support for Home Assistant device control via HA's native LLM API (up to 10 round-trips)
• Configurable system prompt with Jinja2 template support ({{ ha_name }}, {{ now() }}, etc.)
• Configurable temperature (0.0-2.0) and max tokens (64-16384)
• Continue conversation mode (Experimental) — keeps the microphone open after question responses
  Infrastructure
• Shared CopilotRuntimeData with thread-safe token management via asyncio.Lock
• Model fetching extracted to shared utils.py (used by both config flow and runtime)
• Payload sanitization to prevent JSON serialization errors from HA's internal types
• HACS-compatible repository structure