v0.3.2

[0.3.2]

Added

• Local & Enterprise Model Endpoints — bring your own Ollama, LM Studio, vLLM, llama.cpp, SGLang, LiteLLM, or Olla fleet, or an internal enterprise/private-cloud gateway, into the same Copilot Chat model picker as Feima-hosted models
  • Auto-discovery of well-known local ports on startup, manual registration (Feima Local Models: Add Model Endpoint) for endpoints discovery can't reach, and team-shared .feima/endpoints.json (URLs only, never secrets)
  • New Local & Enterprise Models view in the Explorer sidebar — live per-endpoint health indicators, model overrides, and an endpoint editor for correcting or adding model metadata
  • Feima Local Models: Refresh Models plus per-endpoint test-connection/remove/edit commands
• Feima Auto — a new picker entry that automatically routes each request to the best available registered local/enterprise endpoint (local-first, balanced, or most-capable, via the feima.localModels.autoStrategy setting), disclosing which endpoint was used and why on every response
  • Sticks with the same endpoint across a conversation instead of re-deciding every message
  • Feima Local Models: Add My Feima-Hosted Models to Auto — registers your Feima-hosted models (using a Feima API key) as local endpoints so Auto can route to them too
• Lenient repair for malformed tool-call JSON from local/enterprise models, plus a Responses API endpoint option for providers that support it
• Qwen3.8 Flash — Alibaba Cloud, ~1M context, ultra-fast (free tier)
• HY4 Preview — Tencent Hunyuan, 1M context, thinking support (free tier)

Changed

• Renamed the local/enterprise router's picker entry from "Auto" to "Feima Auto", to avoid confusion with VS Code's own built-in Auto (which only sees GitHub-hosted models)

Full Changelog: v0.3.1...v0.3.2