eosllm-vscode 0.1.0

eosllm-vscode 0.1.0 — initial VS Code extension release

Local-first LLM chat in VS Code, powered by the
eosllm inference engine.
Wraps the eosllm-cli --stream-jsonl binary and renders streamed
tokens in a chat webview as they're generated. Everything runs
locally — no telemetry, no network egress.

Features (v0.1.0)

• Activity-bar chat webview — type prompt, watch tokens stream in.
• Settings: eosllm.modelPath, eosllm.maxTokens, eosllm.binDir.
• Commands: eosllm.runPrompt, eosllm.openChat.
• Status-bar tokens-per-second readout while streaming.
• First-launch postinstall fetches the platform-matching engine binary
  from the engine's GitHub Releases page; cached under the extension's
  global storage so subsequent launches start instantly.

Install

The Marketplace + Open VSX publish steps are wired in CI but not
yet active (the maintainer's VSCE_PAT / OVSX_PAT are not set in
this repo's secrets). Sideload the attached .vsix for now:

code --install-extension eosllm-vscode-0.1.0.vsix

(or Extensions panel → … menu → Install from VSIX… and pick
the downloaded file)

After install:

1. Download an engine release binary from
   v0.2.0
   for your platform, or wait for the postinstall script to fetch it.
2. Settings (Ctrl/Cmd+,) → search eosllm → set
   eosllm.modelPath to a Llama-class GGUF on disk.
3. Open the eosllm activity-bar view → type a prompt → Send.

Privacy

• No telemetry, no analytics, no off-machine network calls.
• The only network call ever made is the one-time fetch of the
  engine binary from github.com on first launch (skip-able by
  setting eosllm.binDir to a directory that already contains
  eosllm-cli).
• Permissions requested: none beyond what VS Code grants every
  extension by default.

Requires

• VS Code 1.85.0 or later.
• An eosllm-cli binary (auto-fetched on first launch from the
  matching engine release; supported platforms: linux x86_64,
  linux aarch64, macOS arm64+x86_64 universal, windows x86_64).

Troubleshooting

• "eosllm-cli binary not found" — the postinstall fetch failed
  (no network on first launch?) or you set eosllm.binDir to the
  wrong place. Either re-trigger by reloading the window, or download
  an engine tarball from
  v0.2.0
  and point eosllm.binDir at its bin/ directory.
• macOS: the bundled eosllm-cli is unsigned in this engine
  release. The first time it runs, run
  xattr -dr com.apple.quarantine eosllm-cli from the bin directory
  to clear Gatekeeper.

Changelog

CHANGELOG.md.