v0.2.0

Changelog

• 0e9794c chore: rename repo to hyperstudyio/hyperstudy-agent; add changelog and docs release-notes sync

v0.1.3

Changelog

• 9e68489 feat(serve): add --mtp speculative decoding for gemma4; drop deprecated --no-thinking (#1)

v0.1.2

Changelog

• bed431f feat(serve): --no-thinking flag disables reasoning-model thinking for faster turns

v0.1.1

Changelog

• 1bbc88e fix(serve): clarify LAN URL is local-only; steer to tunnel for hosted HyperStudy

v0.1.0

Changelog

• 8e1efd7 docs: move install instructions to top; add build-from-source option
• 9069b7a fix(verify): 512-token budget for reasoning models + accurate Gemma 4 caveat (single-turn works; #25072 is multi-turn-only)
• 8db8299 feat(serve): named model presets (qwen3.6-moe, gemma4-moe, gemma4-4b) with tool-calling caveats
• 0480b31 fix: ready-wait races child exit, silence cobra double-errors, wrong-key diagnosis in verify, gofmt/mod tidy, README rows
• fae4fe6 fix(ci): use go-version-file so workflows track go.mod toolchain
• cbf56d4 chore: goreleaser, CI workflows, README, installer
• f02b8ed feat: tunnel command (cloudflared quick tunnel)
• c742b56 feat: verify command
• 47ace84 fix(verify): tool-call args validated against respond schema; p95 threshold branches tested
• 3f6d0cb feat: contract verification checks (reachability, auth, tool calls, concurrency)
• ad26007 feat: serve command
• 6718696 feat: llama-server location, launch args, readiness poll
• 60f6409 feat: model tier selection ladder
• 7caf546 feat: hardware detection (RAM, NVIDIA VRAM, Spark)
• 3b919c0 fix(config): align config package with spec (APIKey/Model/Port, EnsureKey bool, zero-value Load)
• e8fc357 feat(config): internal config package with key persistence
• d4dc289 chore: add MIT license, trim gitignore to spec
• 8447772 feat: scaffold hyperstudy-agent CLI