v0.1.3

Release Notes

Added

• (telemetry) push opt-in performance rollups from deployments (#291)

Install amicu-cli 0.1.3

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/andypalmi/amicu/releases/download/v0.1.3/amicu-cli-installer.sh | sh

Install prebuilt binaries via powershell script

powershell -ExecutionPolicy Bypass -c "irm https://github.com/andypalmi/amicu/releases/download/v0.1.3/amicu-cli-installer.ps1 | iex"

Install prebuilt binaries via Homebrew

brew install andypalmi/tap/amicu

Download amicu-cli 0.1.3

File	Platform	Checksum
amicu-cli-aarch64-apple-darwin.tar.xz	Apple Silicon macOS	checksum
amicu-cli-x86_64-apple-darwin.tar.xz	Intel macOS	checksum
amicu-cli-x86_64-pc-windows-msvc.zip	x64 Windows	checksum
amicu-cli-aarch64-unknown-linux-musl.tar.xz	ARM64 MUSL Linux	checksum
amicu-cli-x86_64-unknown-linux-musl.tar.xz	x64 MUSL Linux	checksum

Verifying GitHub Artifact Attestations

The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:

gh attestation verify <file-path of downloaded artifact> --repo andypalmi/amicu

You can also download the attestation from GitHub and verify against that directly:

gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo andypalmi/amicu

v0.1.2

Release Notes

Added

• (windows) Windows release build, PowerShell installer and docs (#287)
• (windows) compile and test cleanly on Windows (#288)
• (windows) run local llama.cpp deployments as Task Scheduler tasks (#289)
• (windows) detect host hardware on Windows (#286)

Install amicu-cli 0.1.2

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/andypalmi/amicu/releases/download/v0.1.2/amicu-cli-installer.sh | sh

Install prebuilt binaries via powershell script

powershell -ExecutionPolicy Bypass -c "irm https://github.com/andypalmi/amicu/releases/download/v0.1.2/amicu-cli-installer.ps1 | iex"

Install prebuilt binaries via Homebrew

brew install andypalmi/tap/amicu

Download amicu-cli 0.1.2

File	Platform	Checksum
amicu-cli-aarch64-apple-darwin.tar.xz	Apple Silicon macOS	checksum
amicu-cli-x86_64-apple-darwin.tar.xz	Intel macOS	checksum
amicu-cli-x86_64-pc-windows-msvc.zip	x64 Windows	checksum
amicu-cli-aarch64-unknown-linux-musl.tar.xz	ARM64 MUSL Linux	checksum
amicu-cli-x86_64-unknown-linux-musl.tar.xz	x64 MUSL Linux	checksum

Verifying GitHub Artifact Attestations

The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:

gh attestation verify <file-path of downloaded artifact> --repo andypalmi/amicu

You can also download the attestation from GitHub and verify against that directly:

gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo andypalmi/amicu

v0.1.1

Release Notes

Added

• (cli) show old -> new values per changed field in plan (#283)
• (cli) Qwen3.5 examples, presets and tiered starters (#281)
• (cli) empty init, --example starters, host-aware plan and amicu catalog (#279)

Fixed

• (cli) redact secrets in plan runtime settings and launch line (#284)

Install amicu-cli 0.1.1

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/andypalmi/amicu/releases/download/v0.1.1/amicu-cli-installer.sh | sh

Install prebuilt binaries via Homebrew

brew install andypalmi/tap/amicu

Download amicu-cli 0.1.1

File	Platform	Checksum
amicu-cli-aarch64-apple-darwin.tar.xz	Apple Silicon macOS	checksum
amicu-cli-x86_64-apple-darwin.tar.xz	Intel macOS	checksum
amicu-cli-aarch64-unknown-linux-musl.tar.xz	ARM64 MUSL Linux	checksum
amicu-cli-x86_64-unknown-linux-musl.tar.xz	x64 MUSL Linux	checksum

Verifying GitHub Artifact Attestations

The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:

gh attestation verify <file-path of downloaded artifact> --repo andypalmi/amicu

You can also download the attestation from GitHub and verify against that directly:

gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo andypalmi/amicu

v0.1.0

Release Notes

Added

• (engine) validate llama.cpp fields, KV default, Modal region pricing (#250)
• (config) memory_check = "warn" to bypass the memory-fit rejection per deployment (#247)
• per-deployment endpoint API keys + shared post-deploy exposure check (#200)
• (hcl) secret-safety enforcement + free-text secret warning (#110)
• (cli) provider-switch destroy step in plan, state migrations and unknown-field handling (#244)
• vendor-agnostic hardware metrics in amicu modal metrics (#175)
• (api) opt-in Sentry error reporting; catalogue and stderr follow-ups (#262)
• (engine) catalogue every Apple Silicon chip; fix broken-pipe panic (#261)
• (server) wire state-backend selection in build_app (#114)
• (state) encryption-at-rest seam for the SaaS state backend (#113)
• (store) Postgres StateBackendAdapter implementation (#123) (#108)

Fixed

• (cli) project discovery, non-interactive prompts, model alias and purge leftovers (#263)
• (cli) show memory estimates rounded instead of raw bytes (#259)
• (engine) count KV cache in memory fit, catalog gaps, review follow-ups (#256)
• (plan) never silently drop the KV cache from the memory estimate (#198)
• (cli) hardware test follow-ups (#258)
• (engine) llama.cpp on CUDA, persistent systemd units and vLLM VRAM check (#245)
• (local) make local apply work under launchd/systemd (#205)
• (plan) parse MoE parameter counts (NxMB, total-A-active) (#199)
• (cli) never panic on closed stderr; quieter stop-wait note (#265)
• (cli) launchd stop wait, project root marker and baseline follow-ups (#264)
• (local) wait for launchd bootout before bootstrap, destroy and drift (#257)
• (state) workspace-safe Modal names, rollback app id, launchd presence, outbox test (#251)
• explicit endpoint exposure — Modal proxy auth fail-closed, loopback bind, no silent local fallback (#194)

Other

• remove references to private repos (#248)
• change detection fingerprints and pure partial-state merge (#235)
• Modal autoscaling via provider-agnostic scaling block (#176)
• Plan shows the real Modal launch argv; add fp8 quantization (#168)
• Reject unknown attributes and blocks in HCL (AMICU-SCHEMA-003) (#162)
• Fix new stable clippy assert_is_empty lint (#159)
• Typed vLLM enforce_eager field (#150)
• Modal catalog-only provider (#153)
• Typed llama.cpp multimodal projector block (#149)
• Typed llama.cpp speculative decoding block with draft-model graph edge (#148)
• vLLM renderer + runtime dispatch + fit checks for vLLM deployments (#147)
• Typed llama.cpp chat template, reasoning format, GGUF sha256 and build pin (#146)
• Consumer/on-prem GPU catalog + accelerator bandwidth field (#145)
• Typed llama.cpp sampling/KV/batch fields + KV-cache estimator (#144)
• Runtime escape hatch: extra_args and env on llama.cpp/vLLM blocks (#140)
• Customize endpoint + candidate review UI (#136)
• remove all em dashes (U+2014) repo-wide (#106)
• Management API: OpenAPI source-of-truth (utoipa) + validate/plan over HTTP (#70)
• Drop TASK-NNN tags for GitHub issue numbers (#61)
• Add optional adapter-owned zone (AZ) as a third region level (#56)
• TASK-067 (3/3): Retype DataClass.residency to Vec, check by jurisdiction (#47)
• TASK-067 (2/3): Retype HardwareProfile.region to Option (#46)
• TASK-067 (1/3): Move Region/Residency into amicu-core (#45)
• Provider capability and pricing sync (#39)
• Region / data-residency validation in the planner (#38)
• Provider health and error normalization (#32)
• Cost-estimation primitives (#30)
• amicu-core polish: ByteSize display, AdapterError API, catalog codes, SecretValue compile-fail guard (#13)
• Define secret-manager adapter trait (#7)
• Common adapter contract — error model, registry, dispatch, planning/apply separation (#5)
• Error-code catalog and plan JSON contract (#3)
• Canonical graph, resource schemas, and versioning/migration policy (#4)
• scaffold Rust workspace, crate boundaries, and CI gates (#1)
• durable saves, backup recovery, secret lint, gitignore (#231)
• amicu-hcl — validate reference edge values in format_graph (#15)
• amicu-hcl format_graph identifier safety (#11)
• HCL parse -> graph -> format round-trip (#6)
• amicu destroy with write-ahead stop records (#229)
• stream image build logs to stderr; show GPU count in plan (#172)
• project-scoped host labels, state v3 and validated cleanup on destroy (#237)
• host drift refresh and Modal rollback on failed update (#236)
• real runtime metrics + amicu modal metrics (#169)
• harden secret-bearing llama.cpp service files (#21)
• process management (launchd/systemd) and health checks (#20)
• llama.cpp adapter — binary detection, GGUF download, command generation (#19)
• harden RequestRecord privacy guard and Resolved invariant (#23)
• local endpoint registration, metrics, and privacy-safe logging (#22)
• discovery — Intel-Mac CPU brand string and narrowed Apple-Silicon fallback (#18)
• hardware discovery (Apple Silicon, CPU-only) (#17)
• Define runtime adapter trait (#10)
• app_status works for Server functions (FlashContainerList fallback) (#166)
• redact gRPC metadata from errors; python on PATH for vLLM image (#165)
• support llama.cpp (llama-server) on Modal, including GGUF delivery (#163)
• send proxy auth token on endpoint probes, explain 401s (#164)
• Modal follow-ups: logs, existing function id, proxy regions, web provider list (#161)
• route amicu apply for vLLM on Modal through ModalSupervisor (#160)
• log streaming, deployment status/health, and stop (#154)
• deploy path — image, web-server Function, publish, stop (#152)
• gRPC plumbing for Modal + amicu modal auth-check (#151)
• orphan-proof stop records and conformance; telemetry data model (#233)
• (deps) bump russh from 0.62.7 to 0.63.2 (#155)
• Management API: runnable server binary (amicu-server) (#74)
• Bump russh 0.54 -> 0.62.7 to clear 12 SSH advisories ([#59](https://github...

v0.1.0-rc.1

Release Notes

Install amicu-cli 0.1.0-rc.1

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/andypalmi/amicu/releases/download/v0.1.0-rc.1/amicu-cli-installer.sh | sh

Install prebuilt binaries via Homebrew

brew install andypalmi/tap/amicu

Download amicu-cli 0.1.0-rc.1

File	Platform	Checksum
amicu-cli-aarch64-apple-darwin.tar.xz	Apple Silicon macOS	checksum
amicu-cli-x86_64-apple-darwin.tar.xz	Intel macOS	checksum
amicu-cli-aarch64-unknown-linux-musl.tar.xz	ARM64 MUSL Linux	checksum
amicu-cli-x86_64-unknown-linux-musl.tar.xz	x64 MUSL Linux	checksum

Verifying GitHub Artifact Attestations

The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:

gh attestation verify <file-path of downloaded artifact> --repo andypalmi/amicu

You can also download the attestation from GitHub and verify against that directly:

gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo andypalmi/amicu