Open source · MIT · runs locally
Gemma, GLM and Qwen,
on your machine.
The Tara Vision inference engine: a local, private and OpenAI-compatible server built on llama.cpp, with one pinned build per release, a curated catalog, tuned presets and measured memory. Made for taracode, open to any client. Nothing leaves your machine.
$ taraengine runtime install
one binary, one pinned runtime, your models. pick your platform:
$ BASE=https://github.com/tara-vision/taraengine/releases/latest/download; curl -fsSLO $BASE/checksums.txt -O $BASE/checksums.txt.sigstore.json
$ cosign verify-blob --bundle checksums.txt.sigstore.json --certificate-oidc-issuer https://token.actions.githubusercontent.com --certificate-identity-regexp 'https://github.com/tara-vision/taraengine/.github/workflows/release.yml@refs/tags/v.*' checksums.txt
$ shasum -a 256 --ignore-missing -c checksums.txt && chmod +x taraengine-darwin-arm64 && sudo mkdir -p /usr/local/bin && sudo mv taraengine-darwin-arm64 /usr/local/bin/taraengine
$ BASE=https://github.com/tara-vision/taraengine/releases/latest/download; curl -fsSLO $BASE/checksums.txt -O $BASE/checksums.txt.sigstore.json
$ cosign verify-blob --bundle checksums.txt.sigstore.json --certificate-oidc-issuer https://token.actions.githubusercontent.com --certificate-identity-regexp 'https://github.com/tara-vision/taraengine/.github/workflows/release.yml@refs/tags/v.*' checksums.txt
$ sha256sum --ignore-missing -c checksums.txt && chmod +x taraengine-linux-amd64 && sudo mkdir -p /usr/local/bin && sudo mv taraengine-linux-amd64 /usr/local/bin/taraengine
$ BASE=https://github.com/tara-vision/taraengine/releases/latest/download; curl -fsSLO $BASE/checksums.txt -O $BASE/checksums.txt.sigstore.json
$ cosign verify-blob --bundle checksums.txt.sigstore.json --certificate-oidc-issuer https://token.actions.githubusercontent.com --certificate-identity-regexp 'https://github.com/tara-vision/taraengine/.github/workflows/release.yml@refs/tags/v.*' checksums.txt
$ sha256sum --ignore-missing -c checksums.txt && chmod +x taraengine-linux-arm64 && sudo mkdir -p /usr/local/bin && sudo mv taraengine-linux-arm64 /usr/local/bin/taraengine
$ sudo rpm -i taraengine_0.2.0_linux_amd64.rpm
then: taraengine runtime install fetches the pinned llama.cpp build for this machine. the homebrew cask and the installer script come with the next release.
$ taraengine doctor
the engine owns everything above the kernels; the math is ggml through llama.cpp
Pinned runtime
One llama.cpp build per release, hashed against the lock before it is unpacked and checked again every time the engine starts.
✓ b11514 verified
Curated catalog
Models pinned by size and sha256, with the projector and the draft model they come with, and a KV cache chosen for each on an RTX 5090 board.
✓ sha256 verified · 7.2 GB
OpenAI and Anthropic APIs
llama-server's surface, proxied as it streams: chat, completions, embeddings and messages, with a Bearer key when you set one.
✓ 200 · 248 tok/s
Tools, thinking and vision
Tool calls, a thinking switch with effort levels, and images where the model reads them.
✓ reasoning_content
Measured
Every chat answer carries its timings, and ps shows what a loaded model holds in memory, measured on NVIDIA and estimated on a Mac.
✓ 10.0 GiB measured
Privacy-first
Localhost by default and a Bearer key when you set one; no telemetry, no update checks, and two hosts contacted only when you ask for a download.
✓ nothing left your machine
Made for taracode
taracode's next release finds the engine at its address with no configuration; any OpenAI-compatible client can use it now.
✓ http://127.0.0.1:8272
$ taraengine doctor --privacy
what leaves your machine: nothing.
No account
No sign-in, no license keys, no waitlist. Install and run.
Zero telemetry
Nothing is collected, ever. Read the data transparency page.
Fully local
Once a model is pulled, the engine works without internet.
$ taraengine sponsor
free, MIT-licensed, no telemetry. sponsors keep the releases, the catalog and the runtime pins coming.
$ install taraengine