These seven samples use the published Foundry Local Rust SDK 2.2.1.
All crates pin foundry-local-sdk = "=2.2.1" and set the application name to
foundry_local_samples.
Important
These samples use foundry-local-sdk from crates.io.
.cargo/config.toml also selects native Runtime 2.2.1, which the SDK acquires during the build.
No local SDK source build is required. See the
get-started documentation for versioned setup
guidance and additional examples.
- A supported Foundry Local platform
- The Rust toolchain
| Directory | Demonstrates |
|---|---|
hello-world |
Manager setup, execution-provider registration, catalog/model lifecycle, and native chat inference |
realtime-transcription |
Native microphone streaming with AudioSession and ItemQueue |
embeddings |
Batched embeddings and similarity comparison with the native EmbeddingsSession |
tool-calling |
Native ChatSession tool definitions, tool-call items, and tool-result items |
web-server |
Starting the built-in service, discovering its URL, calling an OpenAI-compatible endpoint, and stopping it |
vision |
Sending text and a local image through a native ChatSession |
byom |
Registering existing local model assets and running native chat |
Run a package from this directory:
cd samples/rust
cargo run -p hello-worldReplace hello-world with any package name from the table. Real-time
transcription uses the default microphone and runs until Ctrl+C:
cargo run -p realtime-transcriptionVision accepts a catalog alias with task vision-language-chat and a local JPEG or PNG image. BYOM accepts an
already exported, compatible ONNX Runtime GenAI chat model directory containing genai_config.json, weights, and
tokenizer files:
cargo run -p vision -- VISION_MODEL_ALIAS /path/to/image.png
cargo run -p byom -- /path/to/model-directoryThe BYOM sample unregisters its model on normal completion without deleting your files.
The first run may download execution providers and model assets. Later runs reuse cached models.