CLI Reference

All Timber commands and their options.

timber load <model_path>

Compile and cache a model locally.

timber load model.json
timber load model.json --name fraud-detector
timber load pipeline.pkl --format sklearn

Option	Default	Description
--name NAME	Filename stem	Name to register the model as
--format FMT	Auto-detect	Force input format: xgboost, lightgbm, sklearn, catboost, onnx

What it does:

1. Auto-detects (or uses forced) format
2. Parses model into Timber IR
3. Runs 6 optimization passes
4. Emits C99 source code
5. Compiles a native shared library (.so / .dylib)
6. Caches everything in ~/.timber/models/<name>/

timber serve <name>

Serve a loaded model over HTTP.

timber serve my-model
timber serve my-model --port 8080
timber serve my-model --host 127.0.0.1 --port 9000

Option	Default	Description
--host HOST	0.0.0.0	Bind address
--port PORT	11434	Bind port

The server exposes:

• POST /api/predict — run inference
• POST /api/generate — alias for predict (Ollama compat)
• GET /api/models — list loaded models
• GET /api/model/:name — model metadata
• GET /api/health — health check

timber list

List all cached models in a table.

timber list

Output:

NAME            TREES  FEATURES  FRAMEWORK  SIZE
breast-cancer   50     30        xgboost    58.0 KB
fraud-detector  100    45        lightgbm   112.3 KB

timber remove <name>

Remove a cached model and all its artifacts.

timber remove my-model

timber compile

Compile a model to C99 source without caching in the store.

timber compile --model model.json --out ./dist/
timber compile --model model.json --target targets/x86_64_avx2.toml --out ./dist/
timber compile --model model.json --calibration-data train.csv --out ./dist/

Option	Default	Description
--model PATH	Required	Model artifact path
--format FMT	Auto-detect	Input format hint
--target PATH	x86_64_generic	Target spec TOML file
--out DIR	./dist	Output directory
--calibration-data PATH	None	CSV for branch sorting pass

timber inspect <model_path>

Print model summary without compiling.

timber inspect model.json

timber validate

Validate compiled output against reference predictions.

timber validate --artifact ./dist/ --reference model.json --data test.csv

Option	Default	Description
--artifact DIR	Required	Compiled artifact directory
--reference PATH	Required	Original model file
--data PATH	Required	Validation CSV
--tolerance FLOAT	1e-5	Max allowed absolute error

timber bench

Benchmark inference performance.

timber bench --artifact ./dist/ --data test.csv --batch-sizes 1,10,100,1000

Environment Variables

Variable	Default	Description
TIMBER_HOME	~/.timber	Model store root directory
CC	gcc	C compiler to use for compilation