kossisoroyce

kossisoroyce / timber

Public

Ollama for classical ML models. AOT compiler that turns XGBoost, LightGBM, scikit-learn, CatBoost & ONNX models into native C99 inference code. One command to load, one command to serve. 336x faster than Python inference.

pypi.orgprojecttimber-compiler c99 catboost compiler decision-trees gradient-boosting
458
13
100% credibility
Found Mar 02, 2026 at 440 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Timber compiles tree-based machine learning models from popular frameworks into optimized native C code and serves them via a simple local HTTP API for low-latency inference.

How It Works

1
🔍 Discover Timber

You learn about a handy tool that makes your trained prediction models run incredibly fast, like magic for everyday machine learning.

2
📦 Get it ready

You add Timber to your computer with a single simple instruction, and it's all set up in seconds.

3
⚡ Prepare your model

You share your ready-to-use prediction model with Timber, and it instantly turns it into a super-fast version.

4
🚀 Launch the helper

You start the easy web helper, and your model is now live and waiting for questions.

5
📊 Ask for predictions

You send your data over the web, and get back answers in the blink of an eye.

🎉 Lightning results

Your models now zip through predictions at native speed, perfect for quick decisions anywhere.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 440 to 458 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is timber?

Timber brings Ollama-style serving to classical ML models, compiling XGBoost, LightGBM, scikit-learn, CatBoost, and ONNX tree ensembles into native C99 inference binaries. Run `timber load model.json --name fraud-detector` to compile and cache, then `timber serve fraud-detector` for a local HTTP server with Ollama-compatible endpoints like `/api/predict` and `/api/models`. Python CLI delivers 336x faster single-sample inference than Python runtimes, with no Python deps at runtime.

Why is it gaining traction?

It slashes latency to microseconds and shrinks artifacts to ~48KB, beating ONNX Runtime or Treelite on edge hardware while mimicking Ollama's workflow for easy docker/github actions deploys. Reproducible benchmarks, audit reports, and MISRA-C output appeal to teams ditching Python serving overhead without losing dev simplicity.

Who should use this?

Fraud/risk engineers in high-throughput pipelines, IoT devs targeting gateways, or infra teams in finance/healthcare needing deterministic, auditable inference. Suits anyone integrating ollama github api patterns but with tree models over LLMs.

Verdict

Worth a spin for prod classical ML serving—CLI shines, docs include examples and benchmarks. 169 stars and 1.0% credibility flag alpha maturity; validate outputs rigorously before commit.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.