Kev: Train and Run Self-Hosted Jev-Like Decision Models on Qwen 3.5/3.8
Table of Contents4sections
Kev: Train and Run Self-Hosted Jev-Like Decision Models on Qwen 3.5/3.8
Jared Palmer—creator of Formik and Turborepo—has launched Kev, an open-source framework bringing structured decision modeling to self-hosted infrastructure. Accumulating 7,200+ Stars in record time, Kev provides an open alternative to proprietary routing and gating APIs.
🚀 Fast Download via GH-Proxy
Accelerate Git clone operations and Hugging Face weights acquisition via GH-Proxy:
👉 Direct GH-Proxy Acceleration for Kev
(Alternative URL: https://gh-proxy.com/https://github.com/jaredpalmer/kev)
Quick setup:
git clone https://gh-proxy.com/https://github.com/jaredpalmer/kev.git
cd kev
pip install -e .
💡 The Problem Kev Solves
Proprietary decision APIs (such as TypeSafe Jev) provide sub-second routing, but lock developers into external cloud dependencies. For regulated environments, privacy compliance, and latency-critical on-prem setups, sending continuous routing queries over the public internet is untenable.
Kev solves this by packaging the entire pipeline:
- Built on Qwen 3.5 & 3.8 Foundations: Fine-tuned on dense semantic decision pairs and structured control tasks.
- Comprehensive Size Spectrum: Released in 0.8B, 4B, 9B, and 27B checkpoints to match edge devices, workstations, and high-throughput servers.
- Self-Contained Training Harness: Includes data generation tooling, LoRA configurations, and GGUF / vLLM runtime recipes.
🛠 Quick Inference Example
from kev import KevModel
# Load the local 4B decision model
model = KevModel.from_pretrained("jaredpalmer/kev-4b-qwen")
# Route user prompt without generative latency
result = model.classify(
text="Can you review this pull request and check for race conditions?",
labels=["code_review", "documentation", "issue_triage", "chat"]
)
print(f"Action: {result.label} (Score: {result.score:.3f})")
🎯 Target Use Cases
- On-Premise Enterprise Gateways: Enforce data sovereignty by routing sensitive internal telemetry locally.
- Air-Gapped Autonomous Systems: Power embedded robotics, edge defense, or healthcare devices that operate without constant WAN connectivity.
- Cost-Efficient Orchestration: Filter out 60%+ of mundane queries before delegating to multi-billion parameter foundation models.