Bud Decision Studio
Cross-platform desktop app to download eleven open decision models, run a Playground, and serve TypeSafe Jev API on http://127.0.0.1:8420.
Overview
Bud Decision Studio (github.com/BudEcosystem/Bud-Decision-Studio) is Bud Ecosystem's cross-platform desktop app described in the README as the LM Studio of Jev-like System One models. Source and UI live in the Bud-Decision-Studio repository. Installers, get.sh, get.ps1, and GitHub Releases ship from github.com/BudEcosystem/Bud-Decision-Engine, which the README badges as the download target. The project page budecosystem.com/open-source-projects/bud-decision-studio returned HTTP 200 on 2026-10-07. The Studio repo had no LICENSE file when checked the same day.
First launch checks hardware, installs a private Python engine with matching PyTorch, and lets you download eleven open models such as Kev 4B, Lev, Laya Multilingual, Jev-Omni, and CLM 8B. While the app runs, POST /v1/systemone and GET /v1/models on http://127.0.0.1:8420 work with typesafe-sdk when you change only the base URL.
Problem: Local decision models still require separate engine installs, weight downloads, and SDK base URL wiring before you can compare answers to hosted Jev.
Built for: Builders on macOS Apple Silicon, Windows 10/11, or Linux x64/ARM64 who want a desktop Playground and localhost System One API for open models.
First indexed on Jev Directory: 2026-09-30
Creator and team
- Name
- Bud Ecosystem
- Handle
- @BudEcosystem
How Jev is used
- Role in the product flow
- Local System One server for Playground, templates, history, eval, and train flows
- Primitives
- ChoiceScoreNoul
- State in
- Playground state text or JSON, template variables, or API JSON per studio-api.md.
- Decision out
- Per-question probabilities with act_threshold and needs_review flags on template runs.
- Install via Bud-Decision-Engine release or curl get.sh script
- Complete first-run engine setup on GPU or CPU
- Download models from the Models page; open Playground or POST /v1/systemone
- Point TypeSafeClient base_url at http://127.0.0.1:8420 with api_key local
Bud Decision Studio is GUI-first local Jev API hosting. ollaya-dev-ollaya is the Rust CLI and daemon on port 11435 with pull and MCP helpers. jaredpalmer-kev and receptron-laya remain upstream training and ONNX stories; Bud bundles consumer installs and compares published results next to Jev in the Models table. Hosted TypeSafe Jev is still the cloud baseline when you want zero local disk.
Sourced performance claims
- README Train table lists eleven models with before/after task accuracy on held-out examples and general decision deltas on an NVIDIA GB10, such as Kev 4B policy topics 77% to 82%.Source: github.com/BudEcosystem/Bud-Decision-Studio README Teach a model
- README Models table names eleven downloadable checkpoints with sizes from Julia 1 144M through Jev-Omni 12B multimodal.Source: github.com/BudEcosystem/Bud-Decision-Studio README The eleven models
- GitHub BudEcosystem/Bud-Decision-Studio had 99 stars on 2026-10-07. Releases and install scripts reference BudEcosystem/Bud-Decision-Engine.Source: GitHub API, 2026-10-07
Features and stack
Features
- Playground, Evaluate, History, Templates, Train, and API pages
- Studio template versions and act_threshold review routing
- Eleven curated models with per-model example scenarios
- LoRA Train page with export/import of fine-tunes
- curl and typesafe-sdk examples on the API page
Stack
- Tauri desktop shell
- Bundled Python engine with CUDA, Metal, or CPU paths
- Windows, macOS Apple Silicon, Linux x64 and ARM64 including DGX Spark
Pricing: Free desktop download; electricity and GPU hardware are yours. Each model keeps its upstream Hugging Face license linked in-app.
Links
FAQ
- Which repository has releases?
- README install badges and download table point to github.com/BudEcosystem/Bud-Decision-Engine/releases/latest. Bud-Decision-Studio holds source, docs, and media.
- Will typesafe-sdk work without code changes?
- README shows TypeSafeClient with api_key local and base_url http://127.0.0.1:8420 while the app is open.
- How is this different from ollaya?
- ollaya is a terminal daemon and pull workflow. Bud Decision Studio is a desktop GUI with Playground charts, template versioning, and bundled first-run engine setup.
Related learn guides
Original Jev guidance that pairs with this product pattern.
- System One model
The model family behind Jev: parallel typed questions, one forward pass, probabilities you can threshold.
- Where to run Jev
Compare official TypeSafe, Vercel AI Gateway, OpenRouter, Cloudflare Workers AI, and classifier.dev with a fact table.
Related products
Hand-picked neighbors with rich profiles or overlapping tags.