- Category
- Developer Tools
- Rank
- No. 2031Tools index
Previous survey · No. 1965 ·
- Type
- TOOL
- Use case
- Models: Train & Run
- Interfaces
- CLI
- GitHub
- 3 stars
- Latest release
- v0.16.0
- Date
About
Onesie is a Unix-style command-line tool for calling System One models (small, fast classification-style language models such as TypeSafe's Jev) from shell pipelines, scripts, and agentic workflows. It adds caching, batching, deduplication, and streaming, plus a calibration command that tunes thresholds and question wording against labeled data to reduce false positives and negatives.
At a glance
Onesie turns a question about text into a model answer that a shell script can use. It is worth evaluating when a simple text rule cannot capture the distinction you need.
Good fit
Developers prototyping a classification or triage step with a human fallback.
Consider something else when
Use explicit policy for authorization and destructive operations, where a model’s mistaken allow could cause harm.
Start here
Use a labeled set of harmless text inputs and compare clear, ambiguous and wrong responses. Verify that an uncertain result pauses your workflow.
Cost and requirements
Go CLI with several documented hosted decision providers. Access and charges depend on the configured provider.
Limits to know
The provider receives the input state. Scores and thresholds are specific to the model and examples, so check errors before using an exit code to drive an action.
Sources checked 2026-11-05. Source-based assessment.
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.