Backboard.io Now Supports Jev by TypeSafe
Backboard.io adds support for TypeSafe Jev System One Models for structured decisions, featuring Python and TypeScript SDK integration.
Stock photo for illustration only, not from the actual event
- Backboard.io officially supports TypeSafe Jev for structured decisions
- Features Choice, Score, and Noul with Python and TypeScript SDKs
- Integrates via existing message APIs using llm_provider="typesafe"
Backboard.io has announced official support for TypeSafe Jev under its System One Models, tailored specifically for handling structured decision-making processes. Developers can now leverage this capability directly through their existing message API workflows.
This integration allows systems to evaluate both plain text and structured data while seamlessly incorporating conversation history into the evaluation process, ensuring more accurate and context-aware outputs.
The integration of TypeSafe Jev into Backboard.io marks a significant milestone for developers seeking deterministic, structured AI outputs such as probabilities and confidence scores, bypassing the need for complex parsing of raw text responses.
Key capabilities introduced with the Jev model support include:
- Structured decision evaluations including Choice, Score, and Noul true/false probabilities
- Typed answers, probability metrics, confidence levels where supported, and token usage tracking
- Python SDK v1.5.19 now available on PyPI alongside complete TypeScript SDK support
- Published Core Concepts and SDK guides featuring practical examples and full response documentation
Developers can begin integrating by specifying llm_provider="typesafe" and utilizing system_one.questions. The currently supported model is jev-latest, with full documentation available via the official Backboard.io developer resources.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment