Article
CommunityJev Integrations & Adapters
Jev with Agents in Microsoft Foundry
A Microsoft community walkthrough that wires Jev into a Microsoft Foundry agent as an OpenAPI tool, then uses it as a judgment layer that evaluates a generative model's answers against explicit criteria with confidence scores.
- Format
- Article
- Reading time
- 2 min
- Published by
- Community
- Author
- Robin Lester (Microsoft)
- Last verified
- Oct 2026
- Platforms
- API
- Capabilities
- Requires API key
Highlights
- Adds Jev to a Foundry agent as an OpenAPI tool pointed at https://api.typesafe.ai/v1/systemone, with jev-latest as the model.
- The connection key uses the form Bearer apikey_<the rest of your key>; the schema covers Noul, Choice, and Score questions.
- Pricing cited as of September 2026: jev-1.13.0 at $0.042 per million input tokens, free output tokens, and 250,000 tokens per second.
- Positions Jev as a separate judgment layer: a model generates, Jev evaluates against explicit criteria, and deterministic logic acts.
- The worked example asks billing, tone, and urgency questions about a hotel complaint and prints the agent's reply with Jev's output.
Watch out
A Microsoft community blog post by Robin Lester, not a product announcement, and preview access plus a TypeSafe API key are required. It notes Jev does not replace SHAP-style feature attribution, and the pricing and rate limits are dated September 2026 and marked as subject to change.
Questions about this entry
Short answers drawn from the linked project or page.
Treat Jev as a specialised judgment or evaluation component alongside your generative models: a Foundry model or agent produces an output, Jev evaluates it against predefined criteria, and deterministic logic decides whether the workflow may continue.
Create a new agent, add a Tool, and select OpenAPI tool; point it at https://api.typesafe.ai/v1/systemone, create a connection whose key has the form Bearer apikey_<the rest of your key>, and paste the supplied schema.
Three about a hotel complaint: a noul for whether the ticket concerns billing, a choice for the customer's tone with calm, frustrated, and angry options, and a score for urgency ordered from can wait to handle today.
It does not replace SHAP or feature attribution; instead it can act as a separate judgment layer around hosted MaaS models, evaluating inputs, outputs, or proposed actions against explicit criteria and returning structured decisions with confidence scores.
It says you could ask any LLM to classify or score with structured JSON output, while Jev is purpose-built for that workload, and it reports TypeSafe's claim of roughly 70-500 ms typical decision latency with pricing accurate only as of the article date.
The schema is titled TypeSafe Jev System One at version 1.1.0, uses jev-latest as the model enum with a default, requires state, model, and questions, and lists 400, 401, and 422 responses for invalid or unauthenticated requests.
Yes. You first apply for the Jev preview from TypeSafe and create an API key under API keys; the post notes you may also receive free credit, which should be more than sufficient for experimenting given the low per-token cost.
More Jev resources like this
On-call tooling where Jev answers four typed questions per alert - actionable, severity, owning team and duplicate-of - and plain code routes, pages or drops. Includes a root-cause benchmark where Jev and code investigate recorded incidents with no general-purpose LLM in the loop, enforcing budgets and verification.
Jev Integrations & Adapters#community#alerting#incident-response
Community
mingleiw/jev-oncallHTML
mingleiw/jev-oncallHTMLLiteLLM serves decision models on two routes that share one provider list: /v1/decisions for OpenAI-format bodies and /v1/systemone for System One bodies, with Jev available as the typesafe/jev-latest provider.
Jev Integrations & AdaptersDocs#community#litellm#gateway
Community
Akka's post walks from a Jev triage example - eight questions with probabilities - to an effects API judgment call and an AI Gateway that can route to hosted Jev or a self-hosted OpenJev model.
Jev Integrations & AdaptersArticle#community#java#akka
Community
Fly.io's TypeSafeAI connector gives Sprites access to Jev without the API key entering the machine: the Sprites gateway attaches the organization's encrypted key, and each Sprite sets only TYPESAFE_BASE_URL plus a placeholder key.
Jev Integrations & AdaptersArticle#community#fly-io#agents
Community
A Microsoft developer post compares two lunch-ordering agents, one gated by six Choice questions to Jev and one plain GPT-6-astra, to argue that typed decisions belong in front of a language model while the agent loop stays in code.
Jev Integrations & AdaptersArticle#community#microsoft#agents
Community
Pydantic's AI Gateway forwards Jev natively as a bring-your-own-key TypeSafe provider, so one gateway key covers spend views, limits, and input guardrails while TypeSafeModel turns each output field into a question.
Jev Integrations & AdaptersArticle#community#pydantic#gateway
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.