JSON Contracts FAQs

Question 1

What kind of information can be passed to JSON Contracts for model interpretation?

Accepted Answer

You can pass app or system variables as 'context' to JSON Contracts. This context is then used by the contracts (rules and examples) to guide the model's interpretation, ensuring relevant information is available without cluttering user prompts.

Question 2

How does JSON Contracts help in CI/CD pipelines?

Accepted Answer

JSON Contracts includes a powerful CLI tool that enables you to validate and lint your JSON contracts directly within your CI/CD workflows. This ensures contract quality and consistency before they are used by agents in production.

Question 3

Does JSON Contracts generate JSON or call LLMs directly?

Accepted Answer

No, JSON Contracts acts purely as a schema contract registry and validator, not a generator. Your agent is responsible for generating JSON based on the provided contracts, and JSON Contracts ensures the generated output adheres to the defined schema and rules.

Question 4

Can JSON Contracts provide guidance when an agent's JSON output is invalid?

Accepted Answer

Yes. If an agent produces invalid JSON, JSON Contracts can return 'repair contracts' that provide specific instructions and guidance, allowing the agent to correct its output and achieve schema validation.

Question 5

What is JSON Contracts?

Accepted Answer

JSON Contracts is a local server and CLI tool that provides Git-controlled JSON schemas, rules, and examples to AI agents. It validates their natural-language-to-JSON conversions and offers repair guidance to ensure reliable, schema-valid JSON output.

JSON Contracts

JSON Contracts

主要功能

使用案例

主要功能

使用案例