output-validation

Pluggable controls for reliable AI agents: validated outputs, bounded loops and frozen collective decisions.

Turn any LLM into a reliable local JSON API | Contract-first local LLM API builder for structured JSON outputs, model/prompt comparison, schema validation, repair/retry, model switching, and runtime testing.

🛡️ Programmable Guardrails for LLM Applications in Java. A framework-agnostic toolkit for input/output validation, PII masking, and jailbreak detection. The Java alternative to NVIDIA NeMo Guardrails.

Strict JSON schema validator with streaming partial JSON repair, missing field filling, and sanitization

Strict JSON schema validator with streaming partial JSON repair, missing field filling, and sanitization

Calibrated output validation with the Bayes Factor: weigh per-dimension quality scores into a pass, fail, or escalate decision on the Jeffreys scale, not an arbitrary threshold. Zero-dependency, node-free TypeScript.

The new package transforms a user’s plain‑text search phrase into a structured, BM25‑ranked list of PostgreSQL search results. It takes the query text as input, forwards it to an LLM that rewrites the

A practical repository showcasing guardrails for AI and LLM agents, including input/output validation, jailbreak protection, prompt security, and agent safety techniques to build secure and reliable AI systems.

A new package that processes user-provided text input to generate structured summaries of technical challenges, such as the difficulty of implementing resumable LLM streaming. It uses a system prompt

Parallel multi-critic LLM output arbitration using LangGraph. Multiple specialized critics evaluate responses in parallel; central adjudicator synthesizes verdicts with confidence scores. Pydantic contracts ensure type-safe judge outputs.

A unified suite of MCP suppressors that prevent hallucinations, enforce grounding, and stabilise agent behaviour.

Deterministic regex scanner for dangerous payloads in LLM responses — Python port of the npm package, byte-for-byte parity. OWASP LLM02.

A lightweight, engineering-first LLM framework for FastAPI featuring session-based context caching and reliable output validation.

A practical Python toolkit for evaluating, monitoring, and governing LLM outputs in production

AI agent for quality validation of batches of AI agent outputs against scoring criteria, built with Agentic Star.

Listener-side output re-validation decorator against schema/predicate.

Design-by-contract: verify your function inputs and outputs. Includes a large number of generated verifcation functions for convenience.

Verify LLM output against your source documents. Catch hallucinations in RAG pipelines and agentic workflows before they reach users.

A linter for content written by AI agents. Most linters check what the text says. This one checks how it got there whether the agent that wrote a note labelled itself, whether a claim that generalizes was routed to review instead of asserted as fact, and whether the review queue is actually moving.

Enhancing a Customer Support Agent with production-grade safety layers including prompt injection defense, PII masking, structured output validation, and red-team testing.