🔧 An alternative to the 'eval' function in JavaScript that is faster, easier/better to use, and has less security issues.
Efferent composable agent framework, 0.22.0. Node 24.
Portable pieces of Mossbear's evaluation path — command redaction, the LLM adapter seam, and guide normalization.
Scores the answers of an eval against their labels, with one import path for each shape of answer: a core of ratios and totals, ./structured for structured values, ./classification for labels as scikit-learn scores them, and ./promptfoo.
Eval a JS string as if it was a file being required
Cabidela is a small, fast, eval-less, Cloudflare Workers compatible, dynamic JSON Schema validator
Create an in-memory webpack module by evaluating a function.
Agent-native eval tool — eval agents, services, functions, and coding-agent fixtures
Agent evals as data: scenarios a product writes, run against its real agent configuration (persona, tools, tool policy, turn policy) on every channel it serves, from recorded model responses in CI or against the live model, with a CLI that fails the build
A custom version of the handlebars package which, to improve security, does not use `eval` or `new Function`. This means that templates can't be compiled into JavaScript functions in advance and hence, rendering the templates is a lot slower.
json serializable rule engine / boolean expression evaluator
Skills Source for Composable View Memory: DSH skill management, evolution suggestions, evidence trajectories and eval sets.
Evaluate and query json object value at given path
Find your AI agent's real failures and turn each one into a test. Runs your agent (chat, function or multi-step pipeline, TypeScript or Python) on realistic cases or replayed production inputs, reviews every run, and groups what went wrong.
Colocated agent evaluations with independent Vitest checks
A hack to enable use of libraries that depend on a basic form of eval in Atom
Efferent composable agent framework, 0.22.0. Node 24.
JavaScript client for running remote evaluations with Scoundrel.
Reactive LLM eval, tracing, and regression layer as a Convex component.
Evaluation definitions and runner for ADK-based Botpress agents
Extremely fast, eval-free runtime JSON Schema validation. Interprets a schema you do not know ahead of time — no new Function, no build step, runs anywhere.
Protected eval plugin for form.io
Eval dataset loaders (JSONL, CSV, YAML, HTTP) with a lazy iterator API
Run Introspection Recipes from vitest-evals
Evaluates a polynomial.
A tiny, safe, fast JavaScript library for decimal arithmetic expressions.
eval-gpt
Eval Engine SDK for JavaScript/TypeScript
Allows exposing a module via `eval`. Useful if you have a module that needs to attach a bunch of things on to the global scope using implied globals (like function definitions).
Provider- and storage-neutral cassette contracts and deterministic extraction for agent-qa.
StorageImplementation shell for local agent-qa cassette IO behind eval ports.
Replacement for eval that reveals the location of syntax errors
Evaluate streams of JavaScript code in a given context
Mathematical expression evaluator
Framework-neutral helpers for running and reporting LLM evaluations
EvalSutra AI: evidence-first, keyless, offline agent evaluation CLI plus 18 portable agent skills for Claude Code, Cursor, Codex, Gemini CLI and Copilot.
Test your LLM-powered healthcare and life sciences apps with a TypeScript-native, Vitest-based eval runner. No API key required.
Benchmark Claude Code plugins by A/B comparing plugin versions with LLM-judged evaluation prompts.
Evaluation of DMN 1.1 decision tables, limited to S-FEEL (Simple Friendly Enough Expression Language)
Node.js and browser 8 lines implementation of a safe eval
LavaMoat Webpack plugin for running dependencies in Compartments without eval
Jetty × eve — an eve extension that live-grades an agent's turns with Jetty, steers a reply-style bandit from the grades, and reports eve evals into Jetty
Small library to dynamically create and evaluate expression with multiple parameters (even undefined)