@llm-workbench/runtime
v0.3.2
Published
Headless, model-agnostic runtime for LLM workflows — records state, artifacts, rules, human-review gates, traces, cost telemetry, and tamper-evident run bundles.
Maintainers
roymcfarlandReadme
@llm-workbench/runtime
MIT-licensed — the headless, model-agnostic core of LLM Workbench. Records workflow state, artifacts, rules, human-review gates, traces, and cost telemetry, and exports tamper-evident run bundles. No React or framework dependency — runs in the browser, Node, or edge-style runtimes.
npm install @llm-workbench/runtimeAPI surface
| Export | Role |
| --- | --- |
| WorkbenchRuntime / WorkbenchSession | start runs; drive steps and gates; write & patch artifacts; log model I/O and tool calls |
| SchemaRegistry | register JSON Schemas and Ajv-validate artifacts before they become run state |
| MemoryRunRepository, IndexedDB, HTTP adapters | pluggable persistence behind one RunRepository interface |
| parseRunBundleJson / verifyRunBundleIntegrity | import/verify SHA-256-signed, canonical-JSON run bundles |
| summarizeModelTelemetry | typed cost/usage ledger grouped by provider, model, step, user, tenant, plan |
| WorkbenchError | structured errors with stable codes across package boundaries |
Quick start
A complete, runnable example lives in the repository root README.md. It imports the package and exercises gates, artifacts, model-I/O telemetry, and a signed bundle export under plain Node.
Docs
Overview, architecture, scope, and non-goals: repository root README.md and PROJECT.md.