AgentPM™

Harness Provider Lab

@zack/harness-provider-lab

AgentPM Harness lab for process-hosted model, embedding, KnowledgeRuntime, and MemoryRuntime providers.

Use case: harness-provider-labagentpm-runnodeworkspace
Bootstrap
agentpm new @zack/harness-provider-lab@0.1.0 my-project
Next steps
cp .env.local.example .env.local
Weekly starts
1
0%
Last publish
1d ago
v0.1.0
OverviewReadmeSecurity

harness-provider-lab

harness-provider-lab is a publishable AgentPM workflow template for users who want to see every main Harness provider bridge running from local process services.

Generated apps use a local root agent.json that depends on published packages:

  • @zack/react-loop@0.1.0
  • @zack/agentpm-docs@0.1.1
  • @zack/devwork-maintainer-guide@0.1.0
  • @zack/support-customer-state@0.1.0

The generated agentpm.harness.json wires process-hosted ModelRuntime, EmbeddingProvider, external KnowledgeRuntime, external MemoryRuntime, and approval controller services.

What It Demonstrates

  • a deterministic process ModelRuntime that proposes canonical Harness actions
  • a process EmbeddingProvider for the local @zack/agentpm-docs vector Knowledge package
  • a process KnowledgeRuntime for a published vector Knowledge package
  • a process MemoryRuntime for a published Memory Blueprint
  • environment-based credential projection with no literal credentials in Harness config or package metadata
  • the difference between the quick-start embedding template and a serious provider-integration lab

Generate Locally

agentpm new ./template-packages/harness-provider-lab/agent.json app-harness-provider-lab
cd app-harness-provider-lab
cp .env.local.example .env.local
pnpm install
pnpm setup:bindings
agentpm harness --json

Set OPENAI_API_KEY in .env.local before the full run. Preflight uses the same environment projection as a run, so it will report the embedding provider unavailable until the key is present.

Publish Checks

agentpm lint
agentpm publish --dry-run

The source for this Template package can be found here.