@nylorun/create-agent 0.4.0-beta → 0.5.0-beta

package/CHANGELOG.md

@@ -1,5 +1,15 @@
 # Changelog
 
+## 0.5.0-beta
+
+### Minor Changes
+
+- fa1860a: Use standard MODEL_PROVIDER, MODEL, MODEL_PROVIDER_API_KEY, and MODEL_PROVIDER_BASE_URL environment configuration. Export starter Hono apps and provide CLI development and production Node launchers. Existing starters require manual migration. Release preparation must update the creator Runtime compatibility pin with this release.
+
+### Patch Changes
+
+- Update the tested Harness, Runtime, and Studio compatibility combination.
+
 ## 0.4.0-beta
 
 ### Minor Changes

package/README.md

@@ -54,6 +54,6 @@ The renderer is shared by project creation, the isolated starter development run
 
 Repository development: [contributing](../CONTRIBUTING.md). Package publication: [releasing](../RELEASING.md).
 
-Generated projects use `nylorun dev`, with `--no-studio` and `--no-open` options, rather than a copied supervisor script. They contain one `tsconfig.json`; `npm run check` checks without emitting and `npm run build` emits to `dist/`. Repository examples add their own build configuration to exclude tests. Provider/model selection lives in `.env/model.json`, so configuration creates no top-level `config/` directory.
+Generated projects use `nylorun dev`, with `--no-studio` and `--no-open` options, rather than a copied supervisor script. They contain one `tsconfig.json`; `npm run check` checks without emitting and `npm run build` emits to `dist/`. Repository examples add their own build configuration to exclude tests. Provider/model selection and API keys use a conventional `.env` file (`MODEL_PROVIDER`, `MODEL`, `MODEL_PROVIDER_API_KEY`, and optional `MODEL_PROVIDER_BASE_URL`). The starter exports its Hono app; `nylorun start` supplies the production Node launcher. OAuth state is optional in `.nylorun/auth.json`. See [Runtime migration instructions](../runtime/README.md#upgrading-an-existing-starter).
 
 To verify the packed starter after building all packages, run `node create-agent/scripts/smoke-starter.mjs` from the repository root. It installs candidate tarballs into a disposable project and checks type checking, production output, agent assets, development CORS, streaming, watch reload, and shutdown. Add `--serve` to keep the fixture Studio running for browser verification; stop it with Ctrl-C. It uses a deterministic model adapter and makes no provider calls.

package/compatibility.json

@@ -1,5 +1,5 @@
 {
   "harness": "0.12.0-beta",
   "studio": "0.4.2-beta",
-  "runtime": "0.3.0-beta"
+  "runtime": "0.4.0-beta"
 }

package/dist/compatibility.json

@@ -1,5 +1,5 @@
 {
   "harness": "0.12.0-beta",
   "studio": "0.4.2-beta",
-  "runtime": "0.3.0-beta"
+  "runtime": "0.4.0-beta"
 }

package/dist/starter/.env.example

@@ -0,0 +1,6 @@
+# Copy to .env and fill in a provider and model, or run npm run configure.
+MODEL_PROVIDER=
+MODEL=
+MODEL_PROVIDER_API_KEY=
+# For MODEL_PROVIDER=custom only:
+# MODEL_PROVIDER_BASE_URL=https://your-provider.example/v1

package/dist/starter/README.md

@@ -3,6 +3,8 @@
 The creator configures a provider before starting development. For later sessions, run `npm run dev`. If setup was skipped with `--skip-config` or interrupted, run `npm run configure` first, then `npm run dev`.
 Edit agents under `agents/` and compose your Hono app in `src/index.ts`.
 
-`npm run dev` starts your app on port 3000, waits until its agent endpoint is ready, then starts Studio and opens it in your browser. Use `npm run dev -- --no-open` to start Studio without opening a browser. Run `npm run studio` in another terminal to attach Studio separately. `npm run build` creates `dist/`; deploy it alongside `package.json`, installed production dependencies, and private project configuration, then run `npm start`.
+`npm run dev` starts your app on port 3000, waits until its agent endpoint is ready, then starts Studio and opens it in your browser. Use `npm run dev -- --no-open` to start Studio without opening a browser. Run `npm run studio` in another terminal to attach Studio separately. `npm run build` creates `dist/`; deploy it alongside `package.json`, installed production dependencies, and environment variables, then run `npm start`.
 
-Development orchestration is provided by the installed `nylorun` CLI. Use `npm run dev -- --no-studio` for the application alone. Local provider/model selection is stored in `.env/model.json`. The single `tsconfig.json` builds your sources; `npm run check` checks types without emitting files.
+Development orchestration is provided by the installed `nylorun` CLI. Use `npm run dev -- --no-studio` for the application alone. Copy `.env.example` to `.env` and fill in `MODEL_PROVIDER`, `MODEL`, and `MODEL_PROVIDER_API_KEY`, or run `npm run configure`. Custom OpenAI-compatible providers also use `MODEL_PROVIDER_BASE_URL`. Hosting providers can supply these variables directly; no credential files are required for API keys. The single `tsconfig.json` builds your sources; `npm run check` checks types without emitting files.
+
+`src/index.ts` exports the Hono app. `nylorun dev` and `nylorun start` supply the Node server and load `.env` without overriding process variables. OAuth state, when used, lives in ignored `.nylorun/auth.json`. Serverless session behavior and Workers compatibility require separate validation.

package/dist/starter/_gitignore

@@ -1,9 +1,9 @@
 node_modules/
 dist/
 .data/
-.env/*
-!.env/.gitignore
-!.env/README.md
+.env
+.env.local
+.nylorun/
 
 # Legacy local settings remain private until migrated.
 config/model.json

package/dist/starter/package.json

@@ -11,13 +11,12 @@
     "dev": "nylorun dev",
     "studio": "nylorun studio --agent-url http://localhost:3000/agents",
     "build": "node -e \"import('node:fs/promises').then(async ({rm,cp}) => { await rm('dist', { recursive: true, force: true }); await cp('agents', 'dist/agents', { recursive: true, filter: (source) => !source.endsWith('.ts') }); })\" && tsc -p tsconfig.json",
-    "start": "node dist/src/index.js",
+    "start": "nylorun start",
     "check": "tsc --noEmit"
   },
   "dependencies": {
     "@nylorun/harness": "{{HARNESS_VERSION}}",
     "@nylorun/runtime": "{{RUNTIME_VERSION}}",
-    "@hono/node-server": "^2.1.1",
     "hono": "^4.13.7",
     "zod": "^4.1.12"
   },

package/dist/starter/src/index.ts

@@ -1,4 +1,3 @@
-import { serve } from "@hono/node-server";
 import { Hono } from "hono";
 import { Runtime, serveAgents } from "@nylorun/runtime";
 import { agents } from "../agents/index.js";
@@ -21,12 +20,4 @@ app.route(
   serveAgents({ agents, runtime })
 );
 
-serve(
-  {
-    fetch: app.fetch,
-    port: Number(process.env.PORT ?? "3000"),
-  },
-  (info) => {
-    console.log(`Server is running on http://localhost:${info.port}`);
-  },
-);
+export default app;

package/dist/sync.js

@@ -59,6 +59,10 @@ export async function synchronize(root, files, compatibility, options) {
             !managed(path) ||
             path.startsWith("test/") ||
             (path.startsWith("scripts/") && path !== "scripts/dev.mjs") ||
+            path === ".env" ||
+            path === ".env.local" ||
+            path === ".nylorun" ||
+            path.startsWith(".nylorun/") ||
             path.startsWith(".data/") ||
             path.startsWith("node_modules/") ||
             (path.startsWith(".env/") &&

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nylorun/create-agent",
-  "version": "0.4.0-beta",
+  "version": "0.5.0-beta",
   "description": "Create a local Nylorun Harness agent project.",
   "type": "module",
   "license": "Apache-2.0",

package/dist/starter/.env/README.md

@@ -1,4 +0,0 @@
-# Project credentials
-
-Run `npm run configure`. Provider credentials are stored in `auth.json`, and provider/model selection in `model.json`.
-Do not commit credentials. Optional integration environment variables belong in `integrations.env`.

package/dist/starter/.env/_gitignore

@@ -1,3 +0,0 @@
-*
-!.gitignore
-!README.md