@agentrific/server 0.1.0 → 0.1.2

package/CHANGELOG.md

@@ -0,0 +1,13 @@
+# @agentrific/server
+
+## 0.1.2
+
+### Patch Changes
+
+- 72e68ad: Upgrade the embedded Pi SDK to 1.0.0 so Agentrific automatically recognizes existing credentials saved by both Pi v0 and v1, including the newer OpenAI ChatGPT login. Add real SDK integration tests for startup credential discovery, legacy credentials and transcripts, sessions, streaming, tools, model selection, attachments and delivery, with an isolated compatibility check for future Pi releases.
+
+## 0.1.1
+
+### Patch Changes
+
+- f730c2e: Publish the compiled server CLI to npm with automated versioning and release notes.

package/README.md

@@ -1,8 +1,22 @@
-# @agentrific/server
+# Agentrific Server
 
-Agentrific is a self-hosted control plane for long-running coding agents. This package provides its server and the `agentrific` command-line tool for macOS and Linux.
+Run coding agents on your computer or server, and control them from the Agentrific iOS and Android apps. Start work, follow progress, send instructions and review results from your phone while the agents run on your own machine.
 
-Install Node.js 22.19 or newer, Git, and [Pi](https://pi.dev/). Connect your model accounts through Pi using the same OS account that will run the server.
+This package installs the Agentrific server and the `agentrific` command-line tool for macOS and Linux. The mobile apps connect to your server; your repositories, model accounts and agent processes stay on that machine. A paired device can disconnect and reconnect without ending a running agent session.
+
+With Agentrific you can:
+
+- Manage tasks and agent sessions across your Git repositories.
+- Run coding tasks in separate Git worktrees.
+- Follow live progress and send follow-up instructions from your phone.
+- Attach images and files to give agents context.
+- Pair multiple devices with the same server and revoke access when needed.
+
+Learn more at [agentrific.com](https://agentrific.com).
+
+## Install and start
+
+You'll need Node.js 22.19 or newer, Git, and [Pi](https://pi.dev/) on the machine that will run your agents. Install Pi and connect your model accounts using the same OS account that will run the server. See the [model setup guide](https://agentrific.com/docs/pi-model-setup).
 
 ```sh
 npm install -g @agentrific/server
@@ -10,8 +24,26 @@ agentrific init
 agentrific start
 ```
 
-Run `agentrific --help` for commands. The server defaults to port 4318 and stores its configuration and state in `~/.agentrific`. Its initial projects folder is `~/projects`.
+The server defaults to port 4318 and stores its configuration and state in `~/.agentrific`. Its initial projects folder is `~/projects`; you can choose a different folder in the app's server settings.
+
+## Connect your phone
+
+Open the Agentrific app on iOS or Android. In the server terminal, choose Tailscale Serve, Cloudflare Tunnel, or an existing HTTPS endpoint, then press **p** to display a pairing QR code. Scan it in the app and enter the separately printed six-digit PIN.
+
+Once paired, select a project, create a task and start an agent. The server runs the work and streams progress back to your device. Use HTTPS for remote connections; your phone must be able to reach the configured server address.
+
+See the [server setup and pairing guide](https://agentrific.com/docs/secure-server-setup) for connection options and configuration.
+
+## Useful commands
+
+```sh
+agentrific --help          # Show commands and options
+agentrific status         # Show server status
+agentrific pair           # Print a fresh pairing QR code and PIN
+agentrific devices        # List paired devices
+agentrific update --check # Check for a newer server release
+```
 
-Use HTTPS through Tailscale Serve, Cloudflare Tunnel, or an existing HTTPS endpoint when connecting remotely. In the server terminal, choose a connection and press **p** to display a pairing QR code. Pair your Agentrific client with the QR code and the separately printed six-digit PIN.
+## Access and permissions
 
 Devices must pair before accessing server commands and events. Coding agents execute with the server user's permissions, so run the server under an account with access only to the repositories and credentials you intend to use.

package/cli.js

@@ -30,8 +30,20 @@ var init_package = __esm({
   "package.json"() {
     package_default = {
       name: "@agentrific/server",
-      version: "0.1.0",
+      version: "0.1.2",
       description: "Self-hosted control plane for long-running coding agents",
+      keywords: [
+        "agentrific",
+        "ai",
+        "android",
+        "cli",
+        "coding-agents",
+        "ios",
+        "pi",
+        "remote-control",
+        "self-hosted"
+      ],
+      homepage: "https://agentrific.com",
       repository: {
         type: "git",
         url: "git+https://github.com/devular/agentrific.git",
@@ -51,7 +63,7 @@ var init_package = __esm({
         cli: "AGENTRIFIC_DATA_DIR=./data tsx src/cli.ts"
       },
       dependencies: {
-        "@earendil-works/pi-coding-agent": "^0.87.1",
+        "@earendil-works/pi-coding-agent": "1.0.0",
         "@fastify/cors": "^11.3.0",
         "@fastify/multipart": "^10.1.2",
         "@inquirer/select": "^5.2.5",
@@ -2328,10 +2340,10 @@ ${logs}` : ""}`);
 // src/delivery-writer.ts
 import { createAgentSession, SessionManager } from "@earendil-works/pi-coding-agent";
 import { z as z11 } from "zod";
-function deliveryWriter(models2) {
+function deliveryWriter(models2, createSession = createAgentSession) {
   return async (run, input, diff, signal) => {
     signal?.throwIfAborted();
-    const { session } = await createAgentSession({
+    const { session } = await createSession({
       cwd: run.workspacePath,
       model: await models2.resolve(run.model ?? await models2.defaultReference()),
       modelRuntime: models2.runtime,
@@ -2535,11 +2547,11 @@ function toolOutput(value) {
   const text = parsed.success ? parsed.data.content?.filter((part) => part.type === "text").map((part) => part.text || "").join("\n") || "Completed without text output." : boundedJson(value, 4e3);
   return text.length > 4e3 ? `${text.slice(0, 4e3)}\u2026 [truncated]` : text;
 }
-function createPiFactory(files, models2, delivery2) {
+function createPiFactory(files, models2, delivery2, createSession = createAgentSession2) {
   return {
     async open(sessionId, cwd, sessionDirectory, selected = legacyModel, run) {
       const model = await models2.resolve(selected);
-      const { session } = await createAgentSession2({
+      const { session } = await createSession({
         cwd,
         model,
         modelRuntime: models2.runtime,