Prebuiltcards

apps/rowboat/app/api/templates/route.ts

@@ -2,5 +2,6 @@ import { NextResponse } from 'next/server';
 import { templates } from '@/app/lib/project_templates';
 
 export async function GET() {
+  // The templates are now dynamically loaded from JSON files in the templates folder
   return NextResponse.json(templates);
 }

apps/rowboat/app/lib/prebuilt-cards/README.md

@@ -0,0 +1,43 @@
+# Prebuilt Cards Directory
+
+This directory contains JSON files that define prebuilt assistant templates. These templates appear as cards in the "Pre-built Assistants" section of the application.
+
+## How to Add New Prebuilt Cards
+
+1. Create a new JSON file in this directory (e.g., `my-assistant.json`)
+2. The filename (without extension) will be used as the template key
+3. The JSON file should follow the WorkflowTemplate schema structure
+
+## Required Structure
+
+Each prebuilt card JSON file must have:
+- `name`: Display name for the template
+- `description`: Brief description of what the template does
+- `agents`: Array of agent configurations
+- `startAgent`: Name of the starting agent
+- `tools`: Array of tool configurations (optional)
+- `prompts`: Array of prompt configurations (optional)
+- `pipelines`: Array of pipeline configurations (optional)
+
+## Example Prebuilt Cards
+
+See the existing files in this directory:
+- `github-data-to-spreadsheet.json` - Fetches GitHub stats and logs to Google Sheets
+- `meeting-prep.json` - Research meeting attendees and send to Slack
+- `interview-scheduler.json` - Automate interview scheduling with Google Sheets/Calendar
+
+## Template Loading
+
+Prebuilt cards are automatically loaded when the application starts. Simply drop a new JSON file here and restart the application to see it appear in the prebuilt assistants section.
+
+## Location
+
+This directory is located at `app/lib/prebuilt-cards/` to keep the template definitions close to the `project_templates.ts` file that loads them.
+
+## Validation
+
+The system validates that each template has:
+- A valid `agents` array
+- Proper JSON syntax
+
+Invalid templates will be logged as warnings but won't break the application.

apps/rowboat/app/projects/components/build-assistant-section.tsx

@@ -17,7 +17,7 @@ import { Project } from "@/src/entities/models/project";
 import { z } from "zod";
 import Link from 'next/link';
 
-const SHOW_PREBUILT_CARDS = process.env.NEXT_PUBLIC_SHOW_PREBUILT_CARDS === 'true';
+const SHOW_PREBUILT_CARDS = process.env.NEXT_PUBLIC_SHOW_PREBUILT_CARDS !== 'false';
 
 
 

apps/rowboat/src/application/lib/copilot/current_workflow.ts

@@ -7,7 +7,7 @@ The design of the multi-agent system is represented by the following JSON schema
 {workflow_schema}
 \`\`\`
 
-If the workflow has an 'Example Agent' as the main agent, it means the user is yet to create the main agent. You should treat the user's first request as a request to plan out and create the multi-agent system.
+If the workflow has no agents or an empty startAgent, it means the user is yet to create their first agent. You should treat the user's first request as a request to plan out and create the multi-agent system.
 
 ---
 `;

apps/rowboat/src/application/lib/copilot/example_multi_agent_1.ts

@@ -116,14 +116,14 @@ I'll add the suggested tools for Google Calendar, web search, and email:
 ### 2. Create Agents
 
 #### a. Hub Agent (Meeting Assistant Hub)
-I'll edit the Example Agent to become the hub agent:
+I'll create the hub agent:
 
 \`\`\`copilot_change
 // action: edit
 // config_type: agent
-// name: Example Agent
+// name: Meeting Assistant Hub
 {
-  "change_description": "Transformed Example Agent into the main hub agent orchestrating the meeting summary workflow.",
+  "change_description": "Created the main hub agent orchestrating the meeting summary workflow.",
   "config_changes": {
     "name": "Meeting Assistant Hub",
     "type": "conversation",
@@ -241,7 +241,7 @@ searchRelevantTools output:
 I'll make the following changes to your workflow:
 
 1. Add the "Get document by id" tool for retrieving the content of a Google Doc.
-2. Replace the Example Agent with a single user-facing agent that:
+2. Create a single user-facing agent that:
    * Asks the user for the Google Doc ID and their question.
    * Uses the tool to fetch the document content.
    * Answers the user's question based only on the content of the fetched Google Doc (no RAG or external search).
@@ -277,14 +277,14 @@ I'm adding the "Get document by id" tool to fetch the content of a Google Doc by
 }
 \`\`\`
 
-I'm replacing the Example Agent with a user-facing agent that fetches a Google Doc by ID and answers questions based on its content:
+I'm creating a user-facing agent that fetches a Google Doc by ID and answers questions based on its content:
 
 \`\`\`copilot_change
 // action: edit
 // config_type: agent
-// name: Example Agent
+// name: Meeting Assistant Hub
 {
-  "change_description": "Replaced Example Agent with a user-facing agent that fetches a Google Doc by ID and answers user questions based only on its content.",
+  "change_description": "Created a user-facing agent that fetches a Google Doc by ID and answers user questions based only on its content.",
   "config_changes": {
     "name": "Google Doc QnA Assistant",
     "type": "conversation",