Greenroom

One reactive state tree for your run-of-show, financial ledger, and autonomous PayPal treasury because running a 2,000-person conference shouldn't mean reconciling 14 spreadsheets at 2 AM.

Comment

Greenroom — Backstage Financial & Schedule Command Center for Live Events

Built for the "Build What's Next with PayPal and AI" Hackathon
Sponsor Integrations: AG Grid Enterprise (v36.3.0) | Bryntum Scheduler Pro (v7.3.8) | APIMatic (@paypal/paypal-server-sdk v2.5.0) | Render

License: MIT PayPal Sandbox AG Grid Enterprise Bryntum Scheduler Pro Deployed on Render

Table of Contents

  1. Executive Summary & Problem Statement
  2. System Architecture & Data Flow
  3. Unified Domain Model & Cost Mathematics
  4. Sponsor Engineering Deep-Dives
  5. Development with Sponsor MCP Servers
  6. Backend API Reference
  7. Installation, Configuration & Local Execution
  8. End-to-End Verification Scenarios
  9. Repository Structure
  10. License

1. Executive Summary & Problem Statement

Organizing a multi-track technical conference or hackathon (modeled in this project as DevConnect 2026, a 2-day event on November 14–15, 2026) requires synchronizing three tightly coupled operational domains that traditionally live in isolated tools:

  1. Temporal & Resource Scheduling (Run-of-Show): Allocating physical rooms (Main Stage, Workshop Room A, Workshop Room B), technical crews (AV Team, CloudLens Photography, Bay Area Catering, MerchOps), and speakers across a constrained timeline with strict finish-to-start dependencies.
  2. Financial Accounting & Budget Tracking (The Ledger): Tracking hourly vendor labor rates, fixed catering contracts, speaker honoraria, hackathon prize pools, and inbound sponsorship receivables across budget categories.
  3. Payment Execution (Receivables & Payables): Issuing itemized B2B invoices to corporate sponsors, tracking overdue receivables, and disbursing outbound payouts to speakers and contractors once deliverables are completed.

The Operational Failure Mode:
When a live event deviates from plan—for example, a morning AV call or photography shift is extended by two hours, or a keynote is delayed—organizers manually update a Gantt chart, copy-paste hour deltas into a spreadsheet, recompute category subtotals, and separately log into a payment portal to invoice sponsors or pay vendors. Because the schedule, the ledger, and the payment rails do not share a single source of truth, budget overruns and missed sponsor collections go undetected until post-event reconciliation.

The Greenroom Solution:
Greenroom unifies Bryntum Scheduler Pro, AG Grid Enterprise, and PayPal's Sandbox APIs around a single reactive WorkItem[] state tree, paired with an AI operations agent and a natural-language grid state engine:

  • Resizing or dragging an hourly task in Bryntum Scheduler Pro immediately recalculates item cost, updates group subtotals in AG Grid Enterprise, and recomputes event-wide financial KPIs (Total Budget, Sponsor Income, Paid Out, Invoices Paid, Overdue, and Net Position).
  • Typing natural-language view commands into the AG Grid AI bar dynamically reconfigures grid grouping, multi-column sorting, filtering, and column visibility.
  • Conversing with the Greenroom AI operations assistant dispatches real B2B sponsor invoices via PayPal Invoicing v2 and queues outbound speaker/vendor disbursements via PayPal Payouts v1 behind a mandatory human-in-the-loop approval gate.

2. System Architecture & Data Flow

2.1 High-Level Component & Integration Topology

+---------------------------------------------------------------------------------------------------+
|                        REACT 18 FRONTEND (Shared WorkItem[] State Tree)                           |
|                                                                                                   |
|  +-------------------------------------------+     +-------------------------------------------+  |
|  | Bryntum Scheduler Pro v7.3                |     | AG Grid Enterprise v36.3                  |  |
|  | - 11 Stages, Crews & Sponsors             |<--->| - Grouped Financial Ledger (sum hours/$)  |  |
|  | - 17 Finish-to-Start Dependency Chains    |     | - AG Grid AI Natural-Language Bar         |  |
|  | - onEventResizeEnd / onEventDrop hooks    |     | - getStructuredSchema() / setState()      |  |
|  +---------------------+---------------------+     +---------------------+---------------------+  |
|                        |                                                 |                        |
|                        +------------------------+------------------------+                        |
|                                                 |                                                 |
|                                                 v                                                 |
|  +---------------------------------------------------------------------------------------------+  |
|  | Live KPI Header Strip (Total Budget | Sponsor Income | Paid Out | Invoices Paid | Net)      |  |
|  +---------------------------------------------------------------------------------------------+  |
|                                                                                                   |
|  +-------------------------------------------+     +-------------------------------------------+  |
|  | Greenroom AI Operations Sidebar           |     | Human-in-the-Loop ApprovalToast           |  |
|  | - Natural-Language Prompt Bar             |---->| - Gated Approval for Outbound Payouts     |  |
|  | - Live Tool Execution Agent Trace         |     | - Explicit Approve / Reject Controls      |  |
|  +---------------------+---------------------+     +---------------------+---------------------+  |
+------------------------|-------------------------------------------------|------------------------+
                         | POST /api/chat                                  | POST /api/approve
                         | POST /api/grid-ai                               |
                         v                                                 v
+---------------------------------------------------------------------------------------------------+
|                              EXPRESS + TYPESCRIPT BACKEND (Port 3001)                             |
|                                                                                                   |
|  +---------------------------+   +---------------------------+   +-----------------------------+  |
|  | server/agent-handler.ts   |   | server/grid-ai-handler.ts |   | server/paypal-proxy.ts      |  |
|  | - Schedule & Budget AI    |   | - NL to AG Grid State     |   | - APIMatic SDK OAuth2 Auth  |  |
|  | - Autonomous Invoicing    |   | - Filter / Sort / Group   |   | - PayPal Invoicing v2 API   |  |
|  | - Payout Approval Gating  |   |   JSON State Compiler     |   | - PayPal Payouts v1 API     |  |
|  +-------------+-------------+   +-------------+-------------+   +--------------+--------------+  |
+----------------|-------------------------------|--------------------------------|-----------------+
                 |                               |                                |
                 +---------------+---------------+                                |
                                 |                                                |
                                 v                                                v
+-------------------------------------------------+   +---------------------------------------------+
| MULTI-PROVIDER LLM INFERENCE (server/llm.ts)    |   | PAYPAL DEVELOPER SANDBOX PLATFORM           |
| - Groq Cloud / xAI / OpenAI / Google Gemini     |   | - @paypal/paypal-server-sdk v2.5 (APIMatic) |
| - Low-latency structured directive generation   |   | - POST /v1/oauth2/token (Basic Auth)        |
|                                                 |   | - POST /v2/invoicing/invoices (+ /send)     |
|                                                 |   | - POST /v1/payments/payouts (Batch Payout)  |
+-------------------------------------------------+   +---------------------------------------------+

2.2 End-to-End Execution Flows (Invoicing & Approval-Gated Payouts)

FLOW A: AUTONOMOUS SPONSOR INVOICING (INBOUND RECEIVABLE)
=========================================================
[Organizer]
   |  1. Enters prompt: "Send Neural Labs their sponsor invoice"
   v
[React UI (App.tsx)] ---> POST /api/chat { message, workItems, resources }
   |
   v
[Express (agent-handler.ts)] ---> Calls LLM (server/llm.ts)
   |  2. LLM identifies Bronze Sponsorship Package ($1,000, [email protected])
   |     and emits INVOICE_JSON directive
   v
[PayPal Proxy (paypal-proxy.ts)]
   |  3. Calls APIMatic SDK: OAuthAuthorizationController.requestToken(Basic Auth)
   |  4. POST /v2/invoicing/invoices (Creates draft invoice in PayPal Sandbox)
   |  5. POST /v2/invoicing/invoices/{invoiceId}/send (Dispatches invoice to sponsor)
   v
[React UI & AG Grid Enterprise]
   |  6. Logs `tool: create_and_send_invoice` in Agent Trace
   |  7. Flips AG Grid Payment status badge from "draft" -> "sent" and increments Overdue KPI


FLOW B: HUMAN-IN-THE-LOOP SPEAKER PAYOUT (OUTBOUND PAYABLE)
===========================================================
[Organizer]
   |  1. Enters prompt: "Pay Sarah Chen for the keynote"
   v
[Express (agent-handler.ts)] ---> Calls LLM (server/llm.ts)
   |  2. LLM identifies Keynote: AI Frontiers ($350, [email protected])
   |     and emits APPROVAL_JSON:{ action: "payout", amount: 350, recipient: "sarah.chen@..." }
   v
[ApprovalToast Component]
   |  3. Renders modal: "Agent Approval Required — Pay Sarah Chen for Keynote ($350)"
   |  4. Organizer clicks [Approve]
   v
[Express (/api/approve)] ---> Calls sendPayout() in server/paypal-proxy.ts
   |  5. Obtains OAuth2 Bearer Token via APIMatic SDK OAuthAuthorizationController
   |  6. POST /v1/payments/payouts (Dispatches batch payout in PayPal Sandbox)
   v
[React UI, AG Grid & Bryntum Scheduler Pro]
   |  7. Logs `tool: paypal_payouts_create` with Batch ID in Agent Trace
   |  8. Flips AG Grid Payment badge to "completed" (green) & increments Paid Out KPI ($0 -> $350)
   |  9. Transitions Keynote event bar in Bryntum Scheduler Pro to emerald green (#22c55e)

3. Unified Domain Model & Cost Mathematics

Both Bryntum Scheduler Pro and AG Grid Enterprise consume the same canonical WorkItem array defined in src/lib/types.ts. Neither component maintains a disconnected copy of cost or schedule state.

3.1 Core TypeScript Interfaces (src/lib/types.ts)

export interface WorkItem {
  id: string;
  title: string;
  type: 'session' | 'vendor' | 'sponsor' | 'prize' | 'staff';
  resourceId: string;
  startDate: string;          // ISO-8601 datetime (e.g., "2026-11-14T08:30:00")
  endDate: string;            // ISO-8601 datetime
  duration: number;           // Duration in decimal hours
  rateType: 'hourly' | 'fixed';
  rate: number;               // Hourly rate ($/hr) or fixed contract amount ($)
  totalCost: number;          // Dynamically computed line-item total ($)
  budgetCategory: string;     // "AV & Production" | "Speakers" | "Catering" | "Swag & Merch" | "Prizes" | "Sponsorship" | "Operations"
  payDirection: 'inbound' | 'outbound' | 'none';
  paypalType: 'invoice' | 'order' | 'payout' | null;
  paypalId: string | null;
  paypalStatus: 'draft' | 'sent' | 'paid' | 'pending' | 'completed' | null;
  payeeEmail: string | null;  // Outbound payout recipient (PayPal sandbox personal account)
  payerEmail: string | null;  // Inbound invoice recipient (PayPal sandbox business account)
  status: 'scheduled' | 'in_progress' | 'completed' | 'cancelled';
  eventColor?: string;
}

3.2 Real-Time Cost & KPI Derivation (src/lib/budget.ts)

When an event is resized or dragged in Bryntum Scheduler Pro, the duration delta is computed from the start and end timestamps:

duration (hours) = (endDate.getTime() - startDate.getTime()) / 3,600,000

For items where rateType === 'hourly', recalculateCost() updates totalCost:

totalCost = Math.round(rate * duration * 100) / 100

From the updated WorkItem[] state, computeKPIs() derives the six live header metrics:

  • Total Budget (totalOutbound): Sum of totalCost for all items where payDirection === 'outbound'
  • Sponsor Income (totalInbound): Sum of totalCost for all items where payDirection === 'inbound'
  • Paid Out (paidOut): Sum of totalCost for all items where payDirection === 'outbound' and paypalStatus === 'completed'
  • Invoices Paid (invoicesPaid): Sum of totalCost for all items where payDirection === 'inbound' and paypalStatus === 'paid'
  • Overdue Count (overdue): Count of items where payDirection === 'inbound' and paypalStatus === 'sent'
  • Net Position (net): totalInbound - totalOutbound

4. Sponsor Engineering Deep-Dives

4.1 AG Grid Enterprise v36.3 & AG Grid AI Toolkit Engine

Implementation Files:

Key Technical Implementations:

  1. v36 Module Registration & Programmatic Dark Theming:
    Registers AllCommunityModule and AllEnterpriseModule via ModuleRegistry.registerModules() and composes the v36 Quartz dark-blue theme (themeQuartz.withPart(colorSchemeDarkBlue)) to match the surrounding backstage UI without legacy CSS file bloat.
  2. Enterprise Row Grouping & Live Aggregations:
    Columns are grouped by budgetCategory by default (rowGroup: true, groupDefaultExpanded: 1), with enableRowGroup: true enabled across all columns via defaultColDef so the AI engine can dynamically regroup the ledger by type, payDirection, paypalStatus, or status. The duration (Hours) and totalCost (Cost) columns define aggFunc: 'sum', computing live category subtotals whenever Bryntum updates an item's duration.
  3. Custom React JSX Cell Renderers:
    Renders color-coded category badges for type (session, vendor, sponsor, prize, staff), directional flow indicators for payDirection (↓ In, ↑ Out), and translucent status pills for paypalStatus (paid, completed, sent, pending, draft), plus conditional cell styling highlighting line items where totalCost > $500.
  4. Natural-Language AG Grid AI Toolkit Workflow (/api/grid-ai):
    When the organizer enters a query in the AG Grid AI bar (for example, "group by type, sort by cost descending, show only outbound"), LedgerGrid.tsx captures the grid's structured schema (api.getStructuredSchema()) and current state (api.getState()) and posts them to /api/grid-ai. The backend translates the intent into a structured AG Grid state object containing filter.filterModel, sort.sortModel, rowGroup.groupColIds, and columnVisibility.hiddenColIds, which the frontend applies via api.setFilterModel(), api.applyColumnState(), api.setRowGroupColumns(), and api.setColumnsVisible(). A dedicated Reset button restores the default grouping and clears active filters in one click.

4.2 Bryntum Scheduler Pro v7.3 Dependency & Resource Engine

Implementation Files:

Key Technical Implementations:

  1. Public npm Trial Package Aliasing:
    Configured in package.json using npm package aliasing ("@bryntum/schedulerpro": "npm:@bryntum/[email protected]") paired with @bryntum/[email protected] and the @bryntum/schedulerpro/stockholm-dark.css theme, enabling zero-credential npm install in local and CI/CD environments.
  2. Multi-Domain Resource & Dependency Graph Modeling:
    Models 11 distinct resources across physical rooms (Main Stage, Workshop Room A, Workshop Room B), hourly/fixed production crews (AV Team at $85/hr, CloudLens Photography at $200/hr, Bay Area Catering, MerchOps), and corporate sponsors (Acme Corp, ByteScale, Neural Labs). The timeline enforces 17 finish-to-start dependencies (initialDependencies) across Day 1 and Day 2 (e.g., Registration -> Keynote -> MCP Workshop -> Lunch Service -> Open Source Panel -> Hackathon Kickoff).
  3. Bidirectional Cost Sync via onEventResizeEnd and onEventDrop:
    When an organizer drags an event to a different time slot or resizes its duration bar, SchedulerPanel.tsx extracts eventRecord.startDate, eventRecord.endDate, and eventRecord.resourceId, computes the exact hour duration (ms / 3_600_000), and invokes onEventChange(). React state immediately recalculates the item's totalCost and updates AG Grid and the KPI Strip on the same frame.
  4. Payment-State Visual Feedback:
    Event bars dynamically bind their eventColor to operational and payment state—transitioning to emerald green (#22c55e) the moment a speaker or vendor payout is approved and completed via PayPal.

4.3 APIMatic & PayPal Three-Layer Payment Architecture

Implementation Files:

Why a Three-Layer Architecture?
PayPal's @paypal/paypal-server-sdk (v2.5.0, generated by APIMatic) provides strongly-typed controllers for Orders, Payments, Subscriptions, TransactionSearch, Vault, and OAuthAuthorization, while B2B Invoicing and batch Payouts are exposed via @paypal/agent-toolkit and PayPal's REST v1/v2 endpoints. Greenroom combines all three layers cleanly:

Layer Technology / Package Endpoints / Controllers Used Role in Greenroom
Layer 1: Core SDK & Auth @paypal/[email protected] (APIMatic) Client, Environment.Sandbox, OAuthAuthorizationController.requestToken, OrdersController, TransactionSearchController Type-safe OAuth 2.0 client_credentials token generation (Basic Auth), workshop/booth order management, and sandbox transaction lookups.
Layer 2: Inbound Receivables PayPal Invoicing v2 API + @paypal/[email protected] POST /v2/invoicing/invoices
POST /v2/invoicing/invoices/{id}/send		 Autonomous creation and dispatch of itemized sponsor invoices (Gold, Silver, Bronze packages) when triggered via Greenroom AI.
Layer 3: Outbound Payables PayPal Payouts v1 API (Reusing APIMatic SDK OAuth Token) POST /v1/payments/payouts Human-approved batch disbursements for speaker honoraria, hourly AV/photography crews, and hackathon prize winners.

Human-in-the-Loop Financial Safety Gate:
Inbound sponsor invoicing is safe to execute autonomously, so Greenroom AI dispatches invoices immediately and logs the create_and_send_invoice call in the Agent Trace. Outbound payouts move actual funds out of the event account, so Greenroom enforces a strict approval gate:

  1. When asked to "Pay Sarah Chen for the keynote", the agent identifies Sarah Chen's payee email ([email protected]) and honorarium ($350) from the WorkItem state and returns a pendingApproval payload.
  2. The UI displays ApprovalToast.tsx with the exact recipient, description, and dollar amount, giving the organizer explicit Approve and Reject controls.
  3. Only when the organizer clicks Approve does the frontend call POST /api/approve, which invokes sendPayout() against PayPal Sandbox, records the payout_batch_id in the Agent Trace, marks the item completed in AG Grid and Bryntum, and increments Paid Out in the KPI Strip.

4.4 Render Production Blueprint

Implementation File: render.yaml

Greenroom is packaged as an Infrastructure-as-Code Render Blueprint (render.yaml) that builds the TypeScript/Vite frontend (npm install && npm run build) and runs the Express production server (npm run start with NODE_ENV=production), serving both the compiled SPA bundle from dist/ and the /api/* endpoints on a single service with /api/health readiness probes.

5. Development with Sponsor MCP Servers

During engineering and debugging, three sponsor Model Context Protocol (MCP) servers were used to inspect package exports, component props, and SDK signatures:

  1. AG Grid MCP (npx ag-mcp):
    Used to verify AG Grid v36 module registration (AllCommunityModule, AllEnterpriseModule), programmatic theme composition (themeQuartz.withPart(colorSchemeDarkBlue)), React JSX cellRenderer return conventions in ag-grid-react v36, and setFilterModel / applyColumnState / setRowGroupColumns state application methods.
  2. Bryntum MCP (https://mcp.bryntum.com):
    Used to verify @bryntum/schedulerpro-trial v7.3.8 public npm aliasing, CSS theme import paths (@bryntum/schedulerpro/stockholm-dark.css), and the exact event payload shapes for onEventResizeEnd={({ eventRecord }) => ...} and onEventDrop={({ eventRecords }) => ...}.
  3. APIMatic ContextMatic MCP (https://chatbotapi.apimatic.io/mcp/plugins):
    Used to inspect the APIMatic-generated @paypal/paypal-server-sdk v2.5.0 controller hierarchy and verify how OAuthAuthorizationController.requestToken({ authorization: basicAuth }) issues reusable OAuth 2.0 bearer tokens for downstream PayPal REST endpoints.

6. Backend API Reference

Method & Route Request Body Response Payload Description
POST /api/chat { message: string, workItems: WorkItem[], resources: Resource[] } { response: string, toolCalls: ToolCall[], pendingApproval?: PendingApproval, updatedItems?: Partial<WorkItem>[] } Primary Greenroom AI agent endpoint. Handles budget Q&A, autonomous PayPal invoicing, payout approval proposals, and schedule modifications.
POST /api/grid-ai { prompt: string, currentState?: object, schema?: object } { gridState: { filter?, sort?, rowGroup?, columnVisibility? }, explanation: string } AG Grid AI Toolkit endpoint. Converts natural-language view queries into structured AG Grid state objects.
POST /api/approve { action: "payout", recipientEmail: string, amount: number, currency?: string, note?: string } { response: string, result: { status: "payout_sent", batchId: string }, toolCalls: ToolCall[] } Executes human-approved outbound PayPal Sandbox Payouts and returns the batch ID and trace entry.
GET /api/health None { status: "ok", timestamp: string } Health check endpoint for Render zero-downtime deployments.

7. Installation, Configuration & Local Execution

Prerequisites

  • Node.js: v20.x or later
  • PayPal Developer Account: Free Sandbox Client ID & Secret from the PayPal Developer Dashboard
  • LLM API Key: Groq (GROQ_API_KEY), xAI (XAI_API_KEY), OpenAI (OPENAI_API_KEY), or Google Gemini (GOOGLE_GENERATIVE_AI_API_KEY)

Step 1: Clone and Install Dependencies

git clone <YOUR_GITHUB_REPO_URL>
cd greenroom
npm install

Step 2: Configure .env

Copy .env.example to .env and populate your credentials:

cp .env.example .env

# PayPal Developer Sandbox Credentials
PAYPAL_CLIENT_ID=your_paypal_sandbox_client_id
PAYPAL_CLIENT_SECRET=your_paypal_sandbox_client_secret
PAYPAL_ENVIRONMENT=sandbox

# LLM Configuration (supports: groq | xai | openai | google)
LLM_PROVIDER=groq
GROQ_API_KEY=gsk_your_groq_api_key
GROQ_MODEL=openai/gpt-oss-120b

# Server Port
PORT=3001

Step 3: Start the Development Environment

npm run dev

This launches both servers concurrently:

  • Vite React Frontend: http://localhost:5173 (proxies /api/* to port 3001)
  • Express TypeScript API Server: http://localhost:3001

8. End-to-End Verification Scenarios

Scenario 1: Bryntum Schedule Resize to AG Grid Cost Cascade

  1. Locate AV Support: Morning on the AV Team row in Bryntum Scheduler Pro (scheduled 8:00 AM – 12:00 PM, 4 hours at $85/hr = $340).
  2. Drag the right edge of the event bar from 12:00 PM to 2:00 PM (extending duration to 6 hours).
  3. Observe AG Grid Enterprise immediately update AV Support: Morning from 4h / $340 to 6h / $510 (highlighted in amber as > $500), recalculate the AV & Production group subtotal ($4,672.5 -> $4,842.5), and update Total Budget ($13,822.5 -> $13,992.5) and Net (-$5,322.5 -> -$5,492.5) in the header KPI strip.

Scenario 2: AG Grid AI Natural-Language State Control

  1. In the AG Grid AI bar between the scheduler and ledger, enter: text group by type, sort by cost descending
  2. Click Apply. Observe the grid dynamically regroup all 25 rows by type (session, vendor, sponsor, prize, staff) and sort items within each group from highest to lowest cost.
  3. Click Reset to restore the default Budget Category grouping.

Scenario 3: Autonomous PayPal Sponsor Invoicing

  1. Expand the Sponsorship group in AG Grid and note Bronze Sponsorship Package ($1,000 for Neural Labs, [email protected]) has Payment status draft.
  2. In the Greenroom AI sidebar, enter: text Send Neural Labs their sponsor invoice
  3. Click Send to Agent. Observe the Agent Trace log tool: create_and_send_invoice with the PayPal Sandbox invoice dispatch confirmation, the AG Grid row status badge update from draft to sent, and the header Overdue counter increment from 1 to 2.

Scenario 4: Human-in-the-Loop PayPal Speaker Payout

  1. Expand the Speakers group in AG Grid and note Keynote: AI Frontiers ($350 for Sarah Chen, [email protected]).
  2. In the Greenroom AI sidebar, enter: text Pay Sarah Chen for the keynote
  3. Click Send to Agent. Observe the Agent Approval Required ($350) toast appear in the bottom-right corner.
  4. Click Approve. Observe the Agent Trace log tool: paypal_payouts_create with the PayPal Sandbox batch ID, the AG Grid row Payment badge turn green (completed), the Keynote bar in Bryntum Scheduler Pro turn green, and the Paid Out KPI in the header increase from $0 to $350.

9. Repository Structure

greenroom/
├── .github/workflows/
│   └── ci.yml                  # GitHub Actions CI pipeline (typecheck + production build)
├── scripts/
│   └── seed-sandbox.ts         # PayPal Sandbox seeding utility
├── server/
│   ├── index.ts                # Express server & API routes (/api/chat, /api/grid-ai, /api/approve, /api/health)
│   ├── agent-handler.ts        # Greenroom AI agent orchestration & directive execution
│   ├── grid-ai-handler.ts      # Natural-language to AG Grid state compiler
│   ├── paypal-proxy.ts         # APIMatic PayPal Server SDK client, OAuth2, Invoicing v2, Payouts v1
│   └── llm.ts                  # Multi-provider LLM configuration (Groq, xAI, OpenAI, Gemini)
├── src/
│   ├── App.tsx                 # Root layout, unified WorkItem[] state, and approval workflow
│   ├── main.tsx                # Entry point, AG Grid Enterprise registration, Bryntum theme import
│   ├── index.css               # Global dark theme styles & animations
│   ├── components/
│   │   ├── SchedulerPanel.tsx  # Bryntum Scheduler Pro wrapper with dependency & resize/drop handlers
│   │   ├── LedgerGrid.tsx      # AG Grid Enterprise ledger + AG Grid AI bar & Reset control
│   │   ├── KPIStrip.tsx        # Real-time financial KPI header cards
│   │   ├── AIBar.tsx           # Greenroom AI natural-language input component
│   │   ├── AgentTrace.tsx      # Tool invocation & execution trace log
│   │   ├── ApprovalToast.tsx   # Human-in-the-loop PayPal payout approval modal
│   │   └── TransactionLog.tsx  # AG Grid view for PayPal Sandbox transaction search records
│   ├── data/
│   │   ├── resources.ts        # 11 conference resources (stages, crews, sponsors)
│   │   └── work-items.ts       # 25 conference work items & 17 finish-to-start dependencies
│   └── lib/
│       ├── budget.ts           # Hourly cost recalculation & KPI aggregation functions
│       └── types.ts            # Shared TypeScript interfaces
├── .env.example                # Environment variable template
├── index.html                  # HTML entry shell
├── LICENSE                     # MIT License
├── package.json                # Dependencies & build scripts
├── render.yaml                 # Render Blueprint deployment configuration
├── tailwind.config.js          # Tailwind CSS dark palette configuration
├── tsconfig.json               # TypeScript configuration
└── vite.config.ts              # Vite bundler & local API proxy configuration

10. License

This project is licensed under the MIT License — see the LICENSE file for details.

Built With

Share this project:

Updates

Submission history