OpenEdge × LiveKit AI Agent Workshop

Build an AI agent that reads and writes to a Progress OpenEdge database and books automotive service appointments. This workshop is designed for OpenEdge developers with no prior Python/agent experience. You’ll follow a guided path, copying working examples at each step and learning how they fit together.

We’ll use PASOE WEB Transport services for data access, a small Python driver to call those services, and a LiveKit agent (with OpenAI) that converses with users, manages car records, and schedules bookings. We’ll initially run the agent through the LiveKit Agents Playground and if time, implement a React front end and python token service to complete the app.

---

What You’ll Build

• OpenEdge WEB transport services for Car and Booking records.
• A small, testable Python driver (OEDatabaseDriver) that calls those REST endpoints.
• A LiveKit agent that uses function tools to call your driver:
  • Lookup/add car details.
  • Offer/confirm bookings.
• A multi-agent version showing a handoff from an Account agent → Booking agent.
• (Optional, if time permits) A lightweight React + Token server demo UI.
• (Optional, if time permits) A bookings MCP Server and dynamic MCP Client.

You’ll copy the code for each step from the provided examples/ folders and focus on wiring/config + understanding.

---

Prerequisites

• Progress OpenEdge Developers Kit (CLASSROOM edition is sufficient) — Progress OpenEdge Developers Kit
• Visual Studio Code (Code editor for Python and JavaScript) — Visual Studio Code
• Node.js (for React frontend) — Node.js
• Python 3.10+ (venv)
• A modern browser with microphone access (if testing voice)
• Accounts/keys:
  • LiveKit (API key/secret, URL)
  • OpenAI (API key)

You’ll place secrets in a local .env file during setup.

---

Useful Links

• Progress OpenEdge Developers Kit
• Visual Studio Code
• Python downloads
• LiveKit
• LiveKit Agents Playground
• OpenAI Platform
• Node.js

---

Repository Layout (at a glance)

/openedge/            # ABL handlers/services templates for Car/Booking and test app
/python/
  /step2/             # OEDatabaseDriver env sanity-check
  /step3/             # save_car() via POST
  /step4/             # get_car() via GET + dataclass
  /step5/             # Single-agent (car lookup/add) with LiveKit
  /step6/             # Adds booking tools + driver methods
  /step7/             # Multi-agent architecture AccountAgent → BookingAgent handoff
  /step10/            # (Optional) MCP Server and dynamic client
/frontend/            # (Optional) React demo UI
/token-server/        # (Optional) LiveKit token server (Python)
README.md

---

Workshop Steps (copy code from examples; focus on config + understanding)

Step 1: OpenEdge Setup – Server, Database, and REST Service

1. Create working directories:

   C:\OpenEdge\WRK\oeautos
   C:\OpenEdge\WRK\oeautos\db
   

2. Create an empty database:

   prodb oeautos empty
   

3. Load schema file:

   load oeautos.df
   

4. Add oeautos database to Progress OE Explorer and start it (make a note of the port number i.e 25150).

5. Create a new Progress Application Server with the following settings:

   • Instance name: oeautos
   • AdminServer: localhost
   • Location: Local
   • Security model: Developer
   • Instance Directory: C:\OpenEdge\WRK\oeautos\PAS
   • Autostart: checked

6. Edit the oeautos ABL application configuration:

   • Startup parameters: -db oeautos -H localhost -S 25150
   • Add C:\OpenEdge\WRK\oeautos to PROPATH

7. Start the AppServer.

8. In OE Developer:

   • Choose C:\OpenEdge\WRK\oeautos as the new workspace
   • Create a new OpenEdge project named AgentTools
   • Server project
   • Transport: WEB
   • Finish → Open Perspective
   • Delete AgentTools → PASOEContent → WEB-INF → openedge → AgentToolsHandler.cls (we will create our own)
   • Delete Defined Services → AgentToolsService (we will create are own)

9. Add database to project:

   • Right-click project → Properties → Progress OpenEdge → Database Connections → Configure database connections
   • Click New to add new connection for oeautos server.
   • Connection name: oeautos
   • Physical name: db\oeautos
   • Host name: localhost
   • Service/Port: 25150
   • Press Next to Finish, then Apply and Close
   • Select checkbox next to the new connection, Apply and Close

10. Add Server to project:

    • Right-click project → New → Server
    • Select Progress Software Corporation → Progress Application Server for OpenEdge, press Next
    • Press Configure
    • Select your local Explorer connection in the list, and press Edit
    • Enter the correct User name and Password values to connect to OpenEdge Explorer, press Finish
    • Press Apply and close
    • Select [machine-name].oeautos in Progress Application Server for OpenEdge
    • Select oeautos under ABL Application and press Finish

11. Add a new web handler:

    • Name: carHandler
    • Select GET and POST method stubs
    • Copy in template code

12. Add WEB service:

    • Right-click project → New → ABL Service
    • Transport: WEB
    • Service name: carService
    • Under WebHandler, select Select Existing
    • Press Browse
    • Select the class carHandler and press OK
    • Press Add to add a Resource URI
    • Enter /carService as the Resource URI
    • Press Finish

13. Publish the service to the oeautos server:

    • In the Servers panel, right click oeautos in .. and select Add and Remove
    • Select carService in the Available list, and press Add>
    • Press Finish

---

Step 2: Python Driver – Environment & Connectivity

See the Step 2 documentation for details.

1. Install VS Code and Python (if you don't already have them):

   • Visual Studio Code
   • Python downloads

2. Create directory C:\Work\Agent and open it in VS Code.

3. Create and activate a virtual environment:

   py -m venv .venv
   .\.venv\Scripts\Activate.ps1

4. Create a C:\Work\Agent\requirements.txt file and add:

   python-dotenv
   

   Then install:

   pip install -r requirements.txt

5. Create a .env file with:

   OE_SERVICE_URL=http://localhost:8080/AgentTools/web/
   

6. Copy OEDatabaseDriver.py Step 2 code to C:\Work\Agent\OEDatabaseDriver.py (load .env and print OE_SERVICE_URL).

7. Run the code to test:

   py OEDatabaseDriver.py

---

Step 3: Python Driver – Saving Cars

See the Step 3 documentation for details.

1. Copy requirements.txt and OEDatabaseDriver.py from Step 3 into C:\Work\Agent (note: requests has been added to requirements.txt).

2. Install new dependencies:

   pip install -r requirements.txt

3. Review the save_car method.

4. Run the code to test:

   py OEDatabaseDriver.py

5. Use OpenEdge\ViewCars.w to confirm the record has been created.

---

Step 4: Python Driver – Retrieving Cars

See the Step 4 documentation for details.

1. Copy OEDatabaseDriver.py from Step 4 into C:\Work\Agent.

2. Review the get_car method.

3. Run the code to test:

   py OEDatabaseDriver.py

---

Step 5: Introducing the AI Agent (LiveKit Integration)

See the Step 5 documentation for details.

1. Copy contents of Step 5 into C:\Work\Agent.

2. Install new dependencies:

   pip install -r requirements.txt

3. Create a LiveKit account:

   • Go to LiveKit
   • Click Start Building and create an account
   • Create your first project (for example: PUG Challenge)
   • Complete any onboarding steps
   • In Settings → API Keys, reveal the secret and copy the credentials
   • Paste the values into your .env file

4. Create an OpenAI account:

   • Create an account at OpenAI Platform
   • From the account settings, create a new API key
   • Add OPENAI_API_KEY=[YOUR KEY] to your .env file and save

5. Review the code.

6. Start the agent:

   py main.py dev

7. Log into the LiveKit Agents Playground and connect.

8. Verify the agent can create and look up cars in the database.

---

Step 6: Booking Functionality

See the Step 6 documentation for details.

1. Add a new OpenEdge web handler for bookings (bookingHandler) — copy code from Step 6.

2. Add a new ABL Service (bookingService) with transport WEB and service name bookingService.

3. Select existing Web Handler: bookingHandler.

4. Add resource URIs:

   • /booking
   • /booking/{method}

5. Save your changes and publish to the server.

6. Copy in code from Step 6 and review.

7. Start the agent:

   py main.py dev

8. Log into the LiveKit Agents Playground and connect.

9. Verify the agent can create and look up bookings in the database.

---

Step 7: Multi-Agent Architecture

See the Step 7 documentation for details.

1. Copy in code from Step 7 and review.

2. Start the agent:

   py main.py dev

3. Log into the LiveKit Agents Playground and check agent behaviour.

---

(Optional) Bonus: React Frontend + Token Server and MCP Server

Step 8: Frontend (optional)

See the Frontend documentation for details.

1. If you don’t already have it, download and install Node.js.

2. Open a new terminal and change directory to C:\Work.

3. Scaffold a Vite React app:

   npm create vite@latest frontend -- --template react

4. Open the frontend directory in a new VS Code window.

5. Install dependencies:

   npm install
   npm install @livekit/components-react @livekit/components-styles livekit-client --save

6. Delete C:\Work\frontend\src\assets and C:\Work\frontend\public (if present).

7. Copy in contents of frontend\step 1 from the examples.

8. Log in to LiveKit, go to Settings → API Keys → Generate Token.

9. Copy the token into src/components/LiveKitModal.jsx (as indicated in the frontend example).

10. Create a .env file and enter:

VITE_LIVEKIT_URL=[YOUR LIVEKIT URL]

11. Run the front end with:

npm run dev

---

Step 9: Token Server (optional)

See the Token Server documentation for details.

1. Create directory C:\Work\TokenServer and open it in VS Code.

2. Open a new terminal.

3. Create a virtual environment:

   py -m venv .venv

4. Activate the virtual environment:

   .\.venv\Scripts\Activate.ps1

5. Copy the TokenServer example files from the workshop.

6. Install requirements:

   pip install -r requirements.txt

7. Enter LiveKit values into .env.

8. Start the server:

   py server.py

9. Copy contents of frontend\step 2 (optional) and run the frontend if desired:

   npm run dev

---

Step 10: MCP Server and dynamic client (optional)

See the MCP Server documentation for details.

1. Copy in code from Step 10 and review.

2. Install requirements:

   pip install -r requirements.txt

3. Start the agent:

   py main.py dev

4. Log into the LiveKit Agents Playground and check agent behaviour.

---

Testing the Agent (suggested flow)

In the LiveKit Agents Playground:

1. The agent greets and asks for your registration.

2. Provide a reg (e.g., AB12 CDE).

   • If found, it shows car details.
   • If not, it asks for make, model, year and adds the car.

3. The agent asks if you’d like to book an appointment or check existing.

4. Try: “Book me in for the next available date for an annual service.”

   • Agent retrieves next slot, confirms, and saves the booking.

5. Ask “What’s my next booking?” to verify.

---

Environment & Scripts

• Python venv: keep dependencies isolated (python -m venv .venv → activate → pip install -r requirements.txt).
• .env: never commit secrets; store OE_SERVICE_URL, LiveKit & OpenAI keys locally.
• Run steps: each python/stepX folder contains a small script (main.py or similar) to run that step.

---

Troubleshooting

• PASOE service not reachable: confirm it’s published, note the correct port/context, and that OE_SERVICE_URL ends with /rest/.
• LiveKit connection: verify LIVEKIT_URL, LIVEKIT_API_KEY, and LIVEKIT_API_SECRET in .env.
• OpenAI: ensure OPENAI_API_KEY is set and your model name is valid for your account.

---

Appendix A — Python Intro for OpenEdge Developers (Quick Notes)

• Whitespace = blocks (no END.). Indentation is syntax.
• Run files directly (python myfile.py); manage deps with venv + pip (requirements.txt).
• Functions use def and return values directly (multiple returns via tuples).
• Dataclasses (@dataclass) create light DTOs for records like Car.
• Requests library is the go‑to for HTTP calls (requests.get/post, r.json()).
• OO Basics: class, __init__, self (like THIS-OBJECT, but explicit). Inherit with class Sub(Super):.

---

License

MIT

---

Acknowledgements

• LiveKit Agents
• Progress OpenEdge
• OpenAI