watsonx Orchestrate demos

Playground for experimenting with watsonx Orchestrate.

Prerequisites

• Python 3.11 or higher
• uv package manager for Python here

watsonx Orchestrate Setup

Remote environment

• Sign up for a free trial (Try it free button, and log-in with your IBM ID) here
• Create an API key: Click on the avatar in the top right corner, then click on "Settings", then click on "API details", then click on "Generate API key".
• Save the created API key and Service instance URL for later use.

orchestrate CLI tool

Install orchestrate CLI tool:

uv sync
alias orchestrate='uv run orchestrate'
# or even shorter
alias wxo='uv run orchestrate' 

And run orchestrate --help (or wxo --help) to verify the installation.

Local environment

• Get a remote environment set up.

• Get entitlement key from MyIBM

• Fill up the .adk file in the root of the repository with the following content:

  WO_DEVELOPER_EDITION_SOURCE=myibm
  WO_ENTITLEMENT_KEY=<my_entitlement_key>
  WO_INSTANCE=<service_instance_url>
  WO_API_KEY=<wxo_api_key>
  WO_USERNAME=<your_wxo_email>
  WO_PASSWORD=<your_wxo_password_for_cp4d>
  

  OR you can use remote watsonx.ai instance for LLM:

  WO_DEVELOPER_EDITION_SOURCE=myibm
  WO_ENTITLEMENT_KEY=<my_entitlement_key>
  WATSONX_APIKEY=<my_watsonx_api_key>
  WATSONX_SPACE_ID=<my_space_id>
  

• Run the server to start the local environment:

  orchestrate server start --env-file=.adk -l # -l enables observability platform

Attaching to a remote environment

Add the remote environment (if not already added):

orchestrate env add --name my-remote-env --url <SERVICE_INSTANCE_URL>

Activate the remote environment:

orchestrate env activate my-remote-env
# And paste the API key when prompted.

You should see something like this:

[INFO] - Environment 'my-remote-env' is now active

Attach to the local environment

You can run orchestrate env activate local to set your environment or orchestrate chat start to start the UI service and begin chatting.

if you want to enable the observability platform, add -l to the server start command.

You can access the observability platform Langfuse at http://localhost:3010, username: [email protected], password: orchestrate

Stop the local environment

To stop the local environment, run:

orchestrate server stop

Working with local models (optional)

Ollama

Model name: virtual-model/ollama/*

ollama pull llama3.2:latest
export OLLAMA_HOST=0.0.0.0:11434
ollama serve

# Test ollama model
curl --request POST \
  --url http://192.168.68.59:11434/v1/chat/completions \
  --header 'content-type: application/json' \
  --data '{
  "model": "llama3.2:latest",
  "messages": [
    {
      "content": "Hi",
      "role": "user"
    }
  ]
}'

Add the model

orchestrate models import --file models/ollama.yaml
[INFO] - Successfully added the model 'virtual-model/ollama/llama3.1:8b'

Remove the model

orchestrate models remove -n virtual-model/ollama/llama3.1:8b
[INFO] - Successfully removed the model 'virtual-model/ollama/llama3.1:8b'

Usage

To list all available agents in the active environment:

orchestrate agents list

To list all available tools in the active environment:

orchestrate tools list

To list all available connections in the active environment:

orchestrate connections list

Demos

1. IP Reputation Checker - An agent that checks the reputation of IP addresses using VirusTotal

   This demo showcases an agent that can evaluate the security reputation of any IP address using the VirusTotal API. The agent provides detailed information about whether an IP is malicious, which security vendors flagged it, its geographic location, and associated threats.

   Setup

   1. Make sure you have a VirusTotal API key. You can sign up for free at VirusTotal.
   2. Set your API key as an environment variable:

      export VIRUSTOTAL_API_KEY=your_api_key_here

   3. Deploy the agent and required components:

      ./demos/demo_1_deploy.sh

   Usage

   Once deployed, you can interact with the agent by providing any IP address you want to check. The agent will:

   • Fetch the IP's reputation data from VirusTotal
   • Analyze the security verdicts
   • Present findings in a structured format showing malicious indicators, geographic data, and security details

   Cleanup

   To remove the agent, tools, and connections created for this demo:

   ./demos/demo_1_clean.sh

2. YouTube Transcriber - An agent that creates transcripts from YouTube videos

   This demo features an agent that can extract and provide text transcripts from any YouTube video. It retrieves video metadata and generates a readable transcript that captures the spoken content.

   Setup

   1. Deploy the agent and required components:

      ./demos/demo_2_deploy.sh

   Usage

   After deployment, you can interact with the agent by providing a YouTube URL. The agent will:

   • Extract the video information (title, channel, etc.)
   • Generate a text transcript of the video content
   • Format the transcript for readability
   • For longer videos, provide a summary of key points

   Cleanup

   To remove the agent, tools, and connections created for this demo:

   ./demos/demo_2_clean.sh

3. Content Extractor - An agent that extracts content from web pages

   This demo showcases an agent that can extract meaningful content from any web page URL. It retrieves the main text, images, and metadata, providing a structured summary of the page's content.

   Setup

   1. Deploy the agent and required components:

      ./demos/demo_3_deploy.sh

   Usage

   After deployment, you can interact with the agent by providing a URL. The agent will:

   • Fetch the web page content
   • Extract the main text, images, and metadata
   • Provide a structured summary of the page's content

   Cleanup

   To remove the agent, tools, and connections created for this demo:

   ./demos/demo_3_clean.sh

4. Patents generator (WIP) - An agent that generates patents based on a given topic

   This demo features an agent that can generate patent documents based on user-provided topic. It uses a template to create structured patent applications.

   Setup

   1. Deploy the agent and required components:

      ./demos/demo_4_deploy.sh

   Usage

   After deployment, you can interact with the agent by providing a description of the invention. The agent will:

   • Generate a structured patent document using the provided description
   • Format the document according to standard patent application guidelines

   Cleanup

   To remove the agent, tools, and connections created for this demo:

   ./demos/demo_4_clean.sh

5. Procurement Assistant - An agent that analyzes Bill of Materials (BOM) files

   This demo features a specialized agent that can interpret and answer questions about Excel-based Bill of Materials (BOM) files. It helps procurement teams quickly analyze component information.

   Setup

   1. Deploy the agent and required components:

      ./demos/demo_5_deploy.sh

   Usage

   After deployment, add the BOM manually via the Agents builder UI. The agent will:

   • Analyze the provided Excel BOM file
   • Answer specific questions about components, costs, quantities, etc.
   • Help identify procurement needs and opportunities

   Cleanup

   To remove the agent, tools, and connections created for this demo:

   ./demos/demo_5_clean.sh

Notes:

• (Demo 8) Playwright MCP server over SSE: npx @playwright/mcp@latest --port 8931 --host 0.0.0.0

• (Demo 9) Elasticsearch MCP server over SSE:

  git clone https://github.com/elastic/mcp-server-elasticsearch.git
  cd mcp-server-elasticsearch
  npm install
  npm run build
  
  HOST=0.0.0.0 uv tool run mcp-proxy --transport sse --port 9201 --host 0.0.0.0 \
    -e ES_URL http://<ES_IP>:9200 \
    -e ES_API_KEY "" \
    node /Users/barha/Desktop/IBM/projects/mcp-server-elasticsearch/dist/index.js