Skip to content

codegen-tools.md

Latest commit

 

History

History
194 lines (138 loc) · 5.29 KB

codegen-tools.md

File metadata and controls

194 lines (138 loc) · 5.29 KB

Code Generation Tools

The Code Execution Engine provides built-in tools for code generation that can be registered with Wanaku and invoked by AI agents.

Overview

When configured with a code generation package, the engine registers three tools:

Tool Description
searchServicesTool Lists available services (Kamelets) that can be used in orchestrations
readKamelet Reads the YAML content of a specific Kamelet by name
generateOrchestrationCode Returns the orchestration template for code generation

Configuration

Command Line Option

java -jar camel-code-execution-engine.jar \
  --codegen-package <path-or-uri> \
  --name my-service \
  ...

Environment Variable (Docker)

CODEGEN_PACKAGE=/path/to/package

Input Formats

The --codegen-package option accepts two formats:

Format Example Description
Local directory /path/to/my-package Uses a local directory directly
Datastore URI datastore-archive://package.tar.bz2 Downloads and extracts from Wanaku data store

Package Structure

The code generation package must have the following structure:

my-package/
├── config.properties      # Required: Configuration file
├── kamelets/              # Required: Directory containing Kamelet YAML files
│   ├── http-source.kamelet.yaml
│   ├── kafka-sink.kamelet.yaml
│   └── ...
└── templates/             # Required: Directory containing templates
    └── orchestration.txt  # Required: Orchestration template

config.properties

# Comma-separated list of available services
available.services=kamelet:http-source,kamelet:kafka-sink,kamelet:log-action

# Optional: Custom description for the search tool
search.tool.description=Searches for available integration services

# Optional: Namespace for tool registration (defaults to no namespace)
namespace=my.custom.namespace
Property Required Default Description
available.services Yes - Comma-separated list of available services
search.tool.description No "Searches for services to perform the tasks" Description for the searchServicesTool
namespace No null (no namespace) Namespace for tool registration in Wanaku

Kamelet Files

Each file in the kamelets/ directory should be a valid Kamelet YAML file with the .kamelet.yaml extension:

apiVersion: camel.apache.org/v1alpha1
kind: Kamelet
metadata:
  name: http-source
spec:
  definition:
    title: HTTP Source
    description: Fetches data from an HTTP endpoint
  # ... rest of Kamelet spec

Orchestration Template

The templates/orchestration.txt file contains the template returned by the generateOrchestrationCode tool. This template guides the AI agent in generating orchestration code.

Tool Registration

Tools are registered with Wanaku using the service name as the URI scheme:

<service-name>://searchServicesTool
<service-name>://readKamelet
<service-name>://generateOrchestrationCode

For example, with --name code-execution-engine:

code-execution-engine://searchServicesTool
code-execution-engine://readKamelet
code-execution-engine://generateOrchestrationCode

Tool Details

searchServicesTool

Returns a formatted list of available services with context explaining how to use Kamelets.

Parameters: None

Response:

# Context
- The list below contains Kamelets that can be used to assemble the orchestration.
- A Kamelet is a snippet for a Camel route.
- It cannot be used on its own. It MUST be a part of an orchestration flow.
- Kamelets can have arguments. Read the Kamelets before you use them.

---
kamelet:http-source
kamelet:kafka-sink
kamelet:log-action

readKamelet

Reads and returns the complete YAML content of a Kamelet.

Parameters:

Name Type Required Description
name string Yes The Kamelet name (without .kamelet.yaml extension)

Response: The full YAML content of the Kamelet.

generateOrchestrationCode

Returns the orchestration template content.

Parameters: None

Response: The content of templates/orchestration.txt.

Example Usage

Using a Local Directory

For development and testing, you can point to a local directory:

java -jar camel-code-execution-engine.jar \
  --codegen-package /home/user/my-codegen-package \
  --name my-service \
  --registration-url http://localhost:8080 \
  --client-id my-client \
  --client-secret my-secret \
  --registration-announce-address localhost

Using Datastore Archive

For production, upload a tar.bz2 archive to the Wanaku data store and reference it:

java -jar camel-code-execution-engine.jar \
  --codegen-package datastore-archive://codegen-package.tar.bz2 \
  --name code-execution-engine \
  --registration-url http://wanaku.example.com \
  --client-id my-client \
  --client-secret my-secret \
  --registration-announce-address auto

Creating a Package Archive

To create a tar.bz2 archive for upload to the data store:

# From the parent directory of your package
tar -cjf codegen-package.tar.bz2 my-package/

Then upload codegen-package.tar.bz2 to the Wanaku data store.