Vizra ADK Tool Creation
# ClaudeWave Editorial This Claude Code item provides patterns and a template structure for building custom tools in the Vizra ADK framework, including how to implement the ToolInterface, define tool schemas for LLM integration, and structure parameters for database, API, file, and email interactions. Use this when creating new agent tools that need to extend Vizra ADK agent capabilities with specific functionality and proper LLM-readable definitions.
git clone --depth 1 https://github.com/vizra-ai/vizra-adk /tmp/vizra-adk-tool-creation && cp -r /tmp/vizra-adk-tool-creation/resources/boost/skills/vizra-tool-creation ~/.claude/skills/vizra-adk-tool-creationSKILL.md
# Building Vizra ADK Tools
Tools extend agent capabilities by allowing them to interact with databases, APIs, files, and external services.
## Tool Class Structure
Every tool MUST implement `ToolInterface`:
```php
<?php
namespace App\Tools;
use Vizra\VizraADK\Contracts\ToolInterface;
use Vizra\VizraADK\Agents\AgentContext;
class {{ ToolName }}Tool implements ToolInterface
{
/**
* Define the tool's schema for the LLM
*/
public function definition(): array
{
return [
'name' => '{{ tool_name }}',
'description' => '{{ What this tool does - be specific }}',
'parameters' => [
'type' => 'object',
'properties' => [
'param1' => [
'type' => 'string',
'description' => '{{ Parameter description }}'
],
'param2' => [
'type' => 'integer',
'description' => '{{ Parameter description }}'
],
],
'required' => ['param1'],
],
];
}
/**
* Execute the tool with given arguments
*/
public function execute(array $arguments, AgentContext $context): string
{
// Tool logic here
return json_encode([
'status' => 'success',
'data' => $result
]);
}
}
```
## Key Rules
1. **Naming Convention**:
- Class name MUST end with `Tool`
- The `name` in definition should be snake_case
2. **Definition Requirements**:
- `name`: Unique identifier for the tool
- `description`: Clear description of what the tool does (LLM uses this to decide when to use the tool)
- `parameters`: JSON Schema format describing expected inputs
3. **Return Format**:
- Always return JSON-encoded strings
- Include status indicators for success/error
- Provide meaningful error messages
## Parameter Types
```php
'parameters' => [
'type' => 'object',
'properties' => [
// String parameter
'name' => [
'type' => 'string',
'description' => 'The user name'
],
// Integer parameter
'count' => [
'type' => 'integer',
'description' => 'Number of items'
],
// Boolean parameter
'active' => [
'type' => 'boolean',
'description' => 'Whether the item is active'
],
// Enum parameter
'status' => [
'type' => 'string',
'enum' => ['pending', 'approved', 'rejected'],
'description' => 'Current status'
],
// Array parameter
'tags' => [
'type' => 'array',
'items' => ['type' => 'string'],
'description' => 'List of tags'
],
// Object parameter
'metadata' => [
'type' => 'object',
'properties' => [
'key' => ['type' => 'string'],
'value' => ['type' => 'string']
],
'description' => 'Additional metadata'
],
],
'required' => ['name', 'count'],
]
```
## Common Tool Patterns
### Database Query Tool
```php
class DatabaseQueryTool implements ToolInterface
{
public function definition(): array
{
return [
'name' => 'query_database',
'description' => 'Execute a read-only database query to retrieve information',
'parameters' => [
'type' => 'object',
'properties' => [
'table' => [
'type' => 'string',
'description' => 'The table to query'
],
'conditions' => [
'type' => 'object',
'description' => 'WHERE conditions as key-value pairs'
],
'limit' => [
'type' => 'integer',
'description' => 'Maximum number of results'
],
],
'required' => ['table'],
],
];
}
public function execute(array $arguments, AgentContext $context): string
{
$allowedTables = ['users', 'orders', 'products'];
if (!in_array($arguments['table'], $allowedTables)) {
return json_encode([
'status' => 'error',
'message' => 'Table not allowed'
]);
}
$query = DB::table($arguments['table']);
if (isset($arguments['conditions'])) {
foreach ($arguments['conditions'] as $key => $value) {
$query->where($key, $value);
}
}
$results = $query->limit($arguments['limit'] ?? 10)->get();
return json_encode([
'status' => 'success',
'data' => $results,
'count' => $results->count()
]);
}
}
```
### API Integration Tool
```php
class ExternalApiTool implements ToolInterface
{
public function definition(): array
{
return [
'name' => 'call_external_api',
'description' => 'Make requests to the external service API',
'parameters' => [
'type' => 'object',
'properties' => [
'endpoint' => [
'type' => 'string',
'description' => 'API endpoint path'
],
'method' => [
'type' => 'string',
'enum' => ['GET', 'POST'],
'description' => 'HTTP method'
],
'data' => [
'type' => 'object',
'description' => 'Request payload for POST requests'
],
],
'required' => [Create AI agents with Vizra ADK - includes patterns for customer service, data analysis, and content generation agents
Test and evaluate AI agents with automated evaluations, assertions, and LLM-as-a-Judge patterns
Implement persistent memory, session context, and vector memory (RAG) for AI agents
Orchestrate complex multi-agent workflows - sequential, parallel, conditional, and loop patterns