AI-Powered Analytics: Natural Language Data Queries

Democratizing Data Access

Not everyone can write SQL. AI can translate natural language questions into database queries, making data accessible to everyone in your organization.

Building the Query Interface

class NaturalLanguageQuery
{
    private string $schema;

    public function __construct()
    {
        $this->schema = $this->buildSchemaDescription();
    }

    public function query(string $question): array
    {
        $sql = $this->generateSQL($question);
        $results = DB::select($sql);

        return [
            'question' => $question,
            'sql' => $sql,
            'results' => $results,
            'interpretation' => $this->interpretResults($question, $results),
        ];
    }

    private function generateSQL(string $question): string
    {
        $prompt = <<schema}

Generate a SQL query for: "{$question}"

Rules:
- Use only tables and columns from the schema
- Return only the SQL, no explanation
- Use appropriate aggregations
- Limit results to 100 rows
PROMPT;

        return $this->ai->generate($prompt, ['temperature' => 0]);
    }
}

Schema Description

private function buildSchemaDescription(): string
{
    $tables = [
        'users' => ['id', 'name', 'email', 'created_at'],
        'orders' => ['id', 'user_id', 'total', 'status', 'created_at'],
        'products' => ['id', 'name', 'price', 'category', 'stock'],
        'order_items' => ['id', 'order_id', 'product_id', 'quantity', 'price'],
    ];

    $description = "";
    foreach ($tables as $table => $columns) {
        $description .= "{$table}: " . implode(', ', $columns) . "\n";
    }

    return $description;
}

Result Interpretation

private function interpretResults(string $question, array $results): string
{
    $prompt = <<formatResults($results)}

Provide a brief, natural language interpretation of these results.
Include key insights and notable patterns.
PROMPT;

    return $this->ai->generate($prompt);
}

Safety Considerations

class QuerySanitizer
{
    private array $blockedPatterns = [
        '/\bDROP\b/i',
        '/\bDELETE\b/i',
        '/\bUPDATE\b/i',
        '/\bINSERT\b/i',
        '/\bTRUNCATE\b/i',
    ];

    public function sanitize(string $sql): string
    {
        foreach ($this->blockedPatterns as $pattern) {
            if (preg_match($pattern, $sql)) {
                throw new UnsafeQueryException('Write operations not allowed');
            }
        }

        return $sql;
    }
}

Conclusion

Natural language querying democratizes data access. Always implement safety guards, validate generated SQL, and provide clear feedback when queries can't be answered.