DocumentationChatUseCase.php

• Pfad: src/UseCases/Docs/DocumentationChatUseCase.php
• Namespace: UseCases\Docs
• Zeilen: 100 | Größe: 2,801 Bytes
• Geändert: 2025-12-25 13:01:59 | Gescannt: 2025-12-31 10:22:15

Code Hygiene Score: 100

• Dependencies: 100 (25%)
• LOC: 100 (20%)
• Methods: 100 (20%)
• Secrets: 100 (15%)
• Classes: 100 (10%)
• Magic Numbers: 100 (10%)

Keine Issues gefunden.

Dependencies 4

• constructor Infrastructure\Docs\ChunkSearchService
• constructor Infrastructure\AI\OllamaService
• use Infrastructure\AI\OllamaService
• use Infrastructure\Docs\ChunkSearchService

Klassen 1

• DocumentationChatUseCase class Zeile 12

Funktionen 4

• __construct() public Zeile 14
• execute() public Zeile 28
• buildContext() private Zeile 64
• buildPrompt() Zeile 81

Verwendet von 2

• DocsController.php constructor
• DocsController.php use

Versionen 4

• v4 2025-12-25 13:01 | claude-code-hook | modified
  Claude Code Pre-Hook Backup vor Edit-Operation
• v3 2025-12-25 13:01 | claude-code-hook | modified
  Claude Code Pre-Hook Backup vor Edit-Operation
• v2 2025-12-23 08:17 | claude-code-hook | modified
  Claude Code Pre-Hook Backup vor Edit-Operation
• v1 2025-12-23 07:56 | claude-code-hook | modified
  Claude Code Pre-Hook Backup vor Edit-Operation

Code

<?php

declare(strict_types=1);

namespace UseCases\Docs;

// @responsibility: RAG-basierter Chat über Dokumentation

use Infrastructure\AI\OllamaService;
use Infrastructure\Docs\ChunkSearchService;

class DocumentationChatUseCase
{
    public function __construct(
        private ChunkSearchService $searchService,
        private OllamaService $ollamaService
    ) {
    }

    /**
     * Execute documentation chat.
     *
     * @param string $question User question
     * @param string $model LLM model to use
     * @param int $limit Number of context chunks
     * @return array{answer: string, sources: array<int, array{id: int, path: string, title: string, score: float}>}
     */
    public function execute(string $question, string $model = 'mistral', int $limit = 5): array
    {
        // Get relevant chunks via semantic search
        $chunks = $this->searchService->search($question, $limit);

        if (empty($chunks)) {
            return [
                'answer' => 'Leider konnte ich keine relevanten Informationen in der Dokumentation finden.',
                'sources' => [],
            ];
        }

        // Build context from chunks
        $context = $this->buildContext($chunks);

        // Generate answer using LLM
        $prompt = $this->buildPrompt($question, $context);
        $answer = $this->ollamaService->generate($prompt, $model, 0.3);

        // Map sources
        $sources = array_map(static fn (array $chunk): array => [
            'id' => $chunk['doc_id'],
            'path' => $chunk['path'],
            'title' => $chunk['title'],
            'score' => round($chunk['score'], 3),
        ], $chunks);

        return [
            'answer' => $answer,
            'sources' => $sources,
        ];
    }

    /**
     * Build context from chunks.
     */
    private function buildContext(array $chunks): string
    {
        $parts = [];

        foreach ($chunks as $chunk) {
            $part = "## {$chunk['title']}\n";
            $part .= "Pfad: {$chunk['path']}\n";
            $part .= $chunk['content'];
            $parts[] = $part;
        }

        return implode("\n\n---\n\n", $parts);
    }

    /**
     * Build chat prompt.
     */
    private function buildPrompt(string $question, string $context): string
    {
        return <<<PROMPT
            Du bist ein Dokumentations-Assistent. Beantworte die Frage basierend auf dem bereitgestellten Kontext.

            KONTEXT:
            {$context}

            FRAGE:
            {$question}

            ANLEITUNG:
            - Antworte auf Deutsch
            - Sei präzise und hilfreich
            - Wenn der Kontext die Frage nicht beantwortet, sage das ehrlich
            - Verweise auf die relevanten Abschnitte der Dokumentation
            PROMPT;
    }
}