As a Nablarch developer, I want AI to answer Javadoc-level details so that I don't need to look up external Javadoc manually

[kiyotis]

Situation

Nablarch official documentation frequently defers to Javadoc for detailed API information (e.g., "詳細はJavadocを参照"). Nabledge knowledge files currently contain only the documentation content, not the Javadoc itself. As a result, when users ask about specific method behavior (e.g., UniversalDao#exist), the AI lacks the information to answer.

A jar tool that generates Markdown from Nablarch source files downloaded via Maven already exists. The jar will be provided at implementation time.

Pain

Nablarch developers ask nabledge about specific API behavior and receive incomplete answers because the knowledge files do not include Javadoc content. They must look up the Javadoc manually, which defeats the purpose of having an AI assistant.

Benefit

• Developers can ask nabledge about any Nablarch API and receive accurate, complete answers without consulting external Javadoc
• The AI can answer questions about methods, parameters, return values, exceptions, and behavior described only in Javadoc

Success Criteria

• Investigate: Identify the input/output format of the provided jar tool (source file → Markdown)
• Investigate: Determine appropriate integration point in the RBKC pipeline across all 5 versions (v6 / v5 / v1.4 / v1.3 / v1.2)
• Javadoc Markdown is generated for all public Nablarch classes in all 5 versions using the jar tool
• Generated Javadoc content is included in knowledge files and passes rbkc.sh verify for all 5 versions
• nabledge can correctly answer questions about Javadoc-deferred API details (e.g., UniversalDao#exist behavior, parameters, return value)
• Benchmark scores for all 5 versions do not decrease vs. baseline after adding Javadoc knowledge

[kiyotis]

Related user-facing request: nablarch/nabledge#15