feat: story prompt customisation, LLM Q&A chat, and filter execution fix

[NotYuSheng]

Summary

• Story enhancements (closes feat: allow users to customise the story generation prompt #135): optional additional-context field for generation, collapsible info card (default minimised), StoryChat component with continuous LLM conversation, dynamic suggested questions on generation and after each answer
• Hibernate warning fix: remove @CreationTimestamp from StoryEntity.created_at — let DB DEFAULT CURRENT_TIMESTAMP handle it instead of causing a PreparedStatementDetails warning in the batch pipeline
• Filter execution fix: tshark 3.6.2 silently produces no output when -Y is combined with -T fields; replaced with -2 -R (two-pass read filter) in both validateDisplayFilter and filterPackets

Changes

Backend

• GenerateStoryRequest — optional additionalContext body field for POST /story/generate/{fileId}
• StoryQuestionRequest — question + full conversation history for POST /story/{storyId}/ask
• StoryAnswerResponse — includes followUpQuestions: List<String> alongside answer
• StoryService — threads additionalContext into the user prompt; askQuestion returns JSON {answer, followUpQuestions} in one LLM call; suggestedQuestions added to story generation schema and parsed
• StoryEntity — created_at marked insertable=false, @CreationTimestamp removed
• FilterService — validateDisplayFilter and filterPackets changed from -Y to -2 -R

Frontend

• StoryInfoCard — collapsible (default closed), context textarea inside card body with suggestion placeholder text
• StoryChat — chat bubble UI, scrolls within card (no page jump), suggestion buttons above input, refreshed after each answer, conversation history sent with every request
• StoryPage — wires context + storyId through to both components
• Story type — suggestedQuestions?: string[]

Test plan

• Generate a story — 3 suggested questions appear in the chat card
• Ask a question — answer appears, 3 follow-up suggestions replace the previous ones
• Ask a follow-up referencing the previous answer — LLM uses conversation history correctly
• Leave additional-context blank — story generation unchanged
• Add additional context, regenerate — context is appended to prompt
• Execute a filter on a PCAP — matching packets returned (not empty)
• Execute an invalid filter — still gets rejected with a syntax error message

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request introduces an interactive Q&A feature for network traffic stories, enabling users to ask the LLM specific questions about the analysis. Key changes include the addition of a StoryChat frontend component, new backend endpoints for Q&A and context-aware story generation, and updates to the tshark filtering logic. The review feedback suggests improving the API controllers by using UUID types for path variables to leverage Spring's automatic validation and removing redundant fully qualified class names in the service layer.

[NotYuSheng]

/gemini review

[gemini-code-assist]

Code Review

This pull request introduces an interactive Q&A feature for network traffic stories, allowing users to ask follow-up questions to the LLM. Key changes include new backend endpoints and service logic for handling conversation history and suggested questions, along with the addition of a StoryChat frontend component. The StoryInfoCard was also updated to allow users to provide additional context during story generation. Furthermore, tshark command arguments were adjusted for improved filtering compatibility. The review feedback suggests enhancing type safety by using UUID for path variables and simplifying redundant class references.