v0.9.5

What's new

AI Assistant

• Chatbot now answers any question about scan results using your configured LLM (OpenAI, Azure, Ollama, or any OpenAI-compatible endpoint)
• Added batch explain — select "All findings" and say explain to get root causes, top priority action, and team insights across all findings
• Fixed command parsing for natural phrasing like "put value 20 in estimate"
• Write field values directly to GitHub from chat — say set estimate 5, review the preview, then apply it. No AUTO_APPLY=true required for interactive chat.

LLM providers

• Added Ollama support for 100% local, no-API-key usage
• Added openai-compatible provider for any local or remote OpenAI-compatible endpoint
• All LLM settings configurable in sidebar and persisted to .env

Sidebar & settings

• Save all settings to .env button — every sidebar value persists across restarts
• Removed "Save LLM settings" (merged into the single save button)
• Password fields no longer trigger browser save-password prompts
• Scan Scope checkboxes now correctly initialise from .env

Table filters

• Added Updated from / Updated to date pickers to filter findings by last-updated date
• Removed redundant title keyword search (Streamlit's built-in table search covers this)

UI

• AI assistant moved to a popover panel (top-right "AI" button), 600 px wide
• Removed "Accounts to watch" sidebar section (redundant with table filters)

GitHub / project

• Added MIT license
• Added CI workflow (ruff, pyright, pytest on every push/PR)
• Added issue templates and PR template