Since a recent refactor on dev, the prompt and completion token counts of a reply that used tools added up every model call in the loop, even though each later call already re-sends the whole conversation. A reply whose context was 50k tokens showed 90k, the context gauge read over 100% and the next message triggered compaction long before the chat reached the threshold. Prompt and completion tokens show the latest call again, while total tokens and cost still add up across all calls.
New **pt-BR** translations for items introduced in the latest releases, plus a consistency/quality pass across existing strings (grammar, tone, capitalization, pluralization). Placeholders and hotkeys preserved. No logic changes.
* docs: highlight the key points of the security advisory notice
* docs: move per-setting Chat Controls saving to Added
* docs: move accurate upload status to Added
* docs: drop dependency bump entries from Changed
* docs: changelog for faster analytics dashboard
* docs: start 0.12.4 changelog
* docs: link compare view scrolling commit
* docs: update 0.12.4 date
* docs: changelog for last active time, note image upload and prompt history
* docs: changelog for folder menu, accessibility, tool, note and code block fixes and translations
* docs: changelog for permission, archive, database, Windows build and accessibility fixes
* docs: changelog for collection access checks
* docs: changelog for formatted text spacing, FastMCP results and Realtime call end
* docs: changelog for shared chat page reloads
* docs: changelog for keyboard tooltips
* docs: changelog for profile cards, chat preview turns and compaction in short chats
* docs: changelog for chat controls saving and streaming scroll
* docs: changelog for compaction in temporary chats and early new chat updates
* docs: changelog for note text edits and clickable setting labels
* docs: changelog for bulk file actions in Manage Files
* docs: changelog for analytics token chart, profile pictures and setting labels
* docs: changelog for new file and folder in folder menu
* docs: changelog for file browser grid view and hover previews
* docs: changelog for selecting several chats in the sidebar
* docs: changelog for upload processing status
* docs: link closed issues for dictation and skip link
* docs: changelog for Azure reasoning replay and screen reader updates
* docs: link closed issues for Azure reasoning replay and cancelled database writes
* docs: changelog for analytics rework and link closed issues
* docs: move accessibility changes into one Added entry before general improvements
* docs: apply changelog audit fixes for labels, placement and wording
* docs: changelog for analytics comparison, draft variables, OAuth static MCP secrets
* docs: changelog for chat edit and calendar access, web search settings, fork, first message, OAuth API, MinerU key
* docs: changelog for model controls for all users, Python 3.12 image, sidebar and info button accessibility
* docs: changelog for model import errors, DOMPurify 3.4.17, model picker Escape accessibility
* docs: changelog for group save speed, direct tool server checks, PowerPoint titles, hover previews, launcher help, accessibility
* docs: changelog for channel post speed, context gauge threshold, new chat accessibility
* docs: changelog for live chat access after archiving and bulk unsharing
* docs: changelog for archived chats search announcement
* docs: update 0.12.4 changelog date
* docs: changelog for Images settings save, Settings and model picker accessibility
* docs: changelog for sidebar, dropdown and search window accessibility
* docs: drop links from the accessibility entry
* docs: changelog for WebP image editing, tooltip text, Workspace and menu accessibility
* docs: changelog for Archived Chats unarchive focus
* docs: link setting row highlight issue
* docs: update 0.12.4 changelog date
* docs: changelog for hidden unavailable options, zoom accessibility
* docs: changelog for hidden options of disabled features, access button, folder descriptions, knowledge file edits
* docs: changelog for Redis websocket reconnect fix and uvicorn update
* docs: add French to 0.12.4 translation updates
* docs: changelog for organizing chats in shared folders
* docs: changelog for sidebar headings accessibility
* docs: changelog for model picker More menus and Direct Integrations switch
* docs: changelog for swiping open Chat Controls on phones, security advisory notice
* docs: changelog for hidden web search and voice settings and accessibility fixes
* docs: changelog for deleting the open chat, Read Aloud failures and hidden group permission toggles
* docs: changelog for MCP protocol 2026-07-28, stable tool lists and dependency updates
* docs: changelog for LDAP default email domain, Brotli update and hidden terminal and shortcut options
* docs: changelog for first-message terminal uploads and MCP tool argument and error fixes
* docs: changelog for faster folder changes, knowledge and note upload fixes, PDF OCR threads and more hidden options
* docs: changelog for More Concise and Add Details, audit log model usage and knowledge and web page fixes
* docs: changelog for hidden base models permission, sharing changes, file reading tool fixes and accessibility
* docs: changelog for questions that survive page reloads
* docs: changelog for compaction markers in chats and hiding controls for users
The view_file and view_knowledge_file built-in tools let a model ask for a start and end line, but passing only the end line did nothing: the tool returned the first block of characters and dropped the end line without a word. Models that asked for "up to line N" kept getting the same opening slice and no error, so they retried the call over and over. An end line on its own now reads from line 1 up to that line, the same result as passing start line 1 explicitly.
Fixes#33100
When a model reads a file or knowledge-base file in pages with line numbers turned on, the position it was told to continue from pointed past the end of the previous page, so a chunk of text between pages was silently skipped and never returned. The gap grew with every page, dropping a few percent of a large file. Reading now continues from the right place and returns every character exactly once; paging with line numbers off is unchanged.
Fixes#33099