v0.9.15

Fixed

• 🗂️ Expanded workspace chat lists keep their place. When a chat updates, Computer now refreshes the expanded workspace list without shrinking it back to only the newest few chats.

v0.9.14

Changed

• 🔧 Chat handling got a cleanup. Computer now carries conversations forward more cleanly after some internal refactoring.

v0.9.13

Fixed

• 💬 Local model chats stop repeating themselves. When a llama.cpp or OpenAI-compatible connection finishes an answer, Computer now stops there instead of accepting extra repeated parts that could duplicate the final text or run the same tool again.

v0.9.12

Added

• 📱 Unsaved file edits now get a proper warning. Closing an edited file tab now shows Computer's own confirmation dialog, so mobile users and desktop users can choose to keep editing or close without saving.
• 🎛️ Browser streaming now has an encoder setting. Admins can leave streaming on Auto, force hardware encoding, or force software encoding from Admin > Web > Streaming quality > Advanced.

Changed

• 🌍 More of Computer is translated. Admin screens, chat controls, diff settings, file editor labels, skill controls, and planning prompts now use the same language as the rest of the app across all supported languages.
• 📜 The license is clearer for hosted use. The old multi-tenant hosting restriction was removed while the attribution requirements remain in place.

Fixed

• 🌐 Browser tabs work on servers without video hardware. Streaming now defaults to Auto, so headless servers, virtual machines, and other GPU-light hosts can fall back to software encoding instead of failing to open the browser stream.
• 🤖 Claude Code appears when it comes from the Claude desktop app. If Claude Code was installed through Claude desktop on macOS or Windows, Computer can now find it automatically and show it as an available coding agent.

v0.9.11

Added

• 🤖 Gemini is now its own coding agent. Choose Gemini as an agent and Computer will start the Gemini CLI with its Google sign-in flow, keep the chat resumable, and handle image attachments like the other coding agents.

Fixed

• 📱 Signal group chats reply to the right group. When a Signal group message reaches Computer, replies, typing indicators, and attachments now use the group address Signal expects instead of falling back to the sender or losing the group target.

v0.9.10

Added

• 🎚️ You can make borders easier to see. Appearance settings now include a border contrast control, so people who found the softer 0.9.9 borders hard to follow can make workspace sections stand out again without changing themes.

Changed

• 🤖 Coding-agent chats continue more naturally. Codex, Claude Code, Cursor, Cline, Grok, and OpenCode now send your latest request into the existing agent chat instead of replaying the whole conversation back at the agent.

Fixed

• 💬 Open WebUI messages save even when the model arrives as a list. Computer now accepts Open WebUI's model format and keeps the chat turn instead of dropping it during save.
• 📁 Dropping a file onto an open folder uploads it once. File drops on expanded folders now go only to that folder, not both the folder and the current directory.
• 🎚️ The border contrast slider is easier to use. The slider track and handle are more visible in light and dark mode.
• 📦 Fresh installs avoid a known package mismatch. Computer now stays within the supported cryptography release range instead of picking a newer release that can break installs.

v0.9.9

Changed

• 💬 Unread chats stay where you can see them. Chat lists now keep unread finished chats above already-read chats, and they refresh when a chat is marked read.
• 🧭 The app feels a little lighter. Dividers and borders across the interface are softer, so the workspace chrome gets out of the way more.

Fixed

• ⏰ Follow-up timers behave more cleanly. Timer prompts and subagent handoffs stay internal, pending follow-ups stay hidden until they are ready, and old timer state is still understood.
• 🔎 Internal work stays out of normal chat search. Recent-chat search no longer shows hidden timer or subagent chats as if they were regular conversations.
• 🧠 The chat size meter keeps moving for local models. llama.cpp chats now show live size updates even when the server does not send token totals until the end.
• 🔐 File reading avoids credential files. Computer can read normal absolute paths and ~/ paths, but it blocks common credential files, .envrc, and special device files.
• 📱 Signal bot updates no longer disappear. When a Signal chat needs an edited answer, Computer now sends the updated text as a new message instead of silently doing nothing.

v0.9.8

Added

• 🏠 Choose how .cptr is handled in projects. Admins can now turn the automatic .gitignore entry on or off, so local chats, memory, caches, audio samples, and generated files stay private by default unless you choose otherwise.
• 🤖 A llama.cpp connection mode. OpenAI-compatible connections can now be marked as llama.cpp, so Qwen and other local reasoning models can receive prior thinking in the format llama.cpp expects.
• 🧩 Skills can open their own companion files. view_skill can now load files such as references/guide.md from the selected skill folder, without opening broader home-directory access.
• 🌐 Sized browser screenshots. Browser tasks can request a screenshot at a specific size and Computer will restore the normal page view afterward.

Changed

• ⚙️ Admin switches behave like switches. In Models, Agents, Connections, Messaging, and Tool Servers, click the row text to edit and click the switch only to turn something on or off.
• 🔎 Workspace search respects .gitignore. File search and content search now skip ignored folders and files, keeping generated output, caches, and local Computer data out of normal results.
• 🧠 Reasoning replay is stricter by default. Computer now keeps strict OpenAI-compatible connections clean, and only sends llama.cpp-style reasoning_content when the connection is set to llama.cpp.

Fixed

• 🧠 The chat size meter no longer freezes behind llama.cpp. Computer now normalizes token usage from OpenAI-compatible streams and sends live context updates to the chat while a response is still running.
• 💬 Compacted chats survive regenerated answers better. Summaries are saved on a stable kept user message when possible, so regenerating an assistant answer is less likely to lose the compacted history.
• ✂️ Manual compact keeps the latest turn available. Compacting a chat now summarizes the older branch and leaves the newest message ready for the next response.
• 🧠 Prior reasoning is preserved more reliably. Reasoning from ordinary replies, tool-call turns, Chat Completions streams, and Responses streams is less likely to be dropped before the next model request.
• 🎨 Token usage details match the chosen theme. The usage popover now uses Computer's app surface colors instead of the wrong gray background.
• ✨ Messages render more faithfully. Your own messages can show Markdown, and raw markup appears as text instead of disappearing.
• 🗂️ Commit suggestions are less brittle. If the model returns a plain commit message instead of perfect structured output, Computer can still use the useful parts.
• 📝 File edits keep text encoding consistent. Created and edited files are written as UTF-8, which avoids surprises with non-ASCII project files.

v0.9.7

Fixed

• 💬 Chat controls match your appearance. The lower edge of chats now blends in with the look you chose for Computer.

v0.9.6

Added

• ⏰ Follow-up timers for chats. Computer can now return to a chat at the right time to continue work, check again, or follow up.

Changed

• ✨ Clearer change reviews. Changes in files are easier to spot, including larger edits.