Major Features

Image Generation

Now you can generate images via /img when a supported provider is set up.

This flow also allows image attachments for edits, followups, or combination prompts if your provider supports image editing.

Ability for an agent-tool to generate images is coming next update.

Better File Picker UX

Now in the file picker you can drag entire folders and their hierarchy will be preserved in the upload panel. This will only import the top level files in the folder — it will be deep and recursive in a later update.

You can also now drag and drop a file into an existing folder directly without needing to move it.

Lastly, you no longer need to type https in the URL fetcher - it will just assume https when left out.

We also made some large performance improvements, like lazy-loading of folders, for those with thousands of documents in a folder.

Tool Toggle Mid-Session

Now, during agentic chats you can freely toggle tools on/off at will without needing to reload or start a new session. Only tools that have been configured (if configuration is required) are available.

Stop Generation

A large improvement is now that aborting a response actually fully kills the inference as well so that ghost inference does not continue when no longer wanted mid-stream.

New Features

• New create-scheduled-job agent tool - agents can now create scheduled jobs on your behalf, with a dedicated card in the Assistant Panel
• Image generation via the /img command, including rendering and controls in the Assistant Panel
• Foundry Local overhaul - model table with full model management, agent and async chat, and tool calling support
• PiperTTS overhaul - long messages now stream as they are spoken with no timeouts, and the Assistant Panel can speak responses with markdown formatting stripped
• You.com added as a web-search provider for agents
• Gitea data connector
• GitHub Enterprise URLs are now supported in the GitHub data connector
• OpenAI-compatible Whisper API support for speech-to-text (bring your own transcription endpoint)
• input_audio support for audio attachments on the Generic OpenAI provider
• Folder upload in the Document Manager now preserves relative folder paths
• Current date and time is now included in default system prompts so models can answer time-aware questions
• SQL agent connector now supports adding multiple databases at once and includes an SSL toggle for self-signed certificates

Improvements

• Stop generation button now fully aborts agent sessions and inference, including the Assistant Panel and local inference engines
• Agent chats with no tools attached now route through native tooled completions for better response quality
• Model router cooldown now starts when inference completes instead of when the routing rule is detected, so long reasoning responses no longer consume the entire cooldown window
• Smoother chat auto-scrolling during streaming - no more scroll lock or jitter on long messages
• Updated default fallback models for OpenAI and Anthropic
• MCP tools with nested object parameters ($ref/$defs schemas) are now inlined correctly for Anthropic agent tool calls
• Light mode scrollbar colors
• macOS system overlays are now filtered out of screen context capture
• New and updated translations: Catalan, Italian, Indonesian, Lao, Croatian, and Traditional Chinese (zh-TW)
• PiperTTS now chunk-plays content, meaning significantly faster playback times for longer messages

Bug Fixes

• Fixed PiperTTS bug preventing voices from playing and causing a "Worker Timeout" error
• Fixed a native onnxruntime crash during embedding caused by maxConcurrentChunks when using an external embedding provider
• Anthropic responses are now properly streamed for non-streaming REST API workspace chat endpoints
• Fixed workspace model selection for AWS Bedrock
• Fixed Lemonade URL input not saving on change
• Thinking blocks no longer render for models that do not emit them on Generic OpenAI providers
• Fixed Windows filepaths in the Document Manager
• Fixed slash command presets not expanding into prompt text
• Fixed citations regression in the API
• Fixed Generic OpenAI provider requests for Magic Echo
• Watched document re-syncs no longer break pinned document de-duplication
• Agent clarifying-question counter now resets per turn
• Fixed null and incorrect vector counts for Milvus/Zilliz when multiple workspaces exist
• Workspace update API now validates payloads instead of returning a misleading 200 response
• Thread lookups are now scoped to the validated workspace
• Kokoro TTS voice lists now render for pre-0.3.x kokoro-fastapi servers
• Empty "App Integrations" section is now hidden on the Agent Skills page in multi-user mode
• Temperature is now omitted for claude-sonnet-5, which does not support it

Pinned Download Links

Revision 1.16.0: