- Introduced `isRefreshDisabled` prop to `ModelsSelectionPanel` to control refresh button state.
- Updated `ProviderConnectDialog` to conditionally enable refresh functionality based on model support and submission status.
The chat composers submit on Enter without checking whether an IME
composition is active. For Japanese/Chinese/Korean input, pressing Enter
to confirm a conversion fires the submit handler and sends a half-typed
message. Guard the Enter handlers with e.nativeEvent.isComposing in the
main chat editor, the anonymous free composer, and the mention/prompt
picker key navigation.
- Increased ACCESS_TOKEN_LIFETIME_SECONDS from 30 minutes to 60 minutes for improved session duration.
- Introduced getRequestHeadersWithCurrentDesktopAuth function to streamline authorization header management across fetch requests in the chat page.
- Introduced fetchSession function to streamline session fetching logic.
- Updated useSession to handle 401 errors by refreshing the session when necessary.
- Modified getDesktopAccessToken to accept options for forced token refresh, improving desktop authentication flow.
- Refactored fetchZeroContext to include a buildHeaders function for better header management.
- Added a request function to handle 401 errors and refresh sessions as needed.
- Improved overall session management for desktop authentication.
- Implemented session refresh logic in the fetchWithTurnCancellingRetry function to handle 401 errors more gracefully.
- Added a new import for refreshSession utility to facilitate session renewal.
- Changed default AUTH_TYPE in backend configuration to "LOCAL".
- Updated frontend environment configuration to reflect the new default for packaged clients.
- Adjusted runtime authentication resolution to use "LOCAL" as the fallback value.
- Added logging for JSON parsing failures in structured output to aid in debugging.
- Updated the TranscriptTurn model to ignore unknown keys, improving flexibility for future model variations.
- Updated main-agent middleware to clarify that both filesystem reads/writes and knowledge-base retrieval are handled by the `knowledge_base` subagent.
- Introduced `_forward_mention_pins` function to carry `@`-mention pins into subagent state.
- Revised system prompts to reflect the new retrieval method and ensure proper citation handling.
- Removed the `search_knowledge_base` tool and its related tests, consolidating functionality under the `task` tool.
- Enhanced documentation to guide usage of the new retrieval approach and citation practices.
The document saving logic used a hardcoded character slice ([:4000]) for
the document summary content fed to the embedder. For UTF-8 documents
(e.g., Arabic, Chinese, Japanese), characters above U+007F take multiple
bytes but count as 1 character in a slice, potentially producing text
that exceeds the token limit of the embedding model.
Replaced the arbitrary slice with runcate_for_embedding(), which safely
bounds the text using the embedding model's actual tokenizer.
When a codeBlock is nested inside an indented structure like a
bulletListItem, the \_render_block\ function prepends the block's
indentation \prefix\ to every line.
However, for codeBlock elements, only the fence markers (the opening
and closing \\\) should carry the block indentation. Interior code
lines must not have the prefix prepended, because markdown parsers
treat leading spaces inside a code fence as part of the code content.
This fix removes the prefix from interior code lines to prevent
code snippets stored in notes from gaining spurious whitespace.
When the Unstructured API returns a Table element without text_as_html
in its metadata (e.g. local install or free-tier API), the lambda was
raising KeyError: 'text_as_html', crashing the entire document
indexing pipeline for any file containing tables.
Guard the key access with .get() and fall back to the plain extracted
text content (x) so the pipeline continues and the table content is
still indexed, just without HTML formatting.
