3.15.1 (latest)#

Squirro 3.15.1 was released on January 26, 2026.

Learn more about the Squirro Release Process.

Note for administrators

This release introduces important changes that may require adjustments to your existing setup. Learn more

What’s New#

  • Technical preview - OpenSearch, available as an alternative to Elasticsearch. That initial release is intended solely for testing and evaluation, and must not be used in production environments.

  • Technical preview - OpenAI Code Interpreter, for secure code execution in a sandboxed environment. That initial release is intended solely for testing and evaluation, and must not be used in production environments.

Improvements#

AI#

  • Skip the file upload tool when no files are attached to the conversation.

  • Refined retriever tool description to improve the precision and relevance of agent searches.

  • Include document metadata in event stream snippets for streaming chat responses.

  • New default large language model set to GPT-5.2 for enhanced performance, accuracy, and responsiveness.

  • Serialize RequestTimeout as Tuple as expected by the genai service API.

  • langchain-classic package for backward compatibility with legacy chains.

  • Error handling for invalid file objects with clear error messages.

  • Warning triggered when tools are defined but ignored in passthrough mode.

  • agent_id added to prompt schema.

  • waitForImage added to the DynamicImage component for automatic retry when an image fails to load.

  • Support for provider-specific extra arguments allowing advanced Bedrock configurations.

  • AWS credential configuration supporting multiple authentication methods (explicit credentials, AWS profiles, session tokens, and AWS credential chain).

  • Dual-mode chat history storage with database backend support.

  • Intelligent container caching with file deduplication and LRU eviction.

  • Lifecycle management added for artifacts (delete by project/conversation scope).

  • Database-backed conversation storage for GenAI service chat history.

  • safety_identifier introduced for OpenAI API calls to enable user-level tracking and abuse prevention.

  • Automatically generate concise titles (5-7 words) for new conversations, based on the first user question.

  • Conversation import/export endpoints for data migration.

  • Standalone migration script tools/devops/genai_migrate_redis_to_db.py.

  • AWS Bedrock support with two provider types, bedrock_converse using Converse API and bedrock_chat using InvokeModel API.

  • Audit logging for conversation operations using structlog.

  • Storage bucket configuration for GenAI artifacts with automatic setup.

  • Artifact persistence with multiple storage backends.

  • Artifact reference list in responses with download links and file sizes.

  • preserve_filename configuration flag for opt-in filename preservation in file uploads.

  • /projects/<project_id>/arbitrary_prompt endpoint added to GenAI proxy plugin for direct LLM prompt execution.

  • Async-first architecture with proper error handling and retry logic.

  • Converter module for translating between Redis and DB formats.

  • Added field validator to automatically truncate conversation titles to 512 characters.

Platform#

  • Explicitly set a none string for a more visible project setting UI for the reasoning_effort parameter.

  • Layer visibility builder in the dashboard editor migrated to React.

  • Keep the split view open state while navigating between items in Item Detail.

  • New projects now include the Semantic Search pipeline workflow by default. The Default Dashboard has been replaced by the Explore Dashboard, which correctly displays project items and provides an updated interface.

  • Adjust the “Server Admin” terminology to “Workspace Admin” in Project Settings for instances with the Workspaces feature activated.

  • Auto-refresh functionality for the data sources page. Administrators can deactivate it via the frontend.userapp.show-sources-auto-refresh server configuration key.

  • Add a warning when the preserve_filename flag is set for the newsletter bucket.

  • XSS sanitization for the email editor.

Bug Fixes#

  • Dashboard query is highlighted in the item cards of the items widget.

  • User cannot select a facet with a numerical value from the item card.

  • Fix long loading in “from chunks” mode.

  • Ensure proper closing of scan context.

  • Editing the name of the collection, favorite, or search-results agent types does not update the agents dropdown.

  • llm_timeout not being parsed properly from project settings.

  • Reference search, document search and similar items query include dashboard query.

  • Missing tool calls in conversation history for agentic workflows with multiple iterations.

  • Paragraph citation in Chat not highlighted in the old reference format.

  • Timeline widget not showing header.

  • Date and time not displayed in the client timezone.

  • Fixed the arbitrary_prompt endpoint by adding LLM configuration from project settings.

  • Fixed quotes handling in the unanalyzed field search.

  • Pass the itemsCollectionsCollection property to the ItemDetail component in the Chat widget, for the collections context.

  • Lost text fragments in Dynamic UI components leading to incorrect highlighting of referenced text.

  • When semantic search results contain a small amount of items, the frontend fetches the first 2 pages and shows loading indefinitely.

  • Inline citations inside dynamic UI not opened in the item detail.

  • Dashboard time selection not included in the chat context.

  • Passthrough mode condition incorrectly handles the CHAT_WITH_LLM agent type.

  • Query not clearing when changing dashboard programmatically via external navigation, preventing a clean dashboard state transition.

Miscellaneous#

  • Upgraded LangChain libraries for improved performance and security.

Known Issues#

  • In specific scenarios, cross-lingual information retrieval for semantic search may deliver lower-than-expected quality (SQ-28171).

Breaking Changes#

  • Email templates that make external requests to populate the template using the h.requests_get utility are no longer supported.

  • extended_plugin_options was removed from the DataSource class.

  • LangChain v1 removes support for classic chains like LLMChain and Chain base classes. Custom code using these deprecated classes must update imports from langchain.* to langchain_classic.*. See the langchain-classic migration guide.

Installation and Upgrade#

For new installations, find step-by-step instructions on the Install and Manage Squirro with Ansible page (recommended) and Installing Squirro on Linux pages.

To upgrade an existing installation, see the Upgrading Squirro page.