v0.4.2

What's Changed

• CAMP-1941-ba-ai-langkit-hotfix-adjust-prompt-field-main by @patsluth-tsl in #18

Full Changelog: v0.4.1...v0.4.2

v0.2.2post1

Full Changelog: v0.2.2...v0.2.2post1

v0.4.1

What's Changed

• hotfix/upgrade-dependencies by @patsluth-tsl in #16

Full Changelog: v0.4.0...v0.4.1

v0.4.0

Added

• Support for custom slash commands
• Support for interactive endpoint

v0.3.0

Refactored

• Removed dependency on Vector Stores for tools functionality. Projects using tools should be reviewed for compatibility.

Added

• Enhanced Slack AI Chat Messages admin interface with message text display and DM/Channel filtering
• Migrated runners, tools, and vector stores to use Unfold's ModelAdmin helper for improved UI/UX

Fixed

• Corrected application name in setup.cfg configuration

v0.2.6

Added

• Slack Reactions admin.
• Export CSV of reactions actions.
• Improving the slack output formatting to create the chunks based on break lines.

Fixed

• Orchestrator workflow: making the synthesizer node clear the memory after erros.
• Slack large responses: chunks are now connected by the previous chunk ts.

v0.2.5

Added

• SlackAIChatController.process_response more modular for customizations.
• Admin model classes using either Django Admin Model or Unfold Admin Model.
• Model SlackAIChatMessage OneToOne field converted to FK, so we can duplicate threads.

Fixed

• SlackAIChatController.collect_slack_context fixed to collect the proper current user data.

v0.2.4

Added

• Keyword arguments added to the slack_process_incoming_user_slack_message task and some other methods. This allows to pass additional context, like a placeholder message that is to be overwritten with the final bot answer.
• process_message returns the exception if there was any.

Fixed

• Private channel identifier in tests changed to "DXXXXXXXXXM"

Full Changelog: v0.2.3...v0.2.4

v0.2.3

Added

• Now workflow nodes (workers and agents) can define a specific LLM model for themselves.
• Compatibility with version 0.3 of langchain-openai, which introduced native support for formatted_output with Pydantic. This eliminates the need for auxiliary properties like min_length and default.

Fixed

• Synthesizer prompt conditional rule adjustment to also check empty lists of selected nodes.

v0.2.2

• Bugfix: Slack App now it optional