Skip to content

Releases: emcie-co/parlant

Version 1.6.7

23 Feb 17:27
Compare
Choose a tag to compare

[1.6.7] - 2025-02-23

Fixed

  • Fix major issue with tool-call inference of non-string parameters on OpenAI

Version 1.6.6

23 Feb 16:19
Compare
Choose a tag to compare

[1.6.6] - 2025-02-23

Fixed

  • Fix unicode generation issues in gemini
  • Adapt tool calling with optional parameters

Version 1.6.5

20 Feb 12:32
Compare
Choose a tag to compare

[1.6.5] - 2025-02-20

Fixed

  • Improve guideline proposer generation consistency

Version 1.6.4

20 Feb 10:27
Compare
Choose a tag to compare

[1.6.4] - 2025-02-20

Fixed

  • Upgrade to Gemini 2.0 Flash

Version 1.6.3

18 Feb 15:09
Compare
Choose a tag to compare

[1.6.3] - 2025-02-18

Fixed

  • Fix Cerebras generation (as well as other Llama 3 generations)

Version 1.6.2

29 Jan 16:38
Compare
Choose a tag to compare

[1.6.2] - 2025-01-29

Fixed

  • Fix loading DeepSeek service during server boot

Version 1.6.1

20 Jan 20:15
Compare
Choose a tag to compare

[1.6.1] - 2025-01-20

Fixed

  • Fix ToolCaller not getting clear information on a parameter being optional
  • Ensure ToolCaller only calls a tool if all required args were given
  • Improve valid JSON generation likelihood in MessageEventGenerator
  • Improve ToolCaller's ability to correctly run multiple tools at once

Version 1.6.0

19 Jan 17:47
Compare
Choose a tag to compare

[1.6.0] - 2025-01-19

Added

  • Add shot creation helper functions under Shot
  • Add ContextEvaluation in MessageEventGenerator
  • Add a log command under client CLI for streaming logs
  • Add engine lifecycle hooks

Changed

  • Split vendor dependencies to extra packages to avoid reduce installation time
  • Modified ToolCaller shot schema
  • Disable coherence and connection checking by default in the CLI for now

Fixed

  • Improved GuidelineProposer's ability to handle compound actions
  • Improved GuidelineProposer's ability to distinguish between a fulfilled and unfulfilled action
  • Improved GuidelineProposer's ability to detect a previously applied guideline's application to new information
  • Reduced likelihood of agent offering hallucinated services
  • Fix ToolCaller false-negative argument validation from int to float
  • Fix ToolCaller accuracy
  • Fix ToolCaller making up argument values when it doesn't have them
  • Fix some cases where the ToolCaller also calls a less-fitting tool
  • Fix mistake in coherence checker few shots
  • Fix markdown tables in sandbox UI
  • Fix wrong import of RateLimitError
  • Fix PluginServer validation for optional tool arguments when they're passed None
  • Fix utterances sometimes not producing a message

Version 1.5.1

05 Jan 08:22
Compare
Choose a tag to compare

[1.5.1] - 2025-01-05

Fixed

  • Fix server CLI boot

Version 1.5.0

04 Jan 08:03
Compare
Choose a tag to compare

[1.5.0] - 2025-01-04

Added

  • Add DeepSeek provider support (via DeepSeekService)

Changed

  • Change default home dir from runtime-data to parlant-data

Fixed

  • Fix tool-calling test
  • Fix HuggingFace model loading issues