0.29.0

• fixes #163
• fixes #156
• fixes #157 - removed extra padding, seemed unnecessary - configurable is still on the table if feedback demands it
• fixes #151
• fixes #152
• fixes #162
• fixes #168 - intro text should be changed through the world editor
• fixes #171 - added button to request choices, and whatever character is active will be target of the choice creation
• fix bug where canceling a generation would auto cancel the next attempt as well
• scene analysis to help guide conversation and narration (summarizer agent)
• context investigation rework (summarizer agent)
• narrator improvements (prompts revisited, context polished)
• narrator long term memory options exposed (similar to conversation agent)
• narrator can now use a writing style template (set through the scene settings and a on/off toggle in the narrator agent)
• director can now guide actors based on scene analysis
• director can now guide narrator actions based on scene analysis
• director long term memory options exposed (similar to conversation agent)
• summarization with analysis to improve quality of layered history summaries (toggleable)
• summarizer long term memory options exposed (similar to conversation agent)
• long term memory caching
• adds favicon
• favicon will change while talemate is busy
• improvements to context investigation
• implement backend to separate function orchestration from creative writing
• add suggestions tab to world editor - agents can now propose updates to character attributes and description based on story progression.
• add DeepSeek API client
• LLM system prompts can now be overwritten in the global app config as well as specifically within each client config.

0.28.0

• Function to fork copy of scene from specific message
• Dynamic actions: director can now generate clickable choices
• Inline action shortcut: starting a message with @ will trigger generation of dialogue or narration based on the instructions after
  • e.g., @look at the treasure map
• Layered history: The summarization agent can now maintain a layered history, always keeping the entire progress in the context by continually progressing a layered history into more condensed entries. This is toggleable in the agent config.
• improvements to context history
• improvements to RAG prompts
• conversation agent can now use the summarization agent for context investigations using the new layered history. (experimental)
• fixes #147 by adding configuration options for message styles and colors.
• fixes #146 by adding shared actor instructions to the conversation agent, which are inserted towards the end of the context history
• saving a new scenario should now always create a folder name based on the scenario title
• fixes #153 - fix world editor rubberbanding issues by refactoring when changes to inputs are saved.
• fixes #155 where inactive characters could not be deleted
• fixes issues with historic date calculations
• fixes #158 - tab should now cycle through all active ai characters, not just the first.
• fixes #161 - simulation suite issues from recent changes
• fixes issue with automatic character selection during multi-character dialog if client does NOT support coercion
• fixes simulation suite error when attempting to alter existing NPC
• openai client: add missing 4o models

0.27.0

fixes #134 - memory agent configuration improvements
fixes #136 - allow cancellation of ongoing agent actions
fixes #138 - rubberbanding issue when editing text in certain fields in the world editor
fixes #139 - narrator generation options added
fixes #140 - unified process
fixes #142 - A1111 visualizer sampler options added
fixes #143 - visualizer configurable timeout

Untracked Changes

• add toggle to hide / show disabled clients
• fix issue simulation suite breaking when removing or altering character
• fix pytorch install issues
• narrator prompts revised and post generation cleanup improvements
• docker fixes
• openai: add o1 models

0.26.0

This is a big refactor of the world editor and generative creation of characters and scenes.

⚠️ This release substantially changes how state reinforcement templates are handled, while i have made sure that old custom made templates are copied to the new system i'd still advise doing a new separate install of talemate if you have made lots of custom templates to track character states.

fixes #86
fixes #87
fixes #98
fixes #99
fixes #117
fixes #120
fixes #122
fixes #123
fixes #130
fixes #131

Untracked changes / fixes

• allow disabling / re-enabling of clients
• some preset default tweaks (mostly to make scene direction more stable)
• refactored the scoped api to direct more complex scenes
• openai: added gpt4-o and gpt-4o-mini
• anthropic: added claude 3.5 sonnet
• mitral: added mistral-nemo
• added phi3 base template
• autocomplete fixes
• fixes an issue where "Save As.." could cause memory entries to be lost

0.25.6

fixes #110 - Exposed presets to config and UX for editing. Available through Settings -> Presets
fixes #125 - revisit inference parameter defaults and client implemenations

Untracked changes

• Dedicated TabbyAPI client to use more samplers (contributed by @official-elinas, thanks!)
• Reorganized presets into more sane naming based on agent action
• Some refactoring of presets.py (contributed by @official-elinas, thanks!)
• OpenAI client: gpt-4o as default
• Fix issue where audio stop button was visible even though no audio was playing

0.25.5

fixes #124

Untracked Changes / Fixes

• fixes openai compatibility client to use completions endpoint if API Handles prompt template is disabled and chat/completions if its enabled.
• fixes openai compatibility client not passing through presence penalty
• fixes openai compatibility client not saving coercion settings
• fixes issue when setting LLM prompt template if model name contained /
• lock sentence-transformers to 2.2.2 for now since >=2.3.0 seems to break loading of instructor embeddings
• support png tEXt import for character cards (fixes silly tavern card import)
• fixes koboldcpp automatic1111 setup spam
• fixes koboldcpp openai api not passing through presence penalty
• changed mistral.ai default model to 8x22b
• fixed mistral.ai 7b and 8x7b json respones (world state generation failures)

0.25.4

fixes #114 - potential fix for missing vue-cli-service error some people have been experiencing
fixes #119 - ALLOWED_HOSTS var cannot be set error

0.25.3

fixes #112 - env var config for frontend serve
fixes #111 - openai gpt-4o added
fixes #109 - koboldcpp bug fixes / enhancements

• fix rep_pen and max_tokens errors
• auto configure visual agent to kcpp automatic1111 if available (Automatic Setup option needs to be turned on in agent)
• use tokencount endpoint for accurate token counts
• default port changed to 5001

Untracked changes

• fixes issues where saving a new scene could cause recent config changes to revert

0.25.2

fixes #106
fixes #105
fixes #107

Untracked changes

• fixes vue warnings from typo

0.25.1

• remove auto disabling of clients