* move memory agent to directory structure
* chromadb settings rework
* memory agent improvements
embedding presets
support switching embeddings without restart
support custom sentence transformer embeddings
* toggle to hide / show disabled clients
* add memory debug tools
* chromadb no longer needs its dedicated config entry
* add missing emits
* fix initial value
* hidden disabled clients no longer cause enumeration issues with client actions
* improve memory agent error handling and hot reloading
* more memory agent error handling
* DEBUG_MEMORY_REQUESTS off
* relock
* sim suite: fix issue with removing or changing characters
* relock
* fix issue where actor dialogue editor would break with multiple characters in the scene
* remove cruft
* implement interrupt function
* margin adjustments
* fix rubber banding issue in world editor when editing certain text fields
* status notification when re-importing vectorb due to embeddings change
* properly open new client context on agent actions
* move jiggle apply to the end of prompt tune stack
* narrator agent length limit and jiggle settings added - also improve post generation cleanup
* progress story prompt improvements
* narrator prompt and cleanup tweaks
* prompt tweak
* revert
* autocomplete dialogue improvements
* Unified process (#141)
* progress to unified process
* --dev arg
* use gunicorn to serve built frontend
* gunicorn config adjustments
* remove dist from gitignore
* revert
* uvicorn instead
* save decode
* graceful shutdown
* refactor unified process
* clean up frontend log messages
* more logging fixes
* 0.27.0
* startup message
* clean up scripts a bit
* fixes to update.bat
* fixes to install.bat
* sim suite supports generation cancellation
* debug
* simplify narrator prompts
* prompt tweaks
* unified docker file
* update docker compose config for unified docker file
* cruft
* fix startup in linux docker
* download punkt so its available
* prompt tweaks
* fix bug when editing scene outline would wipe message history
* add o1 models
* add sampler, scheduler and cfg config to a1111 visualizer
* update installation docs
* visualizer configurable timeout
* memory agent docs
* docs
* relock
* relock
* fix issue where changing embeddings on immutable scene would hang
* remove debug message
* take torch install out of poetry since conditionals don't work.
* torch gets installed through some dependency so put it back into poetry, but reinstall with cuda if cuda support exists
* fix install syntax
* no need for torchvision
* torch cuda install added to linux install script
* add torch cuda install to update.bat
* docs
* docs
* relock
* fix install.sh
* handle torch+cuda install in docker
* docs
* typo
* TabbyAPI Client Addition and presets refactoring (#126)
* feat: frequency_penalty (will make tabbyAPI custom wrapper)
* feat: add FREQUENCY_PENALTY_BASE and adj. conversation template
* feat: use `client_type` of `openai_compat` to send FIXED preset
* change from client name
* feat: pass client_type into presets.configure(...)
* wip: base TabbyAPI client
* feat: add import to register TabbyAPI client
* feat: adjust `presence_penalty` so it has a range of 0.1-0.5 (higher values will likely degrade performance)
* feat: add additional samplers/settings for TabbyAPI
* feat: keep presence_penalty in a range of 0.1-0.5
* feat: keep min_p in a range of 0.05 to 0.15
* update tabbyapi.py
* feat: add MIN_P_BASE and TEMP_LAST and change to tabbyapi client only for now
* fix: add /v1 as default API route to TabbyAPI
* feat: implement CustomAPIClient to allow all TabbyAPI parameters
* fix: change to "temperature_last" instead of "temp_last"
* feat: convert presets to dictionary mappings to make cleaner/more flexible
* fix: account for original substring/in statements and remove TabbyAPI client call
* fix: move down returning token values as it realistically should never be none, so substrings wouldn't be checked
* chore: remove automatic 'token' import due to IDE
---------
Co-authored-by: vegu-ai-tools <152010387+vegu-ai-tools@users.noreply.github.com>
* tabbyapi client auto-set model name
tabbyapi client use urljoin to prevent errors when user adds trailing slash
* expose presets to config and ux for editing
* some more help text
* tweak min, max and step size for some of the inference parameter sliders
* min_p step size to 0.01
* preset editor - allow reset to defaults
* fix preset reset
* dont perist inference_defaults to config file
* only persist presets to config if they have been changed
* ensure defaults are loaded
* rename config to parameters for more clarity
* update default inference params
textgenwebui support for min_p, frequence_penalty and presence_penalty
* overridable function to clean promp params
* add `supported_parameters` class property to clients and revisit all of the clients to add any missing supported parameters
* ux tweaks
* support_parameters moved to propert function
* top p decrease step size
* only show audio stop button if there is actually audio playing
* relock
* allow setting presence and frequency penalty to 0
* lower default frequency penalty
* frequency and presence penalty step size to 0.01
* set default model to gpt-4o
---------
Co-authored-by: official-elinas <57051565+official-elinas@users.noreply.github.com>
* flip title and name in recent scenes
* fix issue where a message could not be regenerated after applying continuity error fixes
* prompt tweaks
* allow json parameters for commands
* autocomplete improvements
* dialogue cleanup fixes
* fix issue with narrate after dialogue and llama3 (and other models that don't have a line break after the user prompt in their prompt template.
* expose ability to auto generate dialogue instructions to wsm character ux
* use b64_json response type
* move tag checks up so they match first
* fix typo
* prompt tweak
* api key support
* prompt tweaks
* editable parameters in prompt debugger / tester
* allow reseting of prompt params
* codemirror for prompt editor
* prompt tweaks
* more prompt debug tool tweaks
* some extra control for `context_history`
* new analytical preset (testing)
* add `join` and `llm_can_be_coerced` to jinja env
* support factual list summaries
* prompt tweaks to continuity check and fix
* new summarization method `facts` exposed to ux
* clamp mistral ai temperature according to their new requirements
* prompt tweaks
* better parsing of fixed dialogue response
* prompt tweaks
* fix intermittent empty meta issue
* history regen status progression and small ux tweaks
* summary entries should always be condensed
* google gemini support
* relock to install google-cloud-aiplatform for vertex ai inference
* fix instruction link
* better error handling of google safety validation and allow disabling of safety validation
* docs
* clarify credentials path requirements
* tweak error line identification
* handle quota limit error
* autocomplete ux wired to assistant plugin instead of command
* autocomplete narrative editing and fixes to autocomplete during dialog edit
* main input autocomplete tweaks
* allow new lines in main input
* 0.25.0 and relock
* fix issue with autocomplete elsewhere locking out main input
* better way to determine remote service
* prompt tweak
* fix rubberbanding issue when editing character attributes
* add open mistral 8x22
* fix continuity error check summary inclusion of target entry
* docs
* default context length to 8192
* linting
* linting
* improve prompt devtools: test changes, show more information
* some more polish for the new promp devtools
* up default conversation gen length to 128
* openai client tweaks, talemate sets max_tokens on gpt-3.5 generations
* support new openai embeddings (and default to text-embedding-3-small)
* ux polish for character sheet and character state ux
* actor instructions
* experiment using # for context / instructions
* fix bug where regenerating history would mess up time stamps
* remove trailing ]
* prevent client ctx from being unset
* fix issue where sometimes you'd need to delete a client twice for it to disappear
* upgrade dependencies
* set 0.19.0
* fix performance degradation caused by circular loading animation
* remove coqui studio support
* fix issue when switching from unsaved creative mode to loading a scene
* third party client / agent support
* edit dialogue examples through character / actor editor
* remove "edit dialogue" action from editor - replaced by character actor instructions
* different icon for delete
* prompt adjustment for acting instructions
* adhoc context generation for character attributes and details
* add adhoc generation for character description
* contextual generation tweaks
* contextual generation for dialogue examples
fix some formatting issues
* contextual generation for world entries
* prepopulate initial recen scenarios with demo scenes
add experimental holodeck scenario
* scene info
scene experimental
* assortment of fixes for holodeck improvements
* more holodeck fixes
* refactor holodeck instructions
* rename holodeck to simulation suite
* better scene status messages
* add new gpt-3.5-turbo model, better json response coercion for older models
* allow exclusion of characters when persisting based on world state
* better error handling of world state response
* better error handling of world state response
* more simulation suite fixes
* progress color
* world state character name mapping support
* if neither quote nor asterisk is in message default to quotes
* fix rerun of new paraphrase op
* sim suite ping that ensure's characters are not aware of sim
* fixes for better character name assessment
simulation suite can now give the player character a proper name
* fix bug with new status notifications
* sim suite adjustments and fixes and tuning
* sim suite tweaks
* impl scene restore from file
* prompting tweaks for reinforcement messages and acting instructions
* more tweaks
* dialogue prompt tweaks for rerun + rewrite
* fix bug with character entry / exit with narration
* linting
* simsuite screenshots
* screenshots
* no " or * just treat as spoken words
* chromadb perist to db
* collect name should contain embedding so switching between chromadb configurations doesn't brick your scenes
* fix save-as long term memory transfer
* add chroma
* director agent refactor
* tweak director command, prompt reset, ux display
* tweak director message ux
* allow clearing of prompt log
* remove auto adding of quotes if neither quote or * are present
* command to reset long term memory for the scene
* improve summarization template as it would cause some llms to add extra details
* rebuilding history will now also rebuild long term memory
* direct scene template
* fix scene time reset
* dialogue template tweaks
* better dialog format fixing
* some dialogue template adjustments
* adjust default values of director agent
* keep track of scene saved/unsaved status and confirm loading a different scene if current scene is unsaved
* prompt fixes
* remove the collection on recommitting the seen to memory, as the embeddings may have changed
* change to the official python api for the openai client and make it async
* prompt tweaks
* world state prompt parsing fixes
* improve handling of json responses
* 0 seconds ago changed to moments ago
* move memory context closer to scene
* token counts for openai client
* narrator agent option: narrate passage of time
* gitignore
* remove memory id
* refactor world state with persistence to chromadb (wip)
* remove world state update instructions
* dont display blank emotion in world state
* openai gpt-4 turbo support
* conversation agent extra instructions
* track prompt response times
* Yi and UtopiaXL
* long term memory retrieval improvements during conversations
* narrate scene tweaks
* conversation ltm augment tweaks
* hide subconfig if parent config isnt enabled
* ai assisted memory recall during conversation default to off
* openai json_object coersion only on model that supports it
openai client emit prompt processing time
* 0.12.0
* remove prompt number from prompt debug list
* add prompt number back in but shift it to the upper row
* narrate time passage hard content limit restriction for now as gpt-4
would just write a whole chapter.
* relock
* fixes#2: character creator description generation will not honor changes to the content context
* decrease output of base attribute generation from 2-3 sentences to 1-2
sentences
* conversation agent tweaks
set other character names as stopping strings via client context
* xwin llm template
* conversation template tweaks
* fixes#6: agent busy status not always reflected in ux
* conversation min response length requirement reduced
include character base details with conversation prompt
* fixes#4: Prompt log
* reset prompt log on scene load
openai tokens as ? for now
* version to 0.9.0
