61d01984ba
* pytorch update * github workflow for tests * tests set up config * tests download nltk punkt * punkt_tab * fix world state not updating on new initial scene load * fix new character creation from scene tools * py 312 py 313 support remove unreliant cuda detection and just lock poetry with cuda * fix tests * dont auto install cuda * remove unused StrEnum import * separate cuda install * fix auto progress OFF no longer working * fix debug logging options not sticking * disable revision during image prompt generation * prompt tweaks * prompt tweaks * fix some issues with directed character creation * tweak the continue generation button so its less prone to pop into a new line on itsown * fix context db filter field layout * handle error when trying to regnerate passthrough narrator message * prompt tweaks * increase auto direct evaluation length * fix node library on windows * auto direct eval tweaks * prompt tweaks * prompt tweaks * allow updationg of scene phase through director console * add generate action to director console phase intent text fields * prompt tweaks * track iteration count in scene loop always yield to user at initial start / load of a scene * fix issue with Split and Join nodes when passed \n as delimiter * sim suite only generate auto direct requirements if auto direct is enabled sim suite fix issues with title generation * autocomplete button disable until there is text to autcomplete * update installation docs * update scenario tools docs * docs * writing style phrase detection * typo * docs * fix issue where deleting an applied preset group would prevent selection of a different preset group in the affected client * fix @-Instruction is broken when using apostrophes * editor never attempt to fix exposition on user input if input starts with command characters @, ! or / * prompt tweaks * editor revision: automatic revision setting, prompt tweaks, docs * missing files * fix issue where narration responses starting with # would result in empty messages * prompt tweaks * fix issue with autocomplete not working at the beginning of a scene * fix issues where cached guidance would result in no guidance * editor revision analysis now has access to scene analysis and director guidance if it exists * fix issue where all nodes in the node editor would be locked on winsows OS * add `scene/SetIntroducation` node * fix issue where generating narration in a scene with zero characters would always come back blank * SceneIntroduction node state output fixed * node editor docs progress * fix issue with loading scene from file upload no longer working * better handling of what to do when there are no characters in a scene and no default character is defined * typo * silence trigger game loop debug message * docs * GenerateThematicList node * docs * docs * stubs * allow creation of module from existing nodes * move function into plugin * separate graph export functions into own .js file * group from selected nodes * remove debug output * tweak create module dialog * docs * docs * graph tests need to assume auto_progres True * add ModuleProperty node * fix some issues in the module creation modal when extending or copying a module * include module name in module deletion confirm dialog * fix node module copy not setting updated registry * module property name output * docs * docs * initializing a scene from a character card will no longer break the node editor * docs * when greeting text and intro do not match, do the greeting text first. * intro is set during card import, there is no need to ever emit character greetings in addition to the intro, its not really something that matches talemate's design philsophy at this point * docs * dynamic premise modules * tweaks * propagate module level property defaults to node * docs * fix issue where the default character would get added to scenes during load even though there already was a player character * prompt tweaks * tweaks to generate-premise module * docs * infinity quest dynamic story v2 * tweaks limits * fix line endings * prompt tweaks * fix some issues with node property editing * formatting * prompt tweaks * group and comment * add talemate tint node style fix gap in nodes when there are no sockets * node style fixes * docs * docs * icon for swtich nodes * conditional emit status node * don't reset dynamic premise * dynamic premise tweaks * dynamic premis status messages * fix issue with Input value error node * validate that topic is specified * fix issue where cancelling a generation triggered during scene loop init would cause the loop to reinitialize * docs * node error handling improvements * docs * better error handling in generate premise module * a connected socket should always override a property of the same value even if the socket connection is not resolved * dynamic premise expose intro length property * fix some issues with node module creation and add registry path validation * correctly title creative loop module so it can be copied * remove debug message * rename to dynamic storyline for clarity and so it doesn't collide with tutorial * import dynanimc storyline node * docs * gracefully handle a node missing during graph load * docs * make agent list a bit more compact * disable node property editing in inherited nodes * rename editor revision events so they are inline with the other agent events * support instruction injection for director.generate_choices action * normalize dynamic instructions * fix director guidance dynamic instructions * docs * generate choices event add choices property * prompt tweaks * add dynamic instruction node * prompt tweak * fix issue where some events could not be hooked into through event node modules * docs * clean response node * docs * docs * docs * module library tweaks * fix some issues with act-as selection * dont allow creation of new modules until scene has been saved at least once * public node modules dir * sim suite module styles * remove debug messages * fix default node titles * fix number input vlaidation in math nodes * context awareness in contextual generate now includes character info * fix dupe id warnings * alt drag to clone multiple nodes * alt drag to clone multiple nodes * docs * docs * fix issue where some scene modules could leak across scenes * dynamic instructions already included through extra context * prompt tweaks * update defaults * docs * make list init from json * socs * fix issue where not all field definitions would get sent * docs * fix issue causing deep analysis to loop * case insentive regex matching * prompt tweaks * fix node title * fix size issue in scene view when node editor was open on smaller resolutions * fix issue with autocomplete sometime loosing markup * add claude 4 * make director guidance and actor instructions available to autocomplete prompt * fix trim node handling of \n * extract node * extract node trim property * remover cruft * charactermessage node only set character if not already part of message prefix * editor revision unslop mode * fix search scenes showing node module files * prompt tweaks * unslop support unwanted phrase detection * define bad_prose * </FIX> seems to get ommitted a lot so lets handle it * cleanup * return on no issues * fix some issues with character creation * fix some character creation issues * prompt tweaks * contewxtual generate and autocomplete signals added * prefix revision_ * use uuidv4 which is already installed and doesnt come with compatibility issues * editor revision support for contextual generations normalize some event objects * add template picker to worldstate character creation interface * prompt tweaks * dont unslop scene intent or story intent generation * prompt tweaks * prompt tweaks * prompt tweaks * prompt tweaks * prompt tweaks * prompt tweaks * prompt tweaks * prompt tweaks * prompt tweaks * prompt tweaks * fix issue of conversation losing edits through signals * support revisions for world state entry generation * task_instructions * dont show rewrite specific options if unslop method is selected * docs * fix issue with setting locked save file when auto saving was turned on * don't trigger player_turn_start in creative mode * better check for when to fire player_turn_start and when not * node editor crash course fixes * docs * fix issue where sometimes wrong system prompt was selected * add world context to contextual generate node * fix node dragging issue when cloning multiple nodes * support editor revision for summarization generation * summarization support dynamic instructions * dedicated template for summary unslop * pass summarization_history to template vars * prompt tweaks * prompt tweaks * not sure how this got blanked * wording * wording * only display debug log if TALEMATE_DEBUG=1 * fix syntax * fix syntax * remove unused cruft * start-debug * log message cleanup * docs * tweak defaults * sim suite polish * remove unused images * add writing style instructions to revision prompts * missing endif tag * prompt tweaks * prompt tweaks * prompt tweaks * use app-busy * prompt tweaks * update readme and screenshots wording wording * add discord link
5.9 KiB
Settings
General
Fix exposition
If enabled the editor will attempt to fix exposition in the generated dialogue.
It will do this based on the selected format.
Fix narrator messages
Applies the same rules as above to the narrator messages.
Fix user input
Applies the same rules as above to the user input messages.
Add detail
Will take the generate message and attempt to add more detail to it.
Long Term Memory
--8<-- "docs/snippets/tips.md:agent_long_term_memory_settings"
Revision
When :material-typewriter: revision is enabled the editor will analyze and attempt to fix character messages, narrator messages, and contextual generation (such as character attributes, details, world context, etc.).
Currently it can detect the following issues:
- Repetition of phrases / concepts
- Unwanted prose as defined in the writing style
The revision action is split into three parts:
- Find any issues through fuzzy, regex and semantic similarity (embeddings) matching
- Analyze the issues and plan a fix
- Apply the fix
This means it comes at a noticable delay IF it finds issues, but the improvements may be worth it.
Enable Revision
Check this to enable revision.
Automatic Revision
Check this to enable automatic revision - this will analyze each incoming actor or narrator message and attempt to fix it if there are issues.
Automatic Revision Targets
When automatic revision is enabled, you can choose which types of messages to automatically revise:
- Character Messages - Automatically revise actor actions and dialogue
- Narration Messages - Automatically revise narrator actions and descriptions
- Contextual generation - Automatically revise generated context such as character attributes, details, world context, scene intros, etc.
By default, both Character Messages and Narration Messages are enabled. You can enable or disable each type independently based on your preferences.
Revision Method
Which method to use to fix issues.
Dedupe (Fast and dumb)- this is the defaultUnslop (AI assisted)- calls 1 additional prompt after generation to remove repetition, purple prose, unnatural dialogue, and over-descriptionTargeted Rewrite (AI assisted)- analyzes text for specific issues and rewrites only problematic parts
Dedupe (Fast and dumb)
When Dedupe is active it will be restricted to finding repetition and removing it without replacing it with something else, nor understanding the intent or context, so it may sometimes cause disjointed dialogue. This method is much faster as it will never prompt the AI for analysis and fixes. (If repetition detection is set to semantic similarity, it will still use the embedding function to find repetition, but it will not prompt the AI for analysis and fixes.)
Unslop (AI assisted)
When Unslop is active, it calls 1 additional prompt after a generation and will attempt to remove repetition, purple prose, unnatural dialogue, and over-description. May cause details to be lost. This is a general-purpose cleanup method that can also use unwanted prose detection when enabled.
Targeted Rewrite (AI assisted)
When Targeted Rewrite is active, unwanted prose detection becomes available and when issues are detected the AI will attempt to rewrite the message to fix the issues. This method checks for specific problems first, then only rewrites if enough issues are found.
Repetition
Repetition Detection Method
How is repetition detected.
Fuzzy- fuzzy matching will match based on character-level similarity, finding text that is approximately the same with small differences (like typos, missing letters, or minor variations). This is faster but less context-aware than semantic matching.Semantic Similarity- will match based on the semantic meaning of the text using the Memory Agent's embedding function. (default)
!!! warning "Semantic Similarity" Uses the memory agent's embedding function to compare the text. Will use batching when available, but has the potential to do A LOT of calls to the embedding model.
Similarity Threshold
How similar does the text need to be to be considered repetitive. (50 - 100%)
You want to keep this relatively high.
Repetition Range
This is the number of MESSAGES to consider in the history when checking for repetition.
At its default of 15 it means the last 15 messages will be considered.
Repetition Min. Length
The minimum length of a phrase (in characters) to be considered for repetition. Shorter phrases will be ignored.
AI-Assisted Methods (Unslop & Targeted Rewrite)
Once switched to either Unslop or Targeted Rewrite mode, additional settings become available.
Preferences for Rewriting (Targeted Rewrite only)
Test parts of sentences, split on commas
If active this means that when a sentence doesn't produce a hit, if it has one or more commas it will split the sentence on the commas and test each part individually.
Minimum issues
The minimum amount of issues that need to be detected to trigger a rewrite.
Unwanted Prose (Both Unslop & Targeted Rewrite)
Detect Unwanted Prose
Check this to enable unwanted prose detection. Available for both Unslop and Targeted Rewrite methods.
--8<-- "docs/snippets/common.md:editor-revision-unwanted-prose-requirement"
Unwanted Prose Threshold
Similarity threshold for unwanted prose detection. (0.4 - 1.0)
You want to keep this relatively high.