Test

CHANGELOG.md

@@ -12,6 +12,38 @@
 </a>
 </div>
 
+# 3.4
+> **Streaming changes**
+turns out I could've fixed the broken formatting long ago, should work great now
+
+> Prompt conversion and **SystemExperiments** reworked
+
+> **Prompts** (PromptMain, PromptReminder, PromptContinue) from 3.0 removed
+
+most frontends are implementing prompt managers so that hack is not needed
+
+> **PersonalityFormat** and **ScenarioFormat** added
+
+those are hardcoded on ST and will stay available until they're not
+
+scenarios and char description are extracted from their hardcoded strings and replaced by the format you set
+
+- **RealChatPrefix** default is now empty
+
+> **AllSamples** changed
+
+no longer excludes the last two messages when transforming
+
+> **LogMessages** changed
+
+moved to Settings
+
+> **Minor changes**
+
+- Error handling changes
+
+- RenewAlways is now more stable when set to false. still, regenerate once if you swap characters
+
 # 3.3
 added \[DONE\] to end of streams
 
@@ -27,7 +59,7 @@ if set to true, prevents the deletion of chats at any point
 # 3.1
 > **Streaming changes**
 
-if the user did not enable streaming, we will **fake** a non-stream response for compatibility
+if the user did not enable streaming, clewd will **fake** a non-stream response for compatibility
 
 > **LogMessages** added (defaults false)
 
@@ -76,7 +108,7 @@ if set to false, check the `Prompts` section below
 
 > **NoSamples** added (defaults false)
 
-if set to true, replaces any H/A prefix from your frontend into Human/Assistant
+if set to true, replaces every previous message with H/A prefix from your frontend into Human/Assistant
 
 mutually exclusive with `AllSamples`
 
@@ -188,7 +220,7 @@ should be more accurate now
 
 uses the AI's own retry mechanism when you regenerate on your frontend, if you change anything from your last prompt before regenerating it will default to old behavior
 
-> **PromptExperiment** added (defaults true)
+> **PromptExperiments** added (defaults true)
 
 an alternative way to send your prompt to the AI, through a file. set to false if it's bad
 both enabled: https://files.catbox.moe/io1q53.webm

README.md

@@ -34,9 +34,37 @@ nodejs>=20.4.*
 
 ### SettingName: (DEFAULT)/opt1/opt2...
 
+ - `PreventImperson`: (false)/true
+    * true trims the bot reply immediately if he says "Human:", "Assistant:", "H:" or "A:"
+    * making it so it doesn't hallucinate speaking as you __(chance of missing some spicy things)__
+
+ - `PromptExperiments`: (true)/false
+    * true is an alternative way to send your prompt to the AI
+    * experiment before setting to false
+
+ - `RetryRegenerate`: (false)/true
+    * true uses the AI's own retry mechanism when you regenerate on your frontend
+    * instead of a new conversation
+    * experiment with it
+
+ - `SystemExperiments`: (true)/false
+    * only has any effect when `RenewAlways` is false
+    * true alternates between Main+JB+User and JB+User
+    * false doesn't alternate
+
+ - `RenewAlways`: (true)/false
+    * true makes a new conversation context each time
+    * false *tries* to reutilize the same old conversation, sending only your actual last message each time, taking into consideration `SystemExperiments`
+
+ - `StripAssistant`: (false)/true
+    * true strips the "Assistant:" prefix from the last assistant message
+
+ - `StripHuman`: (false)/true
+    * true strips the "Human:" prefix from the last human message
+
  - `AllSamples`: (false)/true
     * mutually exclusive with `NoSamples`
-    * true converts every message except the last two to "sample dialogues"
+    * true converts all real dialogue to "sample dialogue"
     * you're "H" and the AI is "A"
     * whatever the AI replies with is kept (only outgoing)
     * [see this](https://docs.anthropic.com/claude/docs/prompt-troubleshooting-checklist#the-prompt-is-formatted-correctly) for more information
@@ -45,54 +73,29 @@ nodejs>=20.4.*
 
  - `NoSamples`: (false)/true
     * mutually exclusive with `AllSamples`
-    * true converts all "sample dialogues" to real dialogue
+    * true converts all "sample dialogue" to real dialogue
     * you're "Human" and the AI is "Assistant"
     * whatever the AI replies with is kept (only outgoing)
     * [see this](https://docs.anthropic.com/claude/docs/prompt-troubleshooting-checklist#the-prompt-is-formatted-correctly) for more information
     - H->Human
     - A->Assistant
+
+ - `LogMessages`: (false)/true
+    * true logs prompt and reply to `log.txt`
+
+ - `ClearFlags`: (false)/true
+    * possibly snake-oil
+    * clears your warnings
 
  - `PassParams`: (false)/true
     * true will send the temperature you set on your frontend
-    * only values under <=1
+    * only values under <=1.0 >= 0.1
     * this could get your account banned
     * if clewd stops working, set to false
 
- - `PreventImperson`: (false)/true
-    * true trims the bot reply immediately if he says "Human:", "Assistant:", "H:" or "A:"
-    * making it so it doesn't hallucinate speaking as you __(chance of missing some spicy things)__
-
- - `PromptExperiment`: (true)/false
-    * true is an alternative way to send your prompt to the AI
-    * experiment before setting to false
-
- - `ClearFlags`: (false)/true
-    * possibly snake-oil
-    * clears your warnings
-
  - `PreserveChats`: (false)/true
     * true prevents the deletion of old chats at any point
 
- - `RenewAlways`: (true)/false
-    * true makes a new conversation context each time
-    * false *tries* to reutilize the same old conversation, sending only your actual last message each time
-
- - `RetryRegenerate`: (false)/true
-    * true uses the AI's own retry mechanism when you regenerate on your frontend
-    * instead of a new conversation
-    * experiment with it
-
- - `StripAssistant`: (false)/true
-    * true strips the "Assistant:" prefix from the last assistant message
-
- - `StripHuman`: (false)/true
-    * true strips the "Human:" prefix from the last human message
-
- - `SystemExperiments`: (true)/false
-    * only has any effect when `RenewAlways` is false
-    * true alternates between Reminder and Continue prompts
-    * false only uses Reminder
-
 
 
 ## Examples
@@ -105,7 +108,7 @@ nodejs>=20.4.*
 ---
 
 **experimental setup**
-> **PromptExperiment**: true
+> **PromptExperiments**: true
 
 > **SystemExperiments**: true