Small and major additions

Small additions to V3_4, and major to "consistency chain"

README.md

@@ -217,31 +217,39 @@ recommendations: **do not use.**
 
 **For any case, only very high-quality models will work outstandingly well (in most cases); most de-censored (like heretic) or used with specific dataset without proper deeper fine-tuning or/and training will lose "logic integrity" even with very small difference in KL divergence.**
 
-**In assistant mode, nearly any model will work with great improvements in overall performance.**
+**As an assistant, nearly any model will work with great improvements in overall performance.**
 
-**Works well (in ~90% cases) if started from the second iteration (reloading the same character card with same settings, context, first massage and anything else preserved as initially and sending the same input again; which similar to stopping and pressing 'Retry' right after the first word from the first output (which goes either after the first starting message, or with user's input AFTER the first message). Use only if FastForwarding is enabled.**
+**It is recommended to preserve "consistency chain".**   
 
-**Recommeded to preserve "consistency chain".**   
-
-<details><summary>What is consistency chain and how to preserve it:</summary>
+<details><summary>What is consistency chain and how to preserve it (follow if FastForwarding is on):</summary>
 
 Consistency chain is consistent user inputs (a preservation of narrative flow) without any 'Retry' action or any correction for AI's outputs.
 
+**Works well (in >90% cases) if started with the second iteration (reloading the same character card with same settings, context and anything else preserved as initially and sending the same message again; which is similar to stopping and pressing 'Retry' right after any first word from the first output (which goes after AI's initial input, and after user's input) or simply sending the same initial starting message (right from the start for either user or AI).**
+
 If you decide to edit AI's output or user's input, then you need to restore consistency chain.
 
-To restore consistency chain for user's input, you need to send any letter instead of your usual input, then stop right when AI starts writing anything, undo (twice) or remove whatever AI wrote (alongside with inserted letter, {{[INPUT]}} field and any empty fields by the end), then send your desired input once again.
+To restore consistency chain for user's input, you need to send any letter instead of your usual input, then stop right when AI starts writing anything, undo (twice) or remove whatever AI wrote (alongside with inserted letter, {{[INPUT]}} field and any empty field(s) by the end), then send your desired input once again.
 
-To restore consistency chain for AI's output (if you decide to correct it before sending your next input), you need to correct it as usual, then send any letter as user's input after finishing your correction, then stop right when AI starts writing anything, undo (twice) or remove the inserted letter, {{[INPUT]}} field and any empty fields by the end, then send your desired input.
+To restore consistency chain for AI's output (if you decide to correct it before sending your next input), then correct it as usual and send any letter as user's input after finishing your correction, then stop right when AI starts writing anything, undo (twice) or remove the inserted letter, {{[INPUT]}} field and any empty field(s) by the end, then send your desired input.
 
-**It is highly recommended to restore consistency chain during sudden stops (in mid state of Prompt Processing) with any character card (or after any change in Context/World Info).**
+**It is highly recommended to restore consistency chain during sudden stops (in mid state of Prompt Processing) with any character card (but mostly after any change in context/World Info).**
 
-To do so with character card, you will have to start a new session, insert any letter, then stop, reload your desired character card; starting from the second iteration is also recommended.
+To do so with character card, you will have to start a new session, insert any letter, then stop and reload your desired character card; starting from the second iteration is also recommended.
 
 For incomplete stops during user's large Prompt Processing input, you need to:
 - Use same steps from 'consistency chain for user's input' 
 
  For incomplete stops during AI's large Prompt Processing output, you need to:
-- Undo once or remove user's input, send any letter instead of your input, stop right when AI starts writing, undo twice or remove the resulted output alongside with lettered input, {{[INPUT]}} field and any empty fields by the end, then insert your desired input.
+- Undo once or remove user's input, send any letter instead of your input, stop right when AI starts writing anything, undo twice or remove the resulted output alongside with lettered input, {{[INPUT]}} field and any empty fields by the end, then insert your desired input.
+
+**ALWAYS pin and use 100% out of any entries in World Info to preserve excellent quality (and start with second iteration as mentioned earlier)**; World Info's context (Memory) reprocessing is terrible and often destroys overall quality, style, consistency and completely changes the narrative flow, which is very noticeable in comparsion with both options preserved; also any restoration of consistency chain is useless (without following both options) and (in ~95% cases) will degrade the resuls in numerous aspects, so always pin and use 100%.
+
+If you decide to edit character card's context (Memory) and/or World Info during mid Prompt Processing or after any output of AI or load 'improved version' of same character card with altered context and World Info), then you need to do the same steps from 'in mid state of Prompt Processing'.
+
+If you decide to use 'AutoGenerate Memory' option and save the resulted character card after waiting for the generation of summary, then you need to do the same steps from 'in mid state of Prompt Processing' if you decide to use the newly resulted character card again.
+
+TextDB causes context reprocessing in most cases, so use World Info (with 2 mentioned options) to ensure better quality and greater results.
 
 </details>
 <details><summary>List of approved models and potential candidates:</summary>
@@ -513,6 +521,8 @@ Works poorly in most cases with average-quality character cards.
 Insanely good complexity under well-written character cards, with outstanding engagament between characters, developement of events and emotional complexity between multiple characters and so on.
 
 Requires well-written character card in order to work properly or at full potential.
+
+The best one in generic estimation of overall model's performance, since proven to be much more adaptive with overall better performance if used with model that has superior quality.
 </details>
 
 To preserve the versatility, I would like to describe complete and specific sampling values below **Additional fine-tuning:**, to aim perfection for any case.