diff --git a/klite.embd b/klite.embd index 5432d2930..735ea8f7f 100644 --- a/klite.embd +++ b/klite.embd @@ -12,7 +12,7 @@ Current version indicated by LITEVER below. --> @@ -21916,8 +22353,8 @@ Current version indicated by LITEVER below.
Connecting
- - + +
@@ -22114,12 +22551,18 @@ Current version indicated by LITEVER below.
-
-
-
Toggle Group
- +
+ + + +
+
+
+
Toggle Group
+ +
+
-
@@ -22157,40 +22600,80 @@ Current version indicated by LITEVER below.
-
Search Includes Context History ? - If enabled, the entire story/chat history is used as a searchable document
- +
Use Embeddings for Searching ? + If enabled, the text DB is converted to vectors using the KoboldCPP embedding endpoint. If not, it uses the traditional TextDB keyword based search. Generally, embeddings are slower to generate initially (so when making changes to DB the next response might be very slow) - but may understand the context of the sentence better while also potentially supporting multiple languages.
+
-
Num. Search Results ? - Controls how many snippets to retrieve from DB
- +
Delete Embedding cache ? + When embeddings are enabled, they get generated by the KCPP server and then are cached in the web browser. This button clears the browser cache.
+
+
-
Nearby Text Amount ? - Controls amount of nearby text included in the search query -
- +
Add document to TextDB ? + Tries to extract the text from the document, adding (or overwriting) an existing document in the TextDB with the same name.
+
+
-
Search Chunk Size ? - Controls the size of each snippet being searched for -
- +
Search Includes Context History ? + If enabled, the entire story/chat history is used as a searchable document.
+
+ +
+
+
Num. Search Results ?Controls how many search results to retrieve from DB. Each search result is added to the context as a snippet of the DB + document.
+ +
+
+
+
1
+
10
+
+
+ +
+
+
Extra Text For Search ?Adds the specified amount of characters from the context before the last user message to the search query.
+ +
+
+
+
50
+
500
+
+
+ +
+
+
Snippet Size ?Controls the size (in characters) of each snippet taken from the documents.
+ +
+
+
+
100
+
3000
+
+
+ +
+
Please note, as with world info this can cause context reprocessing. It is recommended to have at least 8K context size and a good prompt processing speed to use this feature.

TextDB Storage?Paste as much raw text data here as you like. E.g. background information, reference documents, etc. This text will populate the database that will be chunked and searched by TextDB.
+ class="helptext">Paste as much raw text data here as you like. E.g. background information, reference documents, etc. This text will populate the database that will be chunked and searched by TextDB. This can be split using "[DOCUMENT BREAK]" to split the overall database into smaller documents, and titles can be added with "[DOCUMENT BREAK][Title of document]This is the content of the document".
- +
@@ -23289,6 +23772,11 @@ Current version indicated by LITEVER below. class="helptext">Allows you to change the connected custom endpoint at runtime even in local mode.
+
+
RenderStreamingMarkdown ?Attempt to render markdown when streaming. May result in wonky output or severe lag.
+ +
@@ -24363,6 +24851,9 @@ Current version indicated by LITEVER below.
+ + + diff --git a/koboldcpp.py b/koboldcpp.py index 8bdd8c5f6..44e8e0f25 100644 --- a/koboldcpp.py +++ b/koboldcpp.py @@ -1690,10 +1690,13 @@ def tts_generate(genparams): voice = 1 speaker_json = tts_prepare_voice_json(genparams.get("speaker_json","")) #handle custom cloned voices voicestr = genparams.get("voice", genparams.get("speaker_wav", "")) + oai_voicemap = ["alloy","onyx","echo","nova","shimmer"] # map to kcpp defaults voice_mapping = ["kobo","cheery","sleepy","shouty","chatty"] normalized_voice = voicestr.strip().lower() if voicestr else "" if normalized_voice in voice_mapping: voice = voice_mapping.index(normalized_voice) + 1 + elif normalized_voice in oai_voicemap: + voice = oai_voicemap.index(normalized_voice) + 1 else: voice = simple_lcg_hash(voicestr.strip()) if voicestr else 1 inputs = tts_generation_inputs()