0.25.5 (#121)

* oepnai compat client to /completions instead of chat/completions
openai compat client pass frequency penalty

* 0.25.5

* fix version

* remove debug message

* fix openai compat client not saving coercion settings

* openai compatible client: API handles prompt template switches over to chat/completions api

* wording

* mistral std template

* fix error when setting llm prompt template if model name contained /

* lock sentence transformers to 2.2.2 since >=2.3.0 breaks instructor model loading

* support png tEXt

* openai compat client: fix repetition_penality KeyError issue

* presence_penalty is not equal to repetition_penalty and needs its own dedicated definition

* round presence penalty randomization to one decimal place

* fix filename

* same fixes for presence_penalty ported to koboldcpp client

* kcpp client: remove a1111 setup spam
kcpp client: fixes to presence_penalty jiggle

* mistral.ai: default model 8x22b
mistral.ai: 7b and 8x7b taken out of JSON_OBJECT_RESPONSE_MODELS

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.masonry.api"
 
 [tool.poetry]
 name = "talemate"
-version = "0.25.4"
+version = "0.25.5"
 description = "AI-backed roleplay and narrative tools"
 authors = ["FinalWombat"]
 license = "GNU Affero General Public License v3.0"
@@ -51,7 +51,8 @@ chromadb = ">=0.4.17,<1"
 InstructorEmbedding = "^1.0.1"
 torch = ">=2.1.0"
 torchaudio = ">=2.3.0"
-sentence-transformers="^2.2.2"
+# locked for instructor embeddings
+sentence-transformers="==2.2.2"
 
 [tool.poetry.dev-dependencies]
 pytest = "^6.2"