config.default.toml

[llm_endpoint]

label = "OpenHermes Mistral 7B"
base_url = "http://localhost:8000"
class = "llama_cpp_http_chat"
sysmsg = "You are the living-bookmarks agent, a useful assistant who tries to find links relevant to the user's present interests from a database. The following context section includes some relevant links which were found."
sys_postscript = "Based on this context, please respond to the user's query, ieally with one or more of the links from context."
#  If you need more information, please ask the user for it. If you need to provide a link, please do so. If you need to provide a summary

# https://huggingface.co/docs/transformers/main_classes/text_generation#transformers.GenerationConfig
# Interesting: https://www.reddit.com/r/LocalLLaMA/comments/17vonjo/your_settings_are_probably_hurting_your_model_why/
# Sampling params for llama.cpp https://github.com/ggerganov/llama.cpp/blob/master/common/sampling.h
[model_params]

temperature = 0.7
max_tokens = 2048
min_p = 0.1

[vectordb]

bookmarks_table_name = "test_bookmarks"
embedding_model = "all-MiniLM-L6-v2"