# =================== Lord Of Large Language Multimodal Systems Configuration file =========================== version: 130 binding_name: null model_name: null model_variant: null model_type: null show_news_panel: true # Security measures turn_on_setting_update_validation: true turn_on_code_execution: true turn_on_code_validation: true turn_on_open_file_validation: true turn_on_send_file_validation: true turn_on_language_validation: true force_accept_remote_access: false # Server information headless_server_mode: false allowed_origins: [] # Host information host: localhost port: 9600 app_custom_logo: "" # Genreration parameters discussion_prompt_separator: "!@>" start_header_id_template: "!@>" end_header_id_template: ": " separator_template: "\n" start_user_header_id_template: "!@>" end_user_header_id_template: ": " end_user_message_id_template: "" start_ai_header_id_template: "!@>" end_ai_header_id_template: ": " end_ai_message_id_template: "" system_message_template: "system" use_continue_message: true seed: -1 ctx_size: 4084 max_n_predict: 4096 min_n_predict: 1024 temperature: 0.9 top_k: 50 top_p: 0.95 repeat_last_n: 40 repeat_penalty: 1.2 num_experts_per_token: 2 n_threads: 8 #Personality parameters personalities: ["generic/lollms"] active_personality_id: 0 override_personality_model_parameters: false #if true the personality parameters are overriden by those of the configuration (may affect personality behaviour) extensions: [] user_name: user user_description: "" use_user_name_in_discussions: false use_model_name_in_discussions: false user_avatar: null use_user_informations_in_discussion: false # UI parameters discussion_db_name: default # Automatic updates debug: false debug_show_final_full_prompt: false debug_show_chunks: false debug_log_file_path: "" auto_update: true auto_sync_personalities: true auto_sync_extensions: true auto_sync_bindings: true auto_sync_models: true auto_save: true auto_title: false # Install mode (cpu, cpu-noavx, nvidia-tensorcores, nvidia, amd-noavx, amd, apple-intel, apple-silicon) hardware_mode: nvidia-tensorcores # Automatically open the browser auto_show_browser: true # copy to clipboard copy_to_clipboard_add_all_details: false # -------------------- Services global configurations -------------------------- # Select the active test to speach, text to image and speach to text services active_tts_service: "None" # xtts (offline), openai_tts (API key required), elevenlabs_tts (API key required) active_tti_service: "None" # autosd (offline), dall-e (online) active_stt_service: "None" # whisper (offline), asr (offline or online), openai_whiosper (API key required) active_ttm_service: "None" # musicgen (offline) # -------------------- Services -------------------------- # ***************** STT ***************** stt_input_device: 0 # STT service stt_listening_threshold: 1000 stt_silence_duration: 2 stt_sound_threshold_percentage: 10 stt_gain: 1.0 stt_rate: 44100 stt_channels: 1 stt_buffer_size: 10 stt_activate_word_detection: false stt_word_detection_file: null # ASR STT service asr_enable: false asr_base_url: http://localhost:9000 # openai_whisper configuration openai_whisper_key: "" openai_whisper_model: "whisper-1" # whisper configuration whisper_activate: false whisper_model: base # ***************** TTS ***************** tts_output_device: 0 # Voice service auto_read: false xtts_current_voice: null xtts_current_language: en xtts_stream_chunk_size: 100 xtts_temperature: 0.75 xtts_length_penalty: 1.0 xtts_repetition_penalty: 5.0 xtts_top_k: 50 xtts_top_p: 0.85 xtts_speed: 1 xtts_enable_text_splitting: true xtts_freq: 22050 # openai_whisper configuration openai_tts_key: "" openai_tts_model: "tts-1" openai_tts_voice: "alloy" elevenlabs_tts_key: "" elevenlabs_tts_model_id: "eleven_turbo_v2_5" elevenlabs_tts_voice_stability: 0.5 elevenlabs_tts_voice_boost: 0.5 elevenlabs_tts_voice_id: EXAVITQu4vr4xnSDxMaL # ***************** TTI ***************** use_negative_prompt: true use_ai_generated_negative_prompt: false negative_prompt_generation_prompt: Generate negative prompt for the following prompt. negative prompt is a set of words that describe things we do not want to have in the generated image. default_negative_prompt: (((text))), (((ugly))), (((duplicate))), ((morbid)), ((mutilated)), out of frame, extra fingers, mutated hands, ((poorly drawn hands)), ((poorly drawn face)), (((mutation))), (((deformed))), blurry, ((bad anatomy)), (((bad proportions))), ((extra limbs)), cloned face, (((disfigured))), ((extra arms)), (((extra legs))), mutated hands, (fused fingers), (too many fingers), (((long neck))), ((watermark)), ((robot eyes)) # Image generation service enable_sd_service: false sd_base_url: http://localhost:7860 # Image generation service enable_fooocus_service: false fooocus_base_url: http://localhost:7860 # diffuser diffusers_offloading_mode: sequential_cpu_offload # sequential_cpu_offload diffusers_model: PixArt-alpha/PixArt-Sigma-XL-2-1024-MS # Dall e service key dall_e_key: "" dall_e_generation_engine: "dall-e-3" # Midjourney service key midjourney_key: "" midjourney_timeout: 300 midjourney_retries: 1 # Image generation service comfyui enable_comfyui_service: false comfyui_base_url: http://127.0.0.1:8188/ comfyui_model: v1-5-pruned-emaonly.ckpt # Motion control service enable_motion_ctrl_service: false motion_ctrl_base_url: http://localhost:7861 # ***************** TTT ***************** # ollama service enable_ollama_service: false ollama_base_url: http://localhost:11434 # petals service enable_petals_service: false petals_base_url: http://localhost:8064 petals_model_path: TinyLlama/TinyLlama-1.1B-Chat-v1.0 petals_device: cuda # lollms service enable_lollms_service: false lollms_access_keys : [] # set a list of keys separated by coma to restrict access activate_lollms_server: true activate_ollama_emulator: true activate_openai_emulator: true activate_mistralai_emulator: true # elastic search service elastic_search_service: false elastic_search_url: http://localhost:9200 # vll service enable_vllm_service: false vllm_url: http://localhost:8000 vllm_model_path: TinyLlama/TinyLlama-1.1B-Chat-v1.0 vllm_gpu_memory_utilization: 0.9 vllm_max_model_len: 4096 vllm_max_num_seqs: 256 # Audio media_on: false audio_in_language: 'en-US' auto_speak: false audio_out_voice: null audio_pitch: 1 audio_auto_send_input: true audio_silenceTimer: 5000 # Data vectorization rag_databases: [] # This is the list of paths to database sources. Each database is a folder containing data rag_vectorizer: semantic # possible values semantic, tfidf, openai rag_vectorizer_model: sentence-transformers/bert-base-nli-mean-tokens # The model name if applicable rag_vectorizer_parameters: null # Parameters of the model in json format rag_chunk_size: 512 # number of tokens per chunk rag_overlap: 0 # number of tokens of overlap rag_n_chunks: 4 #Number of chunks to recover from the database rag_clean_chunks: true #Removed all uinecessary spaces and line returns rag_follow_subfolders: true #if true the vectorizer will vectorize the content of subfolders too rag_check_new_files_at_startup: false #if true, the vectorizer will automatically check for any new files in the folder and adds it to the database rag_preprocess_chunks: false #if true, an LLM will preprocess the content of the chunk before writing it in a simple format rag_activate_multi_hops: false #if true, we use multi hops algorithm to do multiple researches until the AI has enough data rag_min_nb_tokens_in_chunk: 10 #this removed any useless junk ith less than x tokens rag_max_n_hops: 3 #We set the maximum number of hop in multi hops rag contextual_summary: false #If activated this will completely replace the rag and instead will use contextual summary activate_skills_lib: false # Activate vectorizing previous conversations skills_lib_database_name: "default" # Default skills database max_summary_size: 512 # in tokens data_vectorization_visualize_on_vectorization: false data_vectorization_activate: true # To activate/deactivate data vectorization data_vectorization_method: "tfidf_vectorizer" #"model_embedding" or "tfidf_vectorizer" data_visualization_method: "PCA" #"PCA" or "TSNE" data_vectorization_sentense_transformer_model: "all-MiniLM-L6-v2" # you can use another model by setting its name here or its path data_vectorization_save_db: true # For each new session, new files data_vectorization_chunk_size: 512 # chunk size data_vectorization_overlap_size: 128 # overlap between chunks size data_vectorization_nb_chunks: 2 # number of chunks to use data_vectorization_put_chunk_informations_into_context: false # if true then each chunk will be preceded by its information which may waste some context space but allow the ai to point where it found th einformation data_vectorization_build_keys_words: true # If true, when querrying the database, we use keywords generated from the user prompt instead of the prompt itself. data_vectorization_force_first_chunk: false # If true, the first chunk of the document will systematically be used data_vectorization_make_persistance: false # If true, the data will be persistant webween runs # Activate internet search activate_internet_search: false activate_internet_pages_judgement: true internet_vectorization_chunk_size: 512 # chunk size internet_vectorization_overlap_size: 0 # overlap between chunks size internet_vectorization_nb_chunks: 4 # number of chunks to use internet_nb_search_pages: 8 # number of pages to select internet_quick_search: false # If active the search engine will not load and read the webpages internet_activate_search_decision: false # If active the ai decides by itself if it needs to do search # Helpers pdf_latex_path: null # boosting information positive_boost: null negative_boost: null current_language: english fun_mode: false # webui configurations show_code_of_conduct: true activate_audio_infos: true