config.yml_template

paths:
    raw_markdown: "./knowledge/raw"
    intermediate: "./knowledge/intermediate"
    processed_track: "./knowledge/processed_files.json"
    vector_index: "./knowledge/faiss_index.bin"
    inverted_index: "./knowledge/inverted_index.json"
models:
    summarization: "gpt-4"
    tag_extraction: "gpt-4"
    vision_caption: "gpt-4o"
    text_embedding: "text-embedding-3-small"
    query_summarization: "gpt-4"
    dialog: "gpt-4"
    answer_summarization: "gpt-4"
vector_db:
    m: 32  # parameter for HNSW
    c: 200  # construction trade-off parameter
    s: 50  # search trade-off parameter
tag_merge:
    chunk_size: 300  # how many tags per chunk per level
    final_threshold: 400  # stop recursion when tags ≤ this
generative:
    max_summary_tokens: 8191  # for embedding using text-embedding-3-small
qa:
    top_k: 5  # first pass
    tag_threshold: 3  # min docs/tag to consider
    max_ref_docs: 10  # cap on |references|
    random_seed: 42
    summarize_chunk_size:  5
    max_summarization_depth:  3