wasserstoff / app /params.yaml
adityasarade's picture
Adding backend files along with docker and requirements.txt
c8769a7
paths:
vector_store_dir: "data/vector_store"
index_file: "index.faiss"
metadata_file: "metadata.pkl"
llm:
model_name: "llama-3.3-70b-versatile"
temperature: 0.3
max_tokens: 3000
embedding:
model_name: "all-MiniLM-L6-v2"
search:
initial_top_k_multiplier: 3
chunk_min_words: 15
ocr:
dpi: 300
tesseract_psm: 6
threshold: 140
prompts:
summarize_doc:
system: "You are a helpful research assistant that summarizes documents."
user: |
Given the document chunks below, write a concise summary (2–3 sentences) of the document's main topic. Also include a citation based on the Page and Paragraph numbers shown.
Document Chunks:
{content}
Respond strictly in this format:
- Answer: <summary>
- Citation: Page X, Para Y
synthesize_themes:
system: "You are a research assistant summarizing themes across documents."
user: |
You are given summaries from several documents. Identify 2–3 core themes across them. Label each theme, list supporting document IDs, and write a short explanation.
Document Summaries:
{context}
Respond in this format:
Theme 1: <Title>
- Supporting documents: Document Name/ID
- Description: <theme explanation>
search:
system: |
You are an AI assistant that synthesizes information into key themes.
Given the following extracted facts from documents, group them into meaningful themes.
Use the format:
Theme 1 – [Title of theme]:
Documents (Document ID/ Name) explain