Reporting by manually trawling through documents is resource intensive. BookWyrm's cite endpoint answers complex business questions by finding and verifying evidence within your source files.
Build reporting apps that answer questions from your documents
from bookwyrm import BookWyrmClient
client = BookWyrmClient()
# Answer business questions from your documents
async def generate_report(chunks, questions):
report = {}
for question in questions:
async for citation in client.stream_citations(
chunks=chunks,
question=question
):
if citation.quality >= 3:
report[question] = {
"answer": citation.answer,
"source": citation.text_span,
"confidence": citation.quality
}
return report
# Generate weekly sales report
questions = [
"Which region had the highest growth in Q3?",
"What are the top three client prospects by sales value?"
]
report = await generate_report(sales_chunks, questions)LLMs hallucinate; BookWyrm cites. Every answer is anchored to specific text chunks with a quality score, ensuring your reports are built on facts, not guesses.
Returns the exact text span used to answer.
Quality scores (1-4) let you filter out weak data.
"Reasoning" fields explain the logic behind the match.
Need a quick answer from a sales log or quarterly report? Query your data directly from the terminal.
bookwyrm cite \
--question "Which region had the highest growth in Q3?" \
--file sales_data.jsonl \
--verboseBuild self-updating reports. Integrate the client into your internal tools to answer lists of questions against new data every night.
# Scan for risks in new contracts
async for citation in client.stream_citations(
chunks=contract_chunks,
question="What are the termination clauses?"
):
if citation.quality >= 3:
add_to_risk_report(citation)Identify top prospects from call transcripts instantly.
Audit policy documents for non-compliant phrasing.
Synthesize findings from hundreds of papers in minutes.
Your data pipeline is the foundation for your agentic workflows. Build it right. Get started with the API that's fast to set up, easy to extend, and built for developers.