Corpus Search

Search across all uploaded documents to find relevant content across your entire collection.

Quick Start

from adx import ADX

dn = ADX()

# Upload some documents
dn.upload("report_q1.pdf")
dn.upload("report_q2.pdf")
dn.upload("budget.xlsx")

# Search across all of them
hits = dn.search_corpus("revenue forecast")
for hit in hits:
    print(f"{hit['filename']} — {hit['text_snippet'][:80]}...")
    print(f"  Location: {hit['citation']}")
    print(f"  Score: {hit['score']}")

Parameters

Parameter	Type	Default	Description
query	str	required	Search query (tokenized, case-insensitive)
file_ids	list[str] | None	None	Limit search to specific files (None = all)
max_results	int	20	Maximum results to return

Search Result

Each hit includes:

Field	Description
file_id	Source document ID
filename	Source filename
text_snippet	Matching text (up to 200 chars)
page_number	Page number (PDFs/DOCX) or None
sheet_name	Sheet name (spreadsheets) or None
score	Relevance score (token hit count)
citation	Human-readable source reference

Results are ranked by score (highest first).

Filtering by File

Search only within specific documents:

hits = dn.search_corpus(
    "total expenses",
    file_ids=["abc123", "def456"],
    max_results=10,
)

Async Usage

from adx import AsyncADX

async def search():
    dn = AsyncADX()
    hits = await dn.search_corpus("quarterly revenue")
    return hits

REST API

curl -X POST http://localhost:8000/v1/search \
  -H "Content-Type: application/json" \
  -d '{"query": "revenue forecast", "max_results": 20}'

vs. Single-Document Search

Feature	search(file_id, query)	search_corpus(query)
Scope	One document	All documents
Speed	Fast	Scales with corpus size
Use case	Drill into a known document	Find which documents are relevant