Working-paper Search

Search prior working papers and findings by meaning.

Industry	Level	Difficulty	Recommended mode
Fiscal auditing	1 of 5	Beginner	flat

Coming soon

A complete, runnable Working-paper Search reference implementation is in progress. The architecture and core code below already work with SrvDB 0.3.

The problem

• Prior-year papers are hard to reuse.
• Findings repeat across entities.
• Keyword search fails.

How SrvDB powers Working-paper Search

1. Working papers and findings are embedded.
2. Audit areas query similar past work.
3. Per-engagement databases.
4. Local storage.

What changes

Today	With Working-paper Search on SrvDB
Starting from scratch	Reusing prior work
Repeated findings missed	Patterns visible
Keyword search	Meaning search

How it works

1. Ingest papers.
2. Embed.
3. Search by area.
4. Reuse.
5. Archive.

SrvDB features it uses

• Exact search (flat): every query compares against every stored vector, so results are exact and explainable.
• Several databases: each database is a directory, so different data types or tenants stay separate.
• Crash-safe persist(): after it returns, data survives a crash or power loss and reopens exactly as committed.
• Fully offline: SrvDB runs in-process and never opens a network connection.

Use cases

• Planning
• Risk assessment
• Testing approaches
• Prior findings
• Standards references
• Training

Core code

embed() is your embedding model (for example sentence-transformers, or a function that turns your records into feature vectors).

import json
import srvdb

db = srvdb.Database("./working-paper-search-db", dimension=768)

def ingest(items):
    """items: list of (id, text, metadata dict)"""
    db.upsert(
        ids=[i for i, _, _ in items],
        embeddings=[embed(t) for _, t, _ in items],
        metadatas=[json.dumps(m) for _, _, m in items],
    )
    db.persist()                      # crash-safe checkpoint

def prior_work(audit_area, k=5):
    hits = db.search(embed(audit_area), k)
    return [(json.loads(db.get(i)), round(score, 3)) for i, score in hits]

Limits and honest notes

• Independence rules apply to sharing.

Build roadmap

1. Prototype: load a sample of real data with flat mode and check that searches return what an expert expects.
2. Measure: if you move to an approximate mode, compare its results with flat on the same queries (recall).
3. Pipeline: add continuous upsert and decide when to call persist().
4. Harden: test restarts and crashes, add logging, review every output path with domain experts.
5. Deploy: ship the database directory with the application or service.

All Fiscal auditing projects

Level	Project	What it does
1 · Beginner	Working-paper Search (this page)	Search prior working papers and findings by meaning.
2 · Intermediate	Finding Drafter	Drafts audit findings and management letters from similar past findings.
3 · Advanced	Duplicate Invoice Detector	Finds duplicate and near-duplicate invoices and vendors.
4 · Expert	Continuous Audit Agent	An agent monitoring entries and contracts continuously.
5 · Frontier	Public Funds Analytics	Cross-entity pattern discovery for public-sector auditors.

See also: Use cases · Live examples (coming soon)