Fact-check Assistant

Matches claims against previously fact-checked claims and sources.

Industry	Level	Difficulty	Recommended mode
Media & journalism	2 of 5	Intermediate	flat

Coming soon

A complete, runnable Fact-check Assistant reference implementation is in progress. The architecture and core code below already work with SrvDB 0.3.

The problem

• The same false claims recirculate.
• Checking takes time.
• Sources are scattered.

How SrvDB powers Fact-check Assistant

1. Fact-checked claims and verdicts are embedded.
2. New claims retrieve matches with verdicts.
3. Sources cited.
4. Updated continuously.

What changes

Today	With Fact-check Assistant on SrvDB
Re-checking old claims	Instant matches
Scattered sources	Linked sources
Slow response	Faster checks

How it works

1. Embed checked claims.
2. Query with new claims.
3. Review matches.
4. Publish.
5. Add verdicts.

SrvDB features it uses

• Exact search (flat): every query compares against every stored vector, so results are exact and explainable.
• True cosine scores: scores in [-1, 1] mean the same in every mode, so thresholds are stable.
• upsert: inserts new items and replaces changed ones, so the knowledge base stays current.
• Crash-safe persist(): after it returns, data survives a crash or power loss and reopens exactly as committed.

Use cases

• Political claims
• Health misinformation
• Viral posts
• Statistics checks
• Quote verification
• Election coverage

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("./fact-check-assistant-db", dimension=384)

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

Limits and honest notes

• Editors make the final call.

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 Media & journalism projects

Level	Project	What it does
1 · Beginner	Archive Search	Search decades of articles by meaning.
2 · Intermediate	Fact-check Assistant (this page)	Matches claims against previously fact-checked claims and sources.
3 · Advanced	Narrative Tracker	Clusters stories and posts into narratives.
4 · Expert	Investigation Workbench	Isolated, multimodal workspace per investigation.
5 · Frontier	Secure Leak Analysis	Air-gapped analysis of large leaked datasets.

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