Doctrine Assistant

Air-gapped question answering over manuals and doctrine.

Industry	Level	Difficulty	Recommended mode
Government & defence	1 of 5	Beginner	flat

Coming soon

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

The problem

• Manuals run to thousands of pages.
• Networks are isolated.
• Answers must cite sources.

How SrvDB powers Doctrine Assistant

1. Manuals are embedded on an isolated machine.
2. Questions retrieve passages; a local LLM answers with citations.
3. No network access at all.
4. Updates via upsert.

What changes

Today	With Doctrine Assistant on SrvDB
Manual lookup	Cited answers in seconds
No AI on isolated networks	Fully offline AI
Inconsistent interpretations	Shared sources

How it works

1. Import documents.
2. Embed locally.
3. Answer with citations.
4. Log usage.
5. Update with new editions.

SrvDB features it uses

• Fully offline: SrvDB runs in-process and never opens a network connection.
• Exact search (flat): every query compares against every stored vector, so results are exact and explainable.
• 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

• Procedure lookup
• Equipment manuals
• Training support
• Policy questions
• Standard operating procedures
• Exercise preparation

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("./doctrine-assistant-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 cite(question, k=5):
    hits = db.search(embed(question), k)
    return [(json.loads(db.get(i)), round(score, 3)) for i, score in hits]

Limits and honest notes

• Content classification rules apply to the database folder.

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 Government & defence projects

Level	Project	What it does
1 · Beginner	Doctrine Assistant (this page)	Air-gapped question answering over manuals and doctrine.
2 · Intermediate	Citizen Service Bot	Answers citizens’ questions from official public guidance.
3 · Advanced	Report Linker	Links related reports and requests across departments.
4 · Expert	Field Intelligence Agent	An offline agent on rugged laptops combining text and imagery.
5 · Frontier	Sovereign AI Stack	A fully reviewed, air-gapped AI stack with a published SBOM.

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