Fihrist

Use the World Lattice to classify books by what they are about, not where their authors are from.

pip install fihrist

Fihrist is the Python package for the World Lattice: a 3,660-leaf, ten-axis faceted classification system for books.

It ships the Lattice artifact, verifies its hash, builds the prompts for enrichment and slotting, and validates model replies. It makes no network calls and never sees an API key. You choose the model; Fihrist gives you the frame and checks the answer.

The problem this exists to solve

Ask a language model to sort a library and it may file an Indian doctor's genetics books under "India." It may file a Japanese lean-manufacturing book under "Japan" and lose Operations entirely. Both of those are real, from the project this package came out of, and both were caught by a human rather than the machine.

That is not a quirk of one model. It is a century-old bias the models learned from us. Open Dewey Decimal at religion:

	Dewey range
Christianity	200–289
All Indic religions	294 — Hinduism is 294.5

Ninety classes against one decimal place. Library of Congress does the same thing structurally: the Americas get two top-level classes; Asia and Africa share one. Train on a century of text organised that way and a model inherits the shape.

The bias shows up as two distinct failures, and you will hit both:

	Failure	What it costs you
Identity magnet	the author's origin substitutes for the book's content	a wrong home
Completeness	one obvious home hides the others	a missing home

The second is the one that quietly ruins a catalogue. A book on Toyota's production system is about operations and Japan. A novel can be literature and psychology. Force one label and you throw away most of what you knew.

What Fihrist does about it

The World Lattice is built so no tradition sits at the centre as the unmarked default: Christianity 12 leaves, Islam 12, Hinduism 11, Buddhism 10; Ayurveda 12 and Traditional Chinese Medicine 10 beside Western humoral / Hippocratic-Galenic medicine 7. The West is named as a tradition rather than assumed as the baseline. Every leaf carries a definition, so a placement can be argued against a written scope note instead of a label.

Because the axes are independent, a book attaches to as many as its content earns. In practice most books touch five to eight. Nothing has to lose for something else to win.

Fihrist does the deterministic half: it builds the prompt that asks a model to place the book, and checks the answer that comes back. Models invent plausible leaves. Fihrist drops anything not in the frame, repairs a leaf filed under the wrong parent, and tells you when a batch came back silently empty.

Named for the Kitāb al-Fihrist (Baghdad, 987 CE), a bookseller's attempt to index every book then known, across every language he could reach.

---

What you need before you start

Fihrist takes an enriched book record, not a title.

A title and an author are not enough, and that is the whole point rather than a missing feature. The lattice slots on what a book contains. Feed it a bare title and the only thing a model can classify on is the author's name and origin — which is precisely the bias the artifact exists to remove.

So the input is a record carrying idea units: the handful of things the book is actually about. Step 2 below shows how to get them.

What Fihrist does and does not do

Fihrist makes no network calls and never sees an API key. It builds prompts and checks answers. The model call is yours, on your account, with your choice of provider.

Fihrist owns	You own
the artifact, verified against its sha256 on load	the API key
leaf and parent lookup, the embedded parity audits	the model call
building the research prompt	running it
building the slot prompt	running it
validating the reply — invented leaves are dropped

Reads as a limitation for about ten seconds, and as a feature forever: no key, no telemetry, no surprise bill, and it works on a plane.

1 — Inspect the artifact

Offline, immediate, free.

import fihrist

lat = fihrist.load()              # hash-verified against the shipped sha256
len(lat)                          # 3660
lat.axes                          # ['Discipline', 'Place', 'Time', 'Form', 'Genre', ...]
len(lat.leaves("Discipline"))     # 1246

lat.find("Meiji")                 # [Leaf('Modern Japan (Meiji–Shōwa)', axis='Discipline', ...)]
lat.parents_of("Modern Japan (Meiji–Shōwa)", "Discipline")
# ['Japanese history']

lat.substantive_audits            # the 6 written parity audits, carried inside the file
lat.audited_axes                  # which axes they cover

The artifact carries its own audits. You do not have to take the parity claim on trust — open lat.substantive_audits and read what was checked, including the one that returned FAIL on the Place axis.

Six of the ten axes have a written audit. Time, Thing/Matter and Form carry a stub rather than a finding and should be treated as unaudited; Genre has none. lat.audited_axes names the six.

From the shell:

fihrist verify
fihrist axes
fihrist leaves --axis Discipline --grep literature

2 — Enrich a book

Fihrist writes the research prompt. You run it against any model that can search the web.

prompt = fihrist.enrich_prompt("I Am a Cat", "Natsume Sōseki")
record = fihrist.parse_enrichment(your_model(prompt))

record["ideaUnits"]
# [{'id': 'the-outsider-narrator-sees-clearest-1',
#   'title': 'The Outsider Narrator Sees Clearest',
#   'tags': ['narration', 'satire', 'detachment']}, ...]

The instruction body is frozen and byte-identical for every book — only the assignment block at the end changes. That constancy is deliberate. A prompt that said more about a Japanese novel than a French one would manufacture the very difference the artifact exists to measure, then hand it to the slotter as though the lattice had caused it.

3 — Slot it

prompt = fihrist.slot_prompt([record])
results = fihrist.validate(your_model(prompt), [record])

results[0].attached[0]

{"axis": "Discipline", "parent": "Japanese history", "leaf": "Modern Japan (Meiji–Shōwa)",
 "confidence": "stretch", "culture": "Japanese", "status": "valid"}

Batch as many books into one slot_prompt call as your context window allows. The lattice index is rendered first and identically every time, so a prompt-caching backend caches it across a whole run.

Validation is not a formality

fihrist.summarise(results)
# {'books': 1, 'attachments': 7, 'rejectedAttachments': 1,
#  'booksWithNoAttachment': 0, 'leavesPerBook': 7.0, 'byAxis': {...}}

Three things worth knowing:

• Invented leaves are dropped. Models produce plausible leaves in the artifact's own house style. An invented one that survives into your results is worse than a missing one, because it looks like evidence. Anything not on the lattice goes to rejected with a reason, never into attached.
• A wrong parent is repaired, and the repair is recorded in parentRepaired. Polyhierarchy means a leaf can hang from several parents; any true one is accepted.
• Check booksWithNoAttachment first. A batch that comes back structurally valid and substantively empty is the quietest way a run fails — the JSON parses, nothing raises, and the numbers are simply thin. That counter is where it shows.

From the shell

fihrist verify                                    # hash check
fihrist axes                                      # the 10 axes
fihrist leaves --axis Genre --grep thriller
fihrist prompt enrich --title "I Am a Cat" --author "Natsume Sōseki"
fihrist prompt slot books.json > prompt.txt
fihrist validate response.json --books books.json --out slotted.json

---

The claim, and its limits

Seven engines — Claude, GPT, DeepSeek, and local Qwen models from 30B to 235B — slotted the same 1,232 books onto this lattice. All of them top-ranked India, China and Japan on the Place axis. A biased engine buries those. None did. The correction lives in the artifact rather than in a prompt or a model, so swapping the model does not bring the bias back.

What is not established: the 425-book companion corpus should be read as a test dataset and model-comparison record, not as a final population-level proof. It is published because the Lattice is meant to be inspected on books beyond the library it was built beside. See the corpus dataset for the records and known limitations.

Two independent things are worth keeping apart. Audit coverage: six axes carry a written parity audit — Discipline, Place, Agent/Entity, Action/Process, Language, Ethnic/National-group. The other four do not, and Place's own audit records a verdict of FAIL with the remediation that followed. Slotting maturity: three axes — Agent/Entity, Action/Process, Thing/Matter — are provisional as slotting targets and marked as such in fihrist.PROVISIONAL_AXES.

The guard rules

fihrist.GUARDS is the anti-bias correction, carried verbatim from the run that produced the published results. Rewording it invalidates every parity claim attached to this artifact. If you need different behaviour, write your own prompt rather than editing those lines, and do not report the result as a Fihrist placement.

Cost

Fihrist itself is free and offline. Steps 2 and 3 are your own model calls — roughly one search-enabled call and one slot call per book.

Just want to look?

The evidence is published, and you can read it without installing anything:

• bk2022/world-lattice — the artifact this package ships, with its parity audits and the two gaps it still has.
• bk2022/corpus-425 — 425 books slotted independently by three engines, 10,634 placements, and the test records around them.

Source repo: bk2022-ui/lattice.

Licence

Code: Apache-2.0. The artifact (world-lattice.json) and the corpus: CC-BY-4.0.

Attribution is the point. If the lattice is doing work for you, cite it.