Grimm Fairy Tales x RedDB

A multi-model data showcase built from 206 canonical Grimm tales.
Graphs, tables, KV, time series, statistics, provenance, and readable stories in one embedded RedDB snapshot.

Open the visual docs  |  Explore questions  |  Run locally  |  RedDB features

---

Fairy tales look simple until you try to ask them precise questions.

Who is the predator when the predator is not a wolf? Which forests behave like thresholds rather than scenery? Which stories reuse the same machinery of abandonment, deception, rescue, punishment, and transformation?

The Grimm corpus is full of patterns that readers feel immediately, but that are hard to materialize as structured data without flattening the stories. This project turns that literary problem into a working RedDB showcase.

The gold layer is a curated canonical collection of 206 Grimm tales. Each tale keeps readable text, provenance back to book editions, KHM/ATU metadata, characters, curated traits, objects, locations, world laws, moral regimes, Propp-like narrative events, branches, and textual evidence. The embedded RedDB snapshot then lets the same corpus be explored as graph, tables, KV, time series, and statistics.

---

The Pitch

If you are...	This repo gives you...
A developer	A concrete embedded RedDB app with CLI queries, generated docs data, and CI publishing.
A researcher	A structured corpus where motifs, characters, provenance, and story functions are queryable.
A literary reader	A way to read the tales and compare canonical text against source versions.
A database person	A compact demo of multi-model workloads using one local RedDB snapshot.

Run It Locally

./grimm setup
./grimm about
./grimm ask
./grimm query "GRAPH CENTRALITY"
./grimm insights stats

Everything above points at the embedded database in output/embedded.rdb. The server/container path is intentionally separate; this showcase is embedded-first.

What RedDB Is Doing

RedDB capability	Where it appears	Why it matters
Graph	tales	Characters, curated traits, motifs, Propp events, provenance edges, and tale structure.
Tables	tale_words, tale_bigrams, tale_vocab	Text evidence, vocabulary, frequency, and lexical exploration.
KV	kv_default	Snapshot metadata and corpus-level facts.
Time series	ingest_log	Ingest counters and timing metrics for the build itself.
Statistics	centrality, fingerprints, ranked counts	Fast insight surfaces on top of the graph and text tables.
Docs export	docs/data/*.json	Static visual docs generated from the embedded snapshot in GitHub Actions.

Snapshot At A Glance

Metric	Count
Canonical tales	206
Graph nodes	8,180
Graph edges	42,532
Curated trait nodes	50
Propp event nodes	2,355
Character nodes	1,563

Collection	Model	Rows/entities
tales	graph	50,712
tale_words	table	53,946
tale_bigrams	table	4,075
tale_vocab	table	206
ingest_log	time series	2,335
kv_default	KV	6

Traits are intentionally curated. The gold YAML still preserves the raw character labels for auditability, but the graph promotes only reusable character traits such as Clever, Helper, Adversary, Witness, Royal, Hidden Or Disguised, and Cursed Or Transformed.

---

Questions This Corpus Can Answer

The README is intentionally question-first. Each block gives the answer, the command that reproduces it, and the RedDB surface being exercised.

Question	Short answer	Try it	RedDB surface
Which predators play the same narrative role across species?	Predator is a narrative job, not just a species.	./grimm ask predators	graph + statistics
Which tales share the same narrative machinery?	Similarity over fingerprints finds structural neighbors beyond title/theme.	./grimm ask tale-machinery	graph + statistics
Can raw word tables ground a narrative pattern?	Word frequencies show where motifs concentrate in the prose itself.	./grimm ask devouring-words	tables
What is structurally central in the corpus?	Centrality surfaces reusable machinery: traits, functions, agency, locations.	./grimm insights stats	graph + statistics
How does one tale unfold as narrative functions?	A tale becomes an ordered chain of Propp-like events with actors and scenes.	./grimm ask propp-hansel	graph
Which books feed a canonical tale?	Gold text stays separate from its source editions, while keeping provenance.	./grimm ask provenance	KV + gold metadata
What did the ingest write, and how expensive was it?	The demo records its own build counters and timing rows.	./grimm ask ingest	time series + KV

Q1. Which predators play the same narrative role across species?

Answer: predator is a narrative job, not just a species. A wolf, a fox, a cannibal band, a witch, an old man, or a bird of prey can occupy the same structural role when the graph says they share arc_predator.

./grimm ask predators

fox_geese_fox                  The Fox
strange_musician_wolf          The Wolf
wolf_man_wolf                  The Wolf
frau_trude                     Frau Trude
hansel_gretel_witch            The Cannibal Witch in the Edible House

Q2. Which tales share the same narrative machinery?

Answer: Cinderella's closest structural neighbors are not only other princess stories. The graph links it to tales with cruel stepfamilies, false brides, oppressed maidens, repeated trials, birds, houses, trees, and symbolic numbers.

./grimm ask tale-machinery

26.5%  the_three_little_men_in_the_forest
25.0%  the_juniper_tree
22.9%  frau_holle
21.2%  allerleirauh_all_kinds_of_fur
20.0%  hansel_and_gretel

Q3. Can raw word tables ground a narrative pattern?

Answer: the graph says wolves matter; the word table lets us check where wolf-language concentrates in the actual text.

./grimm ask devouring-words

the_two_brothers                    18
little_red_cap                      18
the_wolf_and_the_seven_young_kids   17
the_wolf_and_the_fox                17
old_sultan                          11

Q4. What is structurally central in the corpus?

Answer: the center is not only famous characters. The corpus center is made of reusable story machinery: curated traits, narrative functions, appearances, agency, locations, and Propp events.

./grimm insights stats

HAS_TRAIT                 3,181
HAS_NARRATIVE_FUNCTION    4,710
APPEARS_IN                3,824
GAINS_AGENCY_THROUGH      3,313
HAS_LOCATION              2,414
HAS_PROPP_EVENT           2,355

Q5. How does one tale unfold as narrative functions?

Answer: Hansel and Gretel can be read as prose, but also as an ordered chain of events. The witch's imprisonment scene is materialized as a Propp event with actor, function, scene, and evidence.

./grimm ask propp-hansel

propp_order      9
propp_function   func_villainy
actor            hansel_gretel_witch
scene            witch imprisons Hansel, fattens him, and forces Gretel to cook

Q6. Which books feed a canonical tale?

Answer: the gold text is canonical, but not source-less. Hansel and Gretel keeps references to multiple book versions, including source title, book id, match method, and whether that source is the canonical base.

./grimm ask provenance

pg5314-grimm-hunt      canonical      Hansel and Gretel     match: slug
pg2591-grimm-taylor    supplementary  HANSEL AND GRETEL     match: slug
pg11027-grimm-gruelle  supplementary  HANSEL AND GRETHEL    match: khm
pg52521-grimm-olcott   supplementary  HAENSEL AND GRETHEL   match: khm

Q7. What did the ingest write, and how expensive was it?

Answer: the demo records its own ingestion work. ingest_log stores batch timings and counters as time-series rows, so the corpus can explain how it was built.

./grimm ask ingest

nodes_batch_ms
nodes_inserted
edges_batch_ms
edges_inserted
edges_total_ms
ingest_total_ms

---

Visual Docs

The GitHub Pages docs site is the visual side of the showcase.

Mode	What you can do
Reader	Read the gold tale in a clean browser view.
Sources	Inspect original silver book versions without duplicating files in the repo.
Compare	Put canonical text and source text side by side.
Atlas	Browse tale stats, facets, graph counts, and corpus-level questions.
RedDB	See how graph, table, KV, time-series, and statistics data support the same UX.

./grimm export docs
./grimm docs serve

docs/data/*.json is generated, not committed. GitHub Actions rebuilds the embedded snapshot, exports the static JSON, verifies the docs payload, and publishes docs/ to Pages.

Pipeline

input/1-bronze    raw Project Gutenberg books
input/2-silver    extracted books, source texts, canonical texts, branches
input/3-gold      curated 206-tale canonical graph + corpus metadata
output/embedded.rdb
docs/data/*.json  generated visual docs data

Useful Commands

Command	Purpose
./grimm rebuild	Build corpus metadata, validate gold, ingest words, ingest graph.
./grimm export docs	Generate docs/data/*.json from the embedded snapshot.
./grimm docs serve	Serve the Docsify site locally.
./grimm read hansel-and-gretel	Print a canonical gold tale excerpt and provenance.
./grimm query "SELECT COUNT(*) FROM tale_vocab"	Run a raw query against embedded RedDB.
./grimm insights words --word wolf	Explore text-table evidence for one word.

Project Map

Path	Purpose
grimm	Root CLI entry point.
src/embedded	Embedded RedDB ingestion, query, insights, and docs export.
src/shared	Graph loading and SQL helpers.
scripts	Gold validation and corpus build scripts.
input/3-gold	Canonical curated dataset.
docs	Docsify site and generated visual experience.