Datasets:

ente-ai
/

ensu-knowledge-packs

meta_codec string	meta_rows_per_block int64	dataset string	granularity string	selection string	max_chars int64	url_template string	model string	embedding_backend string	source_dim int64	matryoshka bool	doc_prompt string	query_prompt string	dim int64	count int64	quant string	scale int64
zstd-blocks	128	enwiki-2026-07-07	lead-only	pageview-median-top500k + simple-safety-net	600	https://en.wikipedia.org/wiki/{title}	ggml-org/embeddinggemma-300M-GGUF:embeddinggemma-300M-Q8_0.gguf	llama.cpp (Q8_0)	768	true	title: {title} \| text: {text}	task: search result \| query: {query}	512	526,155	int8	127
zstd-blocks	128	simplewiki-2026-07-02	lead-only	null	600	https://simple.wikipedia.org/wiki/{title}	ggml-org/embeddinggemma-300M-GGUF:embeddinggemma-300M-Q8_0.gguf	llama.cpp (Q8_0)	768	true	title: {title} \| text: {text}	task: search result \| query: {query}	512	262,764	int8	127
zstd-blocks	128	enwikibooks-2026-07-02	section-chunk	null	1,400	https://en.wikibooks.org/wiki/{title}	ggml-org/embeddinggemma-300M-GGUF:embeddinggemma-300M-Q8_0.gguf	llama.cpp (Q8_0)	768	true	title: {title} \| text: {text}	task: search result \| query: {query}	512	250,000	int8	127

Ensu Knowledge Packs

Prebuilt on-device retrieval indexes ("knowledge packs") for Ensu, ente's private on-device AI assistant — plus the scripts that generate them. Ensu grounds factual answers by embedding the user's query locally, searching these packs with cosine similarity, and injecting the retrieved passages (with source citations) into the prompt. Everything runs on-device; no query ever leaves the phone.

Layout

Each dataset lives in its own self-contained folder: generation scripts, the generated pack, and dataset-specific docs together. Code shared across packs (the embedder, index writer, dump utilities) lives in shared/.

Folder	Source	Granularity	Rows	Pack size
`fullwiki/`	Full English Wikipedia, latest dump (pageview-curated ~552k)	lead-only (one passage per article)	526,155	~358 MB
`wikibooks/`	English Wikibooks, latest dump	section-chunk (curated, ≤1,400 chars, incl. Cookbook)	250,000	~193 MB
`simplewiki/`	Simple English Wikipedia, latest dump	lead-only (one passage per article)	262,764	~160 MB

fullwiki answers what (definitions, facts) across the head and the mid-tail the on-device model doesn't know; wikibooks answers how/why (explanations, worked examples, procedures). They share model, dim, and quantization, so their cosine scores are comparable and results merge into one global top-k.

fullwiki supersedes simplewiki — it covers the same notable topics (via the pageview safety-net) plus the long tail, in full-English prose. simplewiki is retained for the simpler-language / smallest-footprint use case; ship one or the other, not both.

Inside a dataset folder:

scripts/ — build.py (dump → rows → embeddings → pack) and search.py (quality eval / threshold tuning), plus any dataset-specific analysis tools
data/ — the generated pack: vectors.i8, meta.zst, meta.offsets, manifest.json

Shared: shared/ — common.py (embedder, sharded/resumable embedding, block-zstd meta writer, MetaReader, dump helpers) and requirements.txt.

Pack format

Exactly what the Ensu Rust RetrievalIndex reads:

vectors.i8 — raw row-major count × dim int8; L2-unit vectors × 127
meta.zst — row metadata as independent zstd frames ("blocks") of meta_rows_per_block JSONL rows {id, title, text} each, row-aligned to vectors. Row lookup: block = row / meta_rows_per_block; decompress the frame at offsets[block]..offsets[block+1]; take line row % meta_rows_per_block. A top-k query decompresses at most k blocks (~0.2 ms each on-device).
meta.offsets — little-endian u64 array of n_blocks + 1 byte offsets into meta.zst (last entry = file size)
manifest.json — dim, count, quant, scale, meta_codec, meta_rows_per_block, url_template, dataset (with dump date), granularity, model + prompt provenance

Rows carry no url: it is reconstructed as url_template.format(title=percent-encoded title, spaces → underscores). Packs may add fields the reader ignores if unused — e.g. wikibooks rows carry a section, appended to the reconstructed URL as a # anchor.

Embedding model

Documents are embedded with EmbeddingGemma-300M Q8_0 (ggml-org/embeddinggemma-300M-GGUF) via llama.cpp — the same model artifact Ensu loads on-device for query embedding, so document and query vectors come from a bit-identical model.

Document prompt: title: {title} | text: {text}
Query prompt: task: search result | query: {query}
Mean pooling, L2 normalization
Matryoshka truncation to 512 dims: keep the first 512 of 768 components, re-normalize. The query side must do the same (truncating an already-normalized 768-dim vector and re-normalizing is equivalent).

Both prompts and the truncation are recorded in each pack's manifest.json.

License

Scripts: MIT
Wikipedia & Wikibooks content: CC BY-SA 4.0 (attribution is satisfied in-app by citing each passage's source URL)

Downloads last month: 572

Total file size:

734 MB