sovlm

Sovereign Statistical Language Model. N-grams + Kneser-Ney + BM25 Retrieval + QRNG sampling. No training. No backprop. No GPU. Pure governance.

Part of the SnapKitty Collective sovereign AI stack.

---

Install

pip install -e ".[dev]"
# or
uv pip install -e ".[dev]"

Usage

# Build from your corpus (Lean proofs, Prolog, docs, anything)
sovlm build data/*.lean data/*.pl data/*.md --order 4

# Generate
sovlm generate -p "theorem main : " -m 200

# Stream
sovlm generate -p "The 49th Call is" --stream

# Quantum RNG sampling
sovlm generate -p "lemma " --qrng

# Serve HTTP API
sovlm serve --port 8080

Python API

from sovlm import Generator, build_from_paths
from pathlib import Path

# Build
d, ng, sg, inv, fz = build_from_paths([Path("corpus.md")], ...)

# Generate
gen = Generator(d, kn, sg, fz, inv, blender, sampler)
result = gen.generate("theorem xor_assoc : ", max_tokens=100)

Architecture

Component	Role
Dictionary	Vocabulary sovereignty — you own the tokens
NGramIndex	Local coherence — what follows what
KneserNey	Smoothing — handle unseen contexts
SkipGramIndex	Structural patterns at distance
FuzzyPrefixIndex	LSH near-miss — graceful degradation
InvertedIndex + BM25	Retrieval — what's relevant in corpus
Blender	Arbitration — Markov vs retrieval
Sampler + QRNG	Entropy — unbiased, quantum-seeded choice

SnapKitty Stack Integration

sovlm ──► abjad-machine (SUBLEQ address space)
       ──► sovereign-transformer (plasma gate)
       ──► the-49th-call (Lean 4 verified)
       ──► bob-orchestrator (Byzantine council)

---

Ahmad Ali Parr · SnapKitty Collective · 2026