1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
# Rust
/target
**/*.rs.bk
# Tauri app
app/node_modules
app/dist
app/src-tauri/target
app/src-tauri/gen
# DuckDB reference clone (not part of this repo)
/duckdb
# OS / editor
.DS_Store
*.log
# Models are SHIPPED in-repo so SteelDB works fully offline out of the box (tagger, SPLADE, model2vec,
# gazetteer, OCR, CLIPSeg, and the tuned default LoRA). These are large binaries — tracked via git-lfs
# (see .gitattributes). Nothing under models/ is ignored. Only scratch training checkpoints are excluded:
train/adapters/
train/*.pt
# runtime index caches written next to a corpus
.steeldb-index
# user LLM-provider config (may contain API keys)
/steeldb.json
~/.steeldb/
# Per-corpus growing-gazetteer overlay (learned entities, local to each scanned dir)
.steeldb-gazetteer.json
# Native inference: quantized model cache (built on first run, not committed)
models/steeldb-qwen3-*.gguf
# Public demo corpus — deterministically regenerable via `gen_pokemon_corpus`
/pokemon_corpus
/benchmark_corpus
# HuggingFace access token — a credential, never commit
hf.token
*.token
# Python (demo Spaces)
__pycache__/
*.pyc
# assembled Space bundles (regenerable via spaces/assemble.sh)
/target/spaces
# Registry / signing credentials
*.key
*.pem
# the stress harness builds against the published crate
stress/target/
stress/Cargo.lock