e1bae0c7a0
- Add docs/ folder with English (en/) and Bulgarian (bg/) documentation - Create index.md with language switching and links - English docs: installation, quickstart, architecture, baraql, storage, schema, lsm, btree, vector, graph, fts, columnar, transactions, distributed, protocol, udf, api-binary, api-http, api-websocket - Bulgarian docs: installation, quickstart, architecture, baraql, schema, lsm, btree, vector, graph, fts, transactions, distributed - Update README license to BSD 3-Clause - Add LICENSE file with BSD 3-Clause text
76 lines
1.5 KiB
Markdown
76 lines
1.5 KiB
Markdown
# Vector Search Engine
|
|
|
|
Native HNSW and IVF-PQ indexes for similarity search.
|
|
|
|
## Usage
|
|
|
|
```nim
|
|
import barabadb/vector/engine
|
|
|
|
var idx = newHNSWIndex(dimensions = 128)
|
|
idx.insert(1, @[1.0'f32, 0.0'f32, ...], {"category": "A"}.toTable)
|
|
|
|
# Search
|
|
let results = idx.search(queryVector, k = 10)
|
|
|
|
# With metadata filtering
|
|
let filtered = idx.searchWithFilter(queryVector, k = 10,
|
|
filter = proc(meta: Table[string, string]): bool =
|
|
return meta.getOrDefault("category") == "A")
|
|
```
|
|
|
|
## Index Types
|
|
|
|
### HNSW
|
|
|
|
Hierarchical Navigable Small World graph for approximate nearest neighbor search.
|
|
|
|
```nim
|
|
var hnsw = newHNSWIndex(
|
|
dimensions = 128,
|
|
m = 16, # connections per layer
|
|
efConstruction = 200, # search width during construction
|
|
efSearch = 100 # search width during query
|
|
)
|
|
```
|
|
|
|
### IVF-PQ
|
|
|
|
Inverted File Index with Product Quantization for compression.
|
|
|
|
```nim
|
|
var ivfpq = newIVFPQIndex(
|
|
dimensions = 128,
|
|
numCentroids = 256,
|
|
subQuantizers = 8
|
|
)
|
|
```
|
|
|
|
## Distance Metrics
|
|
|
|
| Metric | Description |
|
|
|--------|-------------|
|
|
| `cosine` | Cosine similarity |
|
|
| `euclidean` | L2 distance |
|
|
| `dotproduct` | Dot product similarity |
|
|
| `manhattan` | L1 distance |
|
|
|
|
## Quantization
|
|
|
|
```nim
|
|
import barabadb/vector/quant
|
|
|
|
# Scalar quantization
|
|
let scalar = scalarQuantize(data, bits = 8)
|
|
|
|
# Product quantization
|
|
let pq = productQuantize(data, subVectors = 8, bits = 8)
|
|
```
|
|
|
|
## SIMD Acceleration
|
|
|
|
```nim
|
|
import barabadb/vector/simd
|
|
|
|
let dist = simdCosineDistance(vec1, vec2)
|
|
``` |