starfuzz/resolver

A module providing declarative entity resolution matching capabilities. Structured records can be compared using weighted components, outputting matched status and confidence levels (High, Medium, Low) for each candidate.

Types

Confidence

</>

Represents the matching confidence category.

pub type Confidence {
  High
  Medium
  Low
}

Constructors

```
High
```
```
Medium
```
```
Low
```

Resolution

</>

A resolved entity match result.

pub type Resolution(candidate) {
  Resolution(
    candidate: candidate,
    score: Float,
    components: List(rank.ComponentScore),
    confidence: Confidence,
    matched: Bool,
  )
}

Constructors

Resolution(
  candidate: candidate,
  score: Float,
  components: List(rank.ComponentScore),
  confidence: Confidence,
  matched: Bool,
)

Resolver

</>

An entity resolver configuration matching a query to candidates.

pub type Resolver(query, candidate) {
  Resolver(
    scorer: fn(query, candidate) -> List(rank.ComponentScore),
    threshold: Float,
    confidence_thresholds: #(Float, Float),
  )
}

Constructors

Resolver(
  scorer: fn(query, candidate) -> List(rank.ComponentScore),
  threshold: Float,
  confidence_thresholds: #(Float, Float),
)

Values

component

</>

pub fn component(
  name: String,
  weight: Float,
  score: Float,
) -> rank.ComponentScore

Helper function to construct a ComponentScore inside the resolver scorer.

new

</>

pub fn new(
  scorer: fn(query, candidate) -> List(rank.ComponentScore),
) -> Resolver(query, candidate)

Creates a new Resolver with a composite scoring function. Defaults to a threshold of 0.7, medium confidence at 0.75, and high confidence at 0.9.

resolve

</>

pub fn resolve(
  resolver: Resolver(query, candidate),
  query: query,
  candidates: List(candidate),
) -> Result(List(Resolution(candidate)), rank.RankError)

Resolves a query entity against a list of candidates. Returns resolution records sorted descending by composite score, with stable ordering.

with_confidence_thresholds

</>

pub fn with_confidence_thresholds(
  resolver: Resolver(q, c),
  medium: Float,
  high: Float,
) -> Resolver(q, c)

Configures the thresholds for Medium and High confidence tiers.

with_threshold

</>

pub fn with_threshold(
  resolver: Resolver(q, c),
  threshold: Float,
) -> Resolver(q, c)

Configures the minimum score threshold to consider a record “matched”.