search

package
v1.2.2 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Mar 26, 2026 License: Apache-2.0 Imports: 6 Imported by: 0

Documentation

Overview

Package search provides search domain types for hybrid code retrieval.

Index

Constants

View Source 
const MaxSnippetIDsPerFind = 1000

MaxSnippetIDsPerFind is the maximum number of snippet IDs per Find call, keeping IN-clause bind parameters within the PostgreSQL 65535 limit.

Variables

This section is empty.

Functions

func EmbeddingFrom 

func EmbeddingFrom(q repository.Query) ([]float64, bool)

EmbeddingFrom extracts the embedding vector from a built query.

func QueryFrom 

func QueryFrom(q repository.Query) (string, bool)

QueryFrom extracts the search query text from a built query.

func SnippetIDsFrom 

func SnippetIDsFrom(q repository.Query) []string

SnippetIDsFrom extracts snippet IDs from conditions on a built query.

func WithEmbedding 

func WithEmbedding(embedding []float64) repository.Option

WithEmbedding passes a pre-computed embedding vector through options.

func WithFilters 

func WithFilters(filters Filters) repository.Option

WithFilters passes search filters through the option system.

func WithQuery 

func WithQuery(query string) repository.Option

WithQuery passes a search query string through options.

func WithSnippetID 

func WithSnippetID(id string) repository.Option

WithSnippetID filters by a single snippet ID.

func WithSnippetIDs 

func WithSnippetIDs(ids []string) repository.Option

WithSnippetIDs filters by multiple snippet IDs.

Types

type BM25Store 

type BM25Store interface {
	// Index adds documents to the BM25 index.
	Index(ctx context.Context, request IndexRequest) error

	// Find performs BM25 keyword search using options.
	// Query text must be passed via WithQuery.
	Find(ctx context.Context, options ...repository.Option) ([]Result, error)

	// DeleteBy removes documents matching the given options.
	DeleteBy(ctx context.Context, options ...repository.Option) error
}

BM25Store defines operations for BM25 full-text search indexing.

type BatchError 

type BatchError func(batchStart, batchEnd int, err error)

BatchError is called when a batch fails during indexing. batchStart and batchEnd are the document offsets of the failed batch; err is the upstream error (e.g. HTTP 429, timeout, auth failure).

type BatchProgress 

type BatchProgress func(completed, total int)

BatchProgress is called after each batch completes during indexing. completed is the running total of documents processed so far; total is the overall number of documents to embed.

type Document 

type Document struct {
	// contains filtered or unexported fields
}

Document represents a generic document for indexing.

func NewDocument 

func NewDocument(snippetID, text string) Document

NewDocument creates a new Document.

func (Document) SnippetID 

func (d Document) SnippetID() string

SnippetID returns the snippet ID.

func (Document) Text 

func (d Document) Text() string

Text returns the document text.

type Embedder 

type Embedder interface {
	Embed(ctx context.Context, texts []string) ([][]float64, error)
}

Embedder converts text into embedding vectors.

type Embedding 

type Embedding struct {
	// contains filtered or unexported fields
}

Embedding represents a snippet with its pre-computed embedding vector.

func NewEmbedding 

func NewEmbedding(snippetID string, vector []float64) Embedding

NewEmbedding creates a new Embedding value object.

func (Embedding) SnippetID 

func (e Embedding) SnippetID() string

SnippetID returns the snippet identifier.

func (Embedding) Vector 

func (e Embedding) Vector() []float64

Vector returns a defensive copy of the embedding vector.

type EmbeddingStore 

type EmbeddingStore interface {
	// SaveAll persists pre-computed embeddings.
	SaveAll(ctx context.Context, embeddings []Embedding) error

	// Find retrieves embeddings matching the given options.
	Find(ctx context.Context, options ...repository.Option) ([]Embedding, error)

	// Search performs vector similarity search using options.
	// Embedding must be passed via WithEmbedding.
	Search(ctx context.Context, options ...repository.Option) ([]Result, error)

	// Exists checks whether any row matches the given options.
	Exists(ctx context.Context, options ...repository.Option) (bool, error)

	// DeleteBy removes documents matching the given options.
	DeleteBy(ctx context.Context, options ...repository.Option) error
}

EmbeddingStore defines persistence operations for vector embeddings.

type Filters 

type Filters struct {
	// contains filtered or unexported fields
}

Filters represents filters for snippet search.

func FiltersFrom 

func FiltersFrom(q repository.Query) (Filters, bool)

FiltersFrom extracts search filters from a built query.

func NewFilters 

func NewFilters(opts ...FiltersOption) Filters

NewFilters creates a new Filters with options.

func (Filters) Authors 

func (f Filters) Authors() []string

Authors returns the author filter.

func (Filters) CommitSHAs 

func (f Filters) CommitSHAs() []string

CommitSHAs returns the commit SHA filter.

func (Filters) CreatedAfter 

func (f Filters) CreatedAfter() time.Time

CreatedAfter returns the created after filter.

func (Filters) CreatedBefore 

func (f Filters) CreatedBefore() time.Time

CreatedBefore returns the created before filter.

func (Filters) EnrichmentSubtypes 

func (f Filters) EnrichmentSubtypes() []string

EnrichmentSubtypes returns the enrichment subtypes filter.

func (Filters) EnrichmentTypes 

func (f Filters) EnrichmentTypes() []string

EnrichmentTypes returns the enrichment types filter.

func (Filters) FilePaths 

func (f Filters) FilePaths() []string

FilePaths returns the file path filter.

func (Filters) IsEmpty 

func (f Filters) IsEmpty() bool

IsEmpty returns true if no filters are set.

func (Filters) Languages 

func (f Filters) Languages() []string

Languages returns the language filter.

func (Filters) SourceRepos 

func (f Filters) SourceRepos() []int64

SourceRepos returns the source repository filter.

func (Filters) With added in v1.1.7 

func (f Filters) With(opts ...FiltersOption) Filters

With returns a copy of f with the given options applied.

type FiltersOption 

type FiltersOption func(*Filters)

FiltersOption is a functional option for Filters.

func WithAuthors 

func WithAuthors(authors []string) FiltersOption

WithAuthors sets the author filter.

func WithCommitSHAs 

func WithCommitSHAs(shas []string) FiltersOption

WithCommitSHAs sets the commit SHA filter.

func WithCreatedAfter 

func WithCreatedAfter(t time.Time) FiltersOption

WithCreatedAfter sets the created after filter.

func WithCreatedBefore 

func WithCreatedBefore(t time.Time) FiltersOption

WithCreatedBefore sets the created before filter.

func WithEnrichmentSubtypes 

func WithEnrichmentSubtypes(subtypes []string) FiltersOption

WithEnrichmentSubtypes sets the enrichment subtypes filter.

func WithEnrichmentTypes 

func WithEnrichmentTypes(types []string) FiltersOption

WithEnrichmentTypes sets the enrichment types filter.

func WithFilePaths 

func WithFilePaths(paths []string) FiltersOption

WithFilePaths sets the file path filter.

func WithLanguages 

func WithLanguages(languages []string) FiltersOption

WithLanguages sets the language filter.

func WithSourceRepos 

func WithSourceRepos(repos []int64) FiltersOption

WithSourceRepos sets the source repository filter.

type Fusion 

type Fusion struct {
	// contains filtered or unexported fields
}

Fusion combines results from multiple search methods using Reciprocal Rank Fusion (RRF) algorithm.

func NewFusion 

func NewFusion() Fusion

NewFusion creates a Fusion with the default RRF constant.

func NewFusionWithK 

func NewFusionWithK(k float64) Fusion

NewFusionWithK creates a Fusion with a custom RRF constant.

func (Fusion) Fuse 

func (f Fusion) Fuse(lists ...[]FusionRequest) []FusionResult

Fuse combines multiple ranked result lists using Reciprocal Rank Fusion. Each input list should be sorted by score (descending). Returns a fused list sorted by combined RRF score.

func (Fusion) FuseTopK 

func (f Fusion) FuseTopK(topK int, lists ...[]FusionRequest) []FusionResult

FuseTopK combines multiple ranked result lists and returns the top K results.

func (Fusion) K 

func (f Fusion) K() float64

K returns the RRF constant used by this service.

type FusionRequest 

type FusionRequest struct {
	// contains filtered or unexported fields
}

FusionRequest represents a fusion request input.

func NewFusionRequest 

func NewFusionRequest(id string, score float64) FusionRequest

NewFusionRequest creates a new FusionRequest.

func (FusionRequest) ID 

func (f FusionRequest) ID() string

ID returns the document ID.

func (FusionRequest) Score 

func (f FusionRequest) Score() float64

Score returns the score.

type FusionResult 

type FusionResult struct {
	// contains filtered or unexported fields
}

FusionResult represents a fusion result.

func NewFusionResult 

func NewFusionResult(id string, score float64, originalScores []float64) FusionResult

NewFusionResult creates a new FusionResult.

func (FusionResult) ID 

func (f FusionResult) ID() string

ID returns the document ID.

func (FusionResult) OriginalScores 

func (f FusionResult) OriginalScores() []float64

OriginalScores returns the original scores from each search method.

func (FusionResult) Score 

func (f FusionResult) Score() float64

Score returns the fused score.

type IndexConfig 

type IndexConfig struct {
	// contains filtered or unexported fields
}

IndexConfig holds the resolved configuration for an Index call.

func NewIndexConfig 

func NewIndexConfig(opts ...IndexOption) IndexConfig

NewIndexConfig applies all options and returns the resolved config.

func (IndexConfig) BatchError 

func (c IndexConfig) BatchError() BatchError

BatchError returns the batch error callback, or nil if none was set.

func (IndexConfig) MaxFailureRate 

func (c IndexConfig) MaxFailureRate() float64

MaxFailureRate returns the maximum fraction of batches that may fail before the Index call returns an error. Default is 0.05 (5%).

func (IndexConfig) Progress 

func (c IndexConfig) Progress() BatchProgress

Progress returns the progress callback, or nil if none was set.

type IndexOption 

type IndexOption func(*IndexConfig)

IndexOption configures the behaviour of an Index call.

func WithBatchError 

func WithBatchError(fn BatchError) IndexOption

WithBatchError registers a callback that is invoked when an individual batch fails during indexing. This allows callers to log each upstream error (HTTP status, timeout, etc.) as it occurs.

func WithMaxFailureRate 

func WithMaxFailureRate(rate float64) IndexOption

WithMaxFailureRate sets the maximum fraction of batches that may fail before the Index call returns an error. The rate is clamped to [0, 1]. A rate of 0 means any single batch failure is fatal.

func WithProgress 

func WithProgress(fn BatchProgress) IndexOption

WithProgress registers a callback that is invoked after each batch of embeddings is generated and saved.

type IndexRequest 

type IndexRequest struct {
	// contains filtered or unexported fields
}

IndexRequest represents a generic indexing request.

func NewIndexRequest 

func NewIndexRequest(documents []Document) IndexRequest

NewIndexRequest creates a new IndexRequest.

func (IndexRequest) Documents 

func (i IndexRequest) Documents() []Document

Documents returns the documents to index.

type MultiRequest 

type MultiRequest struct {
	// contains filtered or unexported fields
}

MultiRequest represents a multi-modal search request.

func NewMultiRequest 

func NewMultiRequest(
	topK int,
	textQuery, codeQuery string,
	keywords []string,
	filters Filters,
) MultiRequest

NewMultiRequest creates a new MultiRequest.

func (MultiRequest) CodeQuery 

func (m MultiRequest) CodeQuery() string

CodeQuery returns the code query.

func (MultiRequest) Filters 

func (m MultiRequest) Filters() Filters

Filters returns the search filters.

func (MultiRequest) Keywords 

func (m MultiRequest) Keywords() []string

Keywords returns the keywords.

func (MultiRequest) TextQuery 

func (m MultiRequest) TextQuery() string

TextQuery returns the text query.

func (MultiRequest) TopK 

func (m MultiRequest) TopK() int

TopK returns the number of results to return.

type Query 

type Query struct {
	// contains filtered or unexported fields
}

Query represents a snippet search query.

func NewQuery 

func NewQuery(text string, searchType Type, filters Filters, topK int) Query

NewQuery creates a new Query.

func (Query) Filters 

func (q Query) Filters() Filters

Filters returns the search filters.

func (Query) SearchType 

func (q Query) SearchType() Type

SearchType returns the search type.

func (Query) Text 

func (q Query) Text() string

Text returns the query text.

func (Query) TopK 

func (q Query) TopK() int

TopK returns the number of results.

type Result 

type Result struct {
	// contains filtered or unexported fields
}

Result represents a generic search result.

func NewResult 

func NewResult(snippetID string, score float64) Result

NewResult creates a new Result.

func (Result) Score 

func (r Result) Score() float64

Score returns the search score.

func (Result) SnippetID 

func (r Result) SnippetID() string

SnippetID returns the snippet ID.

type TokenBudget 

type TokenBudget struct {
	// contains filtered or unexported fields
}

TokenBudget constrains embedding batches to stay within model token limits. It holds a character budget and a maximum batch size: each batch's total (truncated) text must not exceed maxChars, each batch contains at most maxBatchSize documents, and individual texts are truncated to maxChars.

func DefaultTokenBudget 

func DefaultTokenBudget() TokenBudget

DefaultTokenBudget returns a conservative budget of 16 000 characters (~5 300 tokens at ~3 chars/token), safe for 8 192-token models like text-embedding-3-small.

func NewTokenBudget 

func NewTokenBudget(maxChars int) (TokenBudget, error)

NewTokenBudget creates a TokenBudget with the given character limit. maxChars must be positive.

func (TokenBudget) Batches 

func (b TokenBudget) Batches(documents []Document) [][]Document

Batches partitions documents into groups whose total truncated character count stays within the budget and whose size does not exceed maxBatchSize. A single document whose truncated text still exceeds the character budget is placed alone in its own batch.

func (TokenBudget) Truncate 

func (b TokenBudget) Truncate(text string) string

Truncate returns text capped to the character (rune) limit.

func (TokenBudget) WithMaxBatchSize 

func (b TokenBudget) WithMaxBatchSize(n int) TokenBudget

WithMaxBatchSize returns a new TokenBudget with the given maximum number of documents per batch. Values <= 0 are clamped to 1.

type Type 

type Type string

Type represents the type of search to perform. 

const (
	TypeBM25   Type = "bm25"
	TypeVector Type = "vector"
	TypeHybrid Type = "hybrid"
)

Type values.

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.