examples/

Details

Valid go.mod file
Redistributable license
Tagged version
Stable version
Learn more about best practices

Repository

github.com/dotcommander/reliquary

Links

Open Source Insights

README ¶

Reliquary Examples and Demos

Run any example from the repository root:

GOWORK=off go run ./examples/<example-dir>

High-Level Facade

quickstart-rag: Zero-config in-memory hybrid retrieval facade (documents -> Ingest -> Search).
metadata-filtering: Structured metadata filtering (WithFilter) by JSON scalars and reserved fields (filename, document_id).
search-pipeline: Full search pipeline with RRF fusion, external reranking, MMR diversification, and typed explanations.
rag-ingest-retrieve: Manual hybrid retrieval pipeline with BM25, vector search, and MMR reranking.
retrieval-calibration-tune: Retrieval evaluation and hybrid weight calibration against gold relevance sets.

Index Providers (Storage Backends)

sqlite-index: SQLite index adapter (adapter/sqlite) with FTS5 lexical BM25 matching, vector cosine distance, and schema migration.
postgres-index: PostgreSQL index adapter (adapter/postgres) with pgvector and caller-owned pgxpool connection management.

Embedder Providers & Caching

provider-openai-wiring: OpenAI embedding adapter (adapter/openai) configuration with dimension bounds.
provider-ollama-wiring: Ollama embedding adapter (adapter/ollama) with native /api/embed batch vectors and endpoint fallback.
embed-server-quickstart: Local embeddings server adapter with nomic task prefixes and Matryoshka dimensions.
embedding-cache: Transparent embedding cache decorator (embedding/cache + inmem.Store) eliminating redundant provider calls.

Judgment Providers (Evaluation & Reasoning)

judgment-quickstart: Typed Choice/Score/Noul evaluation core with TTL memory cache and reflex-to-deliberate escalation.
judgment-semantic-quickstart: Semantic judgment memory scanning cosine similarity over vectors with local embedding models.
provider-jev-judge: Jev typed evaluation adapter (adapter/jev) mapping Choice, Score, and Noul primitives.
provider-structjudge-wiring: Schema-forced structured LLM judge adapter (adapter/structjudge) with JSON schema completion validation.

Retrieval, Chunking, & Ingest Pipelines

filesystem-ingest: Batch directory tree scanner (pipeline/ingest/fs) with file filtering and pipeline streaming into Reliquary.
chunking-strategies: HeadingAware, SentenceBoundary, SmartBoundary, and Recursive text chunking strategies.
dedup-pipeline: Ingestion deduplication using NormalizedHash (exact/whitespace) and SimHash (near-duplicate clustering).
vector-clustering-pq: K-Means vector clustering (vector/clustering) and Product Quantization (vector/pq) vector compression.

Directories ¶

?	: This menu
/	: Search site
f or F	: Jump to
y or Y	: Canonical URL

Path	Synopsis
chunking-strategies Command chunking-strategies demonstrates Reliquary's chunking strategies (HeadingAware, SentenceBoundary, SmartBoundary, and Recursive).	Command chunking-strategies demonstrates Reliquary's chunking strategies (HeadingAware, SentenceBoundary, SmartBoundary, and Recursive).
dedup-pipeline Command dedup-pipeline demonstrates Reliquary's near-duplicate detection and canonical text normalization primitives.	Command dedup-pipeline demonstrates Reliquary's near-duplicate detection and canonical text normalization primitives.
embed-server-quickstart Command embed-server-quickstart wires a local OpenAI-compatible embedding server into the reliquary facade.	Command embed-server-quickstart wires a local OpenAI-compatible embedding server into the reliquary facade.
embedding-cache Command embedding-cache demonstrates decorating an embedding.Embedder with Reliquary's transparent embedding cache (embedding/cache + inmem.Store).	Command embedding-cache demonstrates decorating an embedding.Embedder with Reliquary's transparent embedding cache (embedding/cache + inmem.Store).
filesystem-ingest Command filesystem-ingest demonstrates scanning and ingesting a directory tree of documents into Reliquary using pipeline/ingest/fs.	Command filesystem-ingest demonstrates scanning and ingesting a directory tree of documents into Reliquary using pipeline/ingest/fs.
internal
examplekit Package examplekit holds demo-only helpers for examples.	Package examplekit holds demo-only helpers for examples.
judgment-quickstart Command judgment-quickstart wires the Embed→Judge stack end to end.	Command judgment-quickstart wires the Embed→Judge stack end to end.
judgment-semantic-quickstart Command judgment-semantic-quickstart wires semantic judgment memory over a local OpenAI-compatible embedding server.	Command judgment-semantic-quickstart wires semantic judgment memory over a local OpenAI-compatible embedding server.
metadata-filtering Command metadata-filtering demonstrates Reliquary's structured metadata filtering during hybrid search.	Command metadata-filtering demonstrates Reliquary's structured metadata filtering during hybrid search.
postgres-index
provider-jev-judge Command provider-jev-judge demonstrates adapting a typed Jev evaluation client to Reliquary's judgment contracts (Choice, Score, and Noul).	Command provider-jev-judge demonstrates adapting a typed Jev evaluation client to Reliquary's judgment contracts (Choice, Score, and Noul).
provider-ollama-wiring Command provider-ollama-wiring demonstrates configuring and wiring the Ollama embedding adapter into Reliquary.	Command provider-ollama-wiring demonstrates configuring and wiring the Ollama embedding adapter into Reliquary.
provider-openai-wiring
provider-structjudge-wiring Command provider-structjudge-wiring demonstrates adapting a structured-output LLM completion client (e.g.	Command provider-structjudge-wiring demonstrates adapting a structured-output LLM completion client (e.g.
quickstart-rag Command quickstart-rag is the short, copyable retrieval path.	Command quickstart-rag is the short, copyable retrieval path.
rag-ingest-retrieve Command rag-ingest-retrieve demonstrates an end-to-end retrieval pipeline assembled from reliquary packages the way a real consumer would:	Command rag-ingest-retrieve demonstrates an end-to-end retrieval pipeline assembled from reliquary packages the way a real consumer would:
retrieval-calibration-tune Command retrieval-calibration-tune demonstrates the retrieval quality-control layer: captured runs, threshold checks, score-reference calibration, and deterministic weight tuning.	Command retrieval-calibration-tune demonstrates the retrieval quality-control layer: captured runs, threshold checks, score-reference calibration, and deterministic weight tuning.
search-pipeline
sqlite-index Command sqlite-index demonstrates Reliquary's SQLite candidate index adapter.	Command sqlite-index demonstrates Reliquary's SQLite candidate index adapter.
vector-clustering-pq Command vector-clustering-pq demonstrates Reliquary's vector clustering and Product Quantization (PQ) compression primitives.	Command vector-clustering-pq demonstrates Reliquary's vector clustering and Product Quantization (PQ) compression primitives.