{"schemaVersion":1,"locale":"es","query":"recall","version":"eb87f4b9d56a6e9b115e3f56a5a5da5afbc03fd9ec2a2077e6e9cd6bd49ccf46","total":453,"matched":8,"method":"BM25 passage retrieval with title weighting; lexical relevance, not a confidence or authority score","offset":0,"items":[{"url":"https://lucioduran.com/es/blog/vector-databases-pgvector-qdrant-pinecone","citationUrl":"https://lucioduran.com/es/blog/vector-databases-pgvector-qdrant-pinecone#question-3","title":"Bases de Datos Vectoriales Comparadas: Ejecutando el Mismo Pipeline RAG en pgvector, Qdrant, Pinecone y Weaviate","heading":"¿Cuánto degrada la calidad de búsqueda vectorial con la cuantización?","kind":"blog","contentLanguage":"es","excerpt":"Con cuantización escalar (float32 a int8), el recall baja 1-3%. Con cuantización binaria más rescoring, el recall baja solo 2-5% mientras reduce la memoria 32x. La cuantización de producto puede perder 5-10% de recall pero ahorra más espacio.","contentHash":"44cf38734c57b337484525fb5871f8a3d99ad05e64c6f4ecedab7ec4790641fa","datePublished":"2025-12-02","dateModified":null,"readUrl":"https://lucioduran.com/api/content?path=%2Fes%2Fblog%2Fvector-databases-pgvector-qdrant-pinecone&section=question-3&format=json","score":7.95957,"representations":{"markdown":"https://lucioduran.com/es/blog/vector-databases-pgvector-qdrant-pinecone.md","json":"https://lucioduran.com/es/blog/vector-databases-pgvector-qdrant-pinecone.json"}},{"url":"https://lucioduran.com/es/projects/rag-evaluation-workbench","citationUrl":"https://lucioduran.com/es/projects/rag-evaluation-workbench#results-retrieval","title":"RAG Evaluation Workbench","heading":"Recuperación: 24 consultas con respuesta","kind":"project","contentLanguage":"es","excerpt":"Recall@3 y NDCG@3 sobre 24 preguntas con respuesta, agrupadas en 12 fuentes. Las preguntas sin respuesta se excluyen de estas métricas. Los intervalos del 95% usan bootstrap por grupos.\n\n| Método | Recall@3 | Intervalo del 95% | NDCG@3 | Intervalo del 95% |\n| --- | --- | --- | --- | --- |\n| BM25 | 1.000 | \\[1.000, 1.000\\] | 0.969 | \\[0.923, 1.000\\] |\n| Densa | 1.000 | \\[1.000, 1.000\\] | 1.000 | \\[1.000, 1.000\\] |\n| Híbrida | 1.000 | \\[1.000, 1.000\\] | 1.000 | \\[1.000, 1.000\\] |","contentHash":"84938b2341b6e5e0b43087524ff3a9aa7e4f64213b68e7d08602169ea6940854","datePublished":"2026-10-21","dateModified":"2026-10-21","readUrl":"https://lucioduran.com/api/content?path=%2Fes%2Fprojects%2Frag-evaluation-workbench&section=results-retrieval&format=json","score":6.98497,"representations":{"markdown":"https://lucioduran.com/es/projects/rag-evaluation-workbench.md","json":"https://lucioduran.com/es/projects/rag-evaluation-workbench.json"}},{"url":"https://lucioduran.com/es/blog/rag-pipeline-architecture-chunking-reranking-evaluation","citationUrl":"https://lucioduran.com/es/blog/rag-pipeline-architecture-chunking-reranking-evaluation#nivel-2-calidad-de-generacin-con-ragas","title":"Arquitectura de Pipelines RAG: Estrategias de Chunking, Búsqueda Híbrida, Reranking y Frameworks de Evaluación","heading":"Nivel 2: Calidad de Generación con RAGAS","kind":"blog","contentLanguage":"es","excerpt":"…m ragas.metrics import (\n    faithfulness,\n    answer_relevancy,\n    context_precision,\n    context_recall,\n)\n\nresults = evaluate(\n    dataset=eval_dataset,  # Preguntas + respuestas ground truth + contextos\n    metrics=[\n        faithfulness,        # ¿La respuesta se mantiene fiel al contexto recuperado?\n        answer_relevancy,    # ¿La respuesta es relevante a la pregunta?\n        context_precision,   # ¿Los chunks recuperados son relevantes?\n        context_recall,      # ¿Se recuperaron todos los chunks necesarios?\n    ],\n)\n\nprint(results)\n# {'faithfulness': 0.89, 'answer_relevancy': 0.91,\n#  'context_precision': 0.82, 'context_recall'…","contentHash":"7dc6a6f36da14b3e894053e87c1ad1b02ea00376844adc8d3e85a5528a98d07a","datePublished":"2026-04-16","dateModified":"2026-10-21","readUrl":"https://lucioduran.com/api/content?path=%2Fes%2Fblog%2Frag-pipeline-architecture-chunking-reranking-evaluation&section=nivel-2-calidad-de-generacin-con-ragas&format=json","score":6.68149,"representations":{"markdown":"https://lucioduran.com/es/blog/rag-pipeline-architecture-chunking-reranking-evaluation.md","json":"https://lucioduran.com/es/blog/rag-pipeline-architecture-chunking-reranking-evaluation.json"}},{"url":"https://lucioduran.com/es/projects/vector-search-portability","citationUrl":"https://lucioduran.com/es/projects/vector-search-portability#demonstrates","title":"Vector Search Portability","heading":"Qué demuestra","kind":"project","contentLanguage":"es","excerpt":"…Contratos comparables entre motores con métricas y semántica de filtros explícitas.\n- Evaluación de recall antes y después de cambios en vectores y metadatos.\n- Análisis de costos del adaptador y preservación de mediciones originales y de confirmación.","contentHash":"b40709d6a609b4aa100843c20be451e2ee1fc412570889a557343d8fc9a8aed6","datePublished":"2026-10-21","dateModified":"2026-10-21","readUrl":"https://lucioduran.com/api/content?path=%2Fes%2Fprojects%2Fvector-search-portability&section=demonstrates&format=json","score":6.14595,"representations":{"markdown":"https://lucioduran.com/es/projects/vector-search-portability.md","json":"https://lucioduran.com/es/projects/vector-search-portability.json"}},{"url":"https://lucioduran.com/es/projects","citationUrl":"https://lucioduran.com/es/projects#project-vector-search-portability","title":"Proyectos","heading":"Vector Search Portability","kind":"section","contentLanguage":"es","excerpt":"Comparación real de hnswlib y sqlite-vec con los mismos vectores, filtros y actualizaciones. Recall frente a búsqueda exacta, carga concurrente y pérdida por postfiltrado.\n\nPython · Licencia: MIT\n\n[Leer la ficha](<https://lucioduran.com/es/projects/vector-search-portability>) · [Repositorio](<https://github.com/lucioduran/vector-search-portability>) · [Release v0.1.0](<https://github.com/lucioduran/vector-search-portability/releases/tag/v0.1.0>)","contentHash":"8ccaf394a698cb4355e930f4f975d9130b551a5eae3cd5a49c7877879ee588bc","datePublished":null,"dateModified":null,"readUrl":"https://lucioduran.com/api/content?path=%2Fes%2Fprojects&section=project-vector-search-portability&format=json","score":5.50253,"representations":{"markdown":"https://lucioduran.com/es/projects.md","json":"https://lucioduran.com/es/projects.json"}},{"url":"https://lucioduran.com/es","citationUrl":"https://lucioduran.com/es#project-vector-search-portability","title":"Lucio Durán","heading":"Vector Search Portability","kind":"section","contentLanguage":"es","excerpt":"Comparación real de hnswlib y sqlite-vec con los mismos vectores, filtros y actualizaciones. Recall frente a búsqueda exacta, carga concurrente y pérdida por postfiltrado.\n\nPython · Licencia: MIT\n\n[Leer la ficha](<https://lucioduran.com/es/projects/vector-search-portability>) · [Repositorio](<https://github.com/lucioduran/vector-search-portability>) · [Release v0.1.0](<https://github.com/lucioduran/vector-search-portability/releases/tag/v0.1.0>)","contentHash":"8a2631059a77f3cd0922423102affdb20a4a81c23ece309c311cddf59287065a","datePublished":null,"dateModified":null,"readUrl":"https://lucioduran.com/api/content?path=%2Fes&section=project-vector-search-portability&format=json","score":5.47238,"representations":{"markdown":"https://lucioduran.com/es.md","json":"https://lucioduran.com/es.json"}},{"url":"https://lucioduran.com/es/access","citationUrl":"https://lucioduran.com/es/access#discovery","title":"Acceso al contenido","heading":"Descubrimiento","kind":"section","contentLanguage":"es","excerpt":"…-index?locale=es>)\n- [Búsqueda en el texto completo](<https://lucioduran.com/api/search?locale=es&q=recall>)\n- [llms.txt](<https://lucioduran.com/llms.txt>)\n- [llms-full.txt](<https://lucioduran.com/llms-full.txt>)\n- [OpenAPI](<https://lucioduran.com/.well-known/openapi.json>)\n- [Sitemap](<https://lucioduran.com/sitemap.xml>)\n- [API catalog (RFC 9727)](<https://lucioduran.com/.well-known/api-catalog>)\n- [JSON Schema](<https://lucioduran.com/.well-known/schemas/publication.json>)\n- [Checklist de capacidades](<https://lucioduran.com/api/capabilities?locale=es>)\n- [Inventario del build con hashes](<https://lucioduran.com/api/publication?locale=e…","contentHash":"ac43a8cfbf3282c50b944fcd3338f974c56a908aecf98fcd6b6237ce0c63b8d4","datePublished":null,"dateModified":null,"readUrl":"https://lucioduran.com/api/content?path=%2Fes%2Faccess&section=discovery&format=json","score":4.65595,"representations":{"markdown":"https://lucioduran.com/es/access.md","json":"https://lucioduran.com/es/access.json"}},{"url":"https://lucioduran.com/es/projects/chunking-ablation-suite","citationUrl":"https://lucioduran.com/es/projects/chunking-ablation-suite#results-coverage","title":"Chunking Ablation Suite","heading":"Evidencia completa en evaluación","kind":"project","contentLanguage":"es","excerpt":"…oques cabe en el presupuesto; esa ablación no informa sensibilidad al umbral.\n\nEl análisis desglosa recall de spans, completitud, bytes de texto, payload vectorial y tiempo de construcción por código, tablas y referencias cruzadas. Los tiempos son observaciones secuenciales de un único equipo. [Resultados y análisis](<https://github.com/lucioduran/chunking-ablation-suite/blob/12193ba09bb1081a05d3ea7a043748089aa2330b/docs/results.md>).","contentHash":"00ec7a67c0ebd6d58a9cd1171170ad05585d4e4a0c31dd5b881bd067a02ccdb5","datePublished":"2026-10-21","dateModified":"2026-10-21","readUrl":"https://lucioduran.com/api/content?path=%2Fes%2Fprojects%2Fchunking-ablation-suite&section=results-coverage&format=json","score":3.77577,"representations":{"markdown":"https://lucioduran.com/es/projects/chunking-ablation-suite.md","json":"https://lucioduran.com/es/projects/chunking-ablation-suite.json"}}],"next":null}