Retrieval and embeddings
Inspect document ingestion, chunks, embeddings, indexes, retrieval quality, freshness, and grounded evidence.
What you will practice
Map the retrieval pipeline; Inspect source documents; Review chunk boundaries; Read embedding metadata.
Use the worked examples when needed, then test your understanding without them.
- 01Map the retrieval pipelineA retrieval pipeline separates ingestion, chunking, embedding, indexing and querying. Locate where evidence disappears before changing the answer-generation prompt.Practice
- 02Inspect source documentsRetrieved evidence is only as useful as its source coverage and validity. Inspect the original document identity and revision before reasoning from a fragment.Practice
- 03Review chunk boundariesChunk boundaries should preserve enough local context to interpret a passage. Separating a condition from its exception can change the meaning of the retrieved evidence.Practice
- 04Read embedding metadataEmbeddings have a model identity and dimension contract. Query and document vectors must use a compatible embedding setup; equal storage format alone is not enough.Practice
- 05Inspect index statusAn index status must be read with document counts and readiness. A process being up does not guarantee that the index contains the expected records.Practice
- 06Check index freshnessFreshness compares the indexed revision with the source revision or update watermark. Wall-clock uptime does not establish that new documents were ingested.Practice
- 07Review retrieval resultsRetrieval results should be judged for relevance, coverage and source quality. A ranking position is not a guarantee that a passage supports the answer.Practice
- 08Compare similarity scoresSimilarity scores are ranking signals under a specific embedding and metric setup. They are not calibrated probabilities of truth or correctness.Practice
- 09Spot a missing citationA citation must connect a claim to source evidence that actually supports it. Merely including a document identifier somewhere in an answer is not adequate grounding.Practice
- 10Diagnose empty retrievalEmpty retrieval can come from filters, missing documents or an unavailable index. Inspect those boundaries before loosening access restrictions or inventing an answer.Practice
- 11Inspect duplicate chunksDuplicate chunks can crowd a result list and reduce source diversity. Review stable document/chunk identifiers before deciding whether two similar passages are redundant.Practice
- 12Review metadata filtersMetadata filters are part of retrieval correctness and access control. A relevant passage from the wrong tenant is not an acceptable result.Practice
- 13Compare retrieval configurationsCompare retrieval configurations on the same query set, corpus and relevance judgments. Otherwise a score change may come from different evidence rather than a better retriever.Practice
- 14Prepare an index rebuild planA safe index rebuild stages a candidate, validates counts and retrieval, and retains the current index until the candidate is accepted. Do not delete the only working index first.Practice
- 15Build retrieval evidenceA retrieval evidence bundle records query, filters, index and embedding revisions, selected source IDs and failure cases. Redact sensitive query content when sharing outside its boundary.Practice
- 16Retrieval checkpointRetrieval readiness requires fresh, accessible, relevant evidence as well as a working index. An answer should not hide a missing or unsupported evidence boundary.Practice