Build the first working tenant pgvector index

Build only what the claim requires: query returns only the requesting tenant's chunks at a pinned index version. Prefer boring, deterministic code over frameworks you cannot yet explain. Run the path twice; identical output on this fixture is a feature, not a lack of creativity.

sql='SELECT chunk_id FROM chunks WHERE tenant_id=%s AND index_version=%s ORDER BY embedding <=> %s LIMIT 1' print({'sql':sql,'params':['cafe-a','v3','[query_vec]']})

Expected evidence: **parameterized tenant+version query**. Read each printed intermediate as part of the argument that the path works—not as decoration.

Narrate the journey from raw input to result for a single example from two tenants with overlapping vocabulary FAQ chunks. If you cannot name an intermediate, the implementation is still too opaque for this lab. Only after this path is solid should you generalize data sources or UI.

While implementing the tenant pgvector index, print or log at least three intermediates that map to the claim (query returns only the requesting tenant's chunks at a pinned index version). Good intermediates are values a teammate could recompute with a calculator or diff. Bad intermediates are framework traces you cannot explain.

Re-run with two tenants with overlapping vocabulary FAQ chunks twice. If the second run differs, either the path is nondeterministic (document the seed) or you have hidden global state—both are lab bugs until named.

Stop adding features once the path supports replace an in-memory list with a versioned vector index that cannot leak across tenants. Extra UI, extra tools, or extra models belong in later chapters. The mastery bar for this page is simply: a deterministic end-to-end path with intermediates.

At the implementation stage for vector-db-lab, the job is narrower than finishing a product demo. You are creating one progressive evidence piece about two tenants with overlapping vocabulary FAQ chunks that later pages inherit without redefining success. Keep that fixture small enough to inspect by hand, keep outputs copy-pasteable as text, and refuse to narrate this baseline as if it were a production SLA: in-memory cosine search on the same fixtures.

For this page specifically, success looks like a deterministic path with printed intermediates while still centering the user decision to replace an in-memory list with a versioned vector index that cannot leak across tenants. If you cannot point to a file, command, or assertion that proves that for the tenant pgvector index, stay on this page instead of advancing.

Glossary: vector database · Glossary: vector index · Cheatsheet: RAG quality

Previous · Next