Learn the controls and knobs

The main controls are embedding model and dimension; similarity metric; index type; HNSW search breadth; candidate top-k; metadata filters; hybrid keyword weight; quantization; replication; and refresh strategy. Each should be treated as a hypothesis: “changing X will move metric Y under workload Z.” Change one family of controls at a time, record the version, and compare against a baseline.

Start with controls that bound harm—permissions, limits, split integrity, or validation—before controls that polish average quality. Use a small sweep instead of one lucky setting. A setting that wins on one example can lose on a different length, language, class, tenant, or traffic pattern. Keep defaults explicit in configuration so a provider or library update cannot silently redefine the experiment.

A useful control sheet has five columns: control, current value, predicted benefit, predicted cost, and rollback trigger. Fill it using the tradeoff below rather than intuition alone: Approximate indexes trade a little recall for major speed gains. Higher search breadth improves recall but costs latency. Filtering after vector search can miss eligible results; filtering during search needs index support. A managed service reduces operations while raising cost and portability concerns.

A catalog embeds product descriptions and stores category, region, and tenant metadata. “Waterproof trail shoe” retrieves semantically related items, but a tenant filter is applied inside the query and a lexical component preserves exact model-number matches.

The worked number is for normalized vectors a and b, cosine similarity is a·b; vectors [1,0] and [0.8,0.6] have similarity 0.8 because both have length 1. State the unit and denominator whenever you report it. A percentage without a denominator can conceal a tiny sample; a latency without a percentile can conceal slow users; a similarity score without a labeled task can conceal irrelevant neighbors. Compare the observed value with a threshold chosen before seeing the final test result.

Now test the tempting shortcut. Suppose the team optimizes only the most visible metric. The result may look better while the system becomes less trustworthy. The reason is concrete: Approximate indexes trade a little recall for major speed gains. Higher search breadth improves recall but costs latency. Filtering after vector search can miss eligible results; filtering during search needs index support. A managed service reduces operations while raising cost and portability concerns. This is why the decision record must include both the intended gain and the tolerated regression. If the tolerated regression is unknown, the change is not ready for a consequential workflow.

• Prefer a measured baseline over a persuasive demo.
• Keep versions, inputs, and thresholds reproducible.
• Separate syntactic success from semantic correctness and authorization.
• Escalate or abstain when evidence falls outside the contract.
• Re-evaluate when data, traffic, models, providers, or user goals change.

These rules turn the topic into an engineering decision rather than a slogan. They also make disagreement productive: another person can challenge the assumptions, rerun the evaluation, and reach a documented conclusion.

Select three plausible values for one control while freezing the others. Run the same representative cases at every value and record task quality, p95 latency, unit cost, and failure count. Do not pick the winner from the average alone: inspect the worst case and important slices. Next, repeat one run to estimate natural variation. If the difference between two settings is smaller than run-to-run variation, the evidence does not support declaring a winner. Save the configuration beside the results so the experiment is reproducible after a model or library upgrade.