bench_cache


Description

{“what”:“Cache of per-turn benchmark results keyed by run, provider, model, arm, and turn.”,“why”:“Without it, repeated benchmark queries would recompute costs and token counts.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:15.146918278+00:00”}

Columns

Column Type Size Nulls Auto Default Children Parents Comments
id bigserial 19 √ nextval('bench_cache_id_seq'::regclass)

{“what”:“Surrogate primary key; unique row identifier.”,“why”:“Without it, no stable reference for updates or deletes.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:15.170993834+00:00”}

run_id uuid 2147483647 null

{“what”:“UUID of the benchmark run this result belongs to.”,“why”:“Without it, results from different runs cannot be separated.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:15.194178671+00:00”}

provider text 2147483647 null

{“what”:“Name of the LLM provider (e.g., OpenAI, Anthropic).”,“why”:“Without it, cost and token metrics cannot be attributed to a vendor.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:15.206958272+00:00”}

model text 2147483647 null

{“what”:“Model identifier as used by the provider (e.g., gpt-4o).”,“why”:“Without it, comparisons across model versions are impossible.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:15.222059993+00:00”}

arm text 2147483647 null

{“what”:“Experiment arm label (e.g., control, variant) for A/B testing.”,“why”:“Without it, results cannot be grouped by experimental condition.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:15.254091713+00:00”}

turn int4 10 null

{“what”:“Conversation turn number, starting at 1.”,“why”:“Without it, per-turn token and cost breakdowns are lost.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:15.284684031+00:00”}

prompt_tokens int8 19 null

{“what”:“Number of tokens in the prompt for this turn.”,“why”:“Without it, prompt cost cannot be computed or audited.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:15.465260864+00:00”}

cached_tokens int8 19 null

{“what”:“Number of prompt tokens served from cache.”,“why”:“Without it, cache savings and effective cost are unknown.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:15.563739783+00:00”}

completion_tokens int8 19 null

{“what”:“Number of tokens generated in the completion.”,“why”:“Without it, completion cost and latency estimates are impossible.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:16.419391809+00:00”}

cost_reported_usd float8 17,17 √ null

{“what”:“Cost in USD as reported by the provider, if available.”,“why”:“Without it, provider-billed cost cannot be compared to estimates.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:16.456865453+00:00”}

cost_est_usd float8 17,17 null

{“what”:“Estimated cost in USD computed from token counts and pricing.”,“why”:“Without it, no cost baseline exists when provider reporting is missing.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:16.490040009+00:00”}

status int4 10 null

{“what”:“Cache entry status code (e.g., 0=valid, 1=stale).”,“why”:“Without it, stale or invalid entries cannot be filtered out.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:16.516495418+00:00”}

ts timestamptz 35,6 now()

{“what”:“Timestamp when the cache entry was created.”,“why”:“Without it, cache freshness and eviction decisions are impossible.”,“by”:“fleet-annotator-worker”,“at”:“2026-09-23T04:04:16.543278138+00:00”}

Indexes

Constraint Name Type Sort Column(s)
bench_cache_pkey Primary key Asc id
bench_cache_run_idx Performance Asc/Asc/Asc/Asc run_id + provider + arm + turn

Relationships