Feedback
Every time you disagree with a score, it lands here. This is the record of where your judgment and SLATE's diverge — material for improving your role definition and target set.
Costs
Every extract, compare, and derive call, logged from api.js — the one place all three go through. The summary cards below are lifetime totals (never trimmed); the table is the most recent 500 calls.
| Time | Kind | Profile | Model | Input | Output | Cached | Write (5m) | Write (1h) | Candidate | Cost | Duration | Outcome |
|---|
Settings
Scoring
Auto-score
Current source batch
Free-text label for the boolean search string producing your current leads. Automatically attached to every profile you save until you change it, and shown as the source on the Prospects screen. Required — you can't save a profile without one.
Exclusion rules
Optional hard dealbreakers, checked before any AI call — instant, free, deterministic. This is rubric.json and ships empty; add a rule here only for something that should disqualify a candidate no matter what role or context says, e.g. {"id": "excl-example", "name": "No management experience", "field": "management_experience", "operator": "is_false"} — field names must match the extraction schema in extract.js. Rules switched off on a role's Overview move to disabled_exclusions and aren't evaluated. Fit itself isn't scored here; that happens in the Role & Targets comparison.
Saved list
0 profile(s) saved.
Model pricing
USD per million tokens, set in pricing.js from Anthropic's published per-MTok rates — re-check that file against the current pricing page if a model's rate changes. Cache write is split into 5-minute and 1-hour TTL rates (1.25x and 2x base input respectively) since the API bills them differently. "Last verified" is set in code, not here.
Cost log debug data
When on, stores the full request and response for the 20 most recent API calls, viewable by expanding a row on the Costs screen. Off by default. Turning this off clears everything already stored, not just future calls.
Extraction cache
Extraction (scrape → structured JSON) is cached by the scraped content itself, separately from the Stage 2 score cache — a role or target-set change never empties this, only a change to the extraction prompt or the scraping selectors does.
0 cached extraction(s). Extract prompt version: . Selectors version: . Extract model: (read-only — set in extract.js).