Select a report
Select a query and report from the navigation to begin.
System Ranking
🔑 (top-right) to set your OpenRouter API keyv1.0 — 2026-06-24
Before creating nuggets, you need an OpenRouter API key for LLM-powered features.
🔑 in the top-right cornerThe key is stored in your browser's localStorage and persists across sessions.
Enter your name in the Username field (top-right). This is recorded as the creator of any nuggets you create.
In Creation phase:
You can also select text from the Query box at the top.
Use the Freetext notes area to describe what this nugget should capture. This helps the LLM generate a better nugget question.
Click Canonicalize to have the LLM generate a formal nugget question from your spans and notes.
| Category | Meaning |
|---|---|
| Must Have | Critical information - reports lacking this are poor |
| Should Have | Important but not critical |
| Avoid | Information that should NOT appear (e.g., hallucinations) |
Click Check Impact to preview how this nugget grades across all reports:
Click Commit to save the nugget permanently. It appears in the Nuggets panel on the right.
| State | Action |
|---|---|
| Find Quote (green) | Click to extract quote from current report |
| Show Quote (blue) | Click to highlight the quote in the source |
| Hide Quote (blue) | Click to remove highlight |
| No Quote (gray) | No supporting quote found |
The Grade All button (top-right) grades all enabled nuggets against all reports for the current query.
Switch to QC phase using the tab at the top.
| Category | Range | Default | Effect |
|---|---|---|---|
| Must Have | 0-10 | 1.0 | Higher = more important in ranking |
| Should Have | 0-10 | 1.0 | Higher = more important in ranking |
| Avoid | -10-0 | -1.0 | Negative = penalizes reports containing this |
Use the checkbox next to each nugget to include or exclude it from ranking calculations.
Click the dot button next to a nugget to "solo" it:
| Column | Meaning |
|---|---|
| # | Rank position |
| System | Report/system name |
| Nug | Satisfied/Total nuggets |
| Avg | Average grade (0-5) |
| Cov | Coverage % (grade 4+ count) |
| Score | Weighted score |
Switch to Observe phase for read-only analysis.
Click All Queries / This Query button to switch between single-query and cross-query rankings.
| Metric | Meaning | Ideal |
|---|---|---|
| Discriminative | Covered by 10-80% of systems | High count |
| Universal | Covered by >80% of systems | May be too easy |
| Hard | Covered by <10% of systems | May need refinement |
Good nugget bank: Mostly discriminative nuggets, few universal/hard extremes.
Click Export to download a JSONL file containing all nuggets, grades, and metadata.
The exported file integrates with nugget-based judges and other evaluation tooling.