Summary report

The dense, comprehensive analytical dump. Nine sub-sections of statistics across the whole manuscript. Useful for cross-checking anything you saw in another report, or for producing the sort of overall snapshot you'd hand to an editor.

Endpoint: /summary_report — up to 50,000 words per run.

summary-report report card

What it measures

The endpoint returns a nine-section dump of statistics computed over the whole manuscript:

  • Basic stats — total words, total sentences, total paragraphs, estimated reading time. The foundation counts every other section depends on.
  • Readability — every major readability formula (Flesch Reading Ease, Flesch-Kincaid Grade, Coleman-Liau, ARI, SMOG, Dale-Chall, Gunning-Fog) side-by-side so you can see whether they agree. Different formulas weight syllables, characters, and word familiarity differently — agreement means the grade level is solid; disagreement tells you something specific about the shape of your prose (e.g. long words with short sentences produce different formula outputs than short words with long sentences).
  • Sentiment — average emotional tone (positive vs negative) plus a variance number showing how much the tone shifts across the book. A flat sentiment line often means a flat read.
  • Advanced metrics — vocabulary richness (Yule's K, hapax ratio, MTLD), complexity indices, and other measures mostly useful for cross-checking against the headline scores.
  • Sentence variation — how mixed your sentence lengths are. A varied mix reads better than a monotonous flow; this section reports the standard deviation and coefficient of variation of your sentence-length distribution.
  • Showing vs telling — the dramatised (character action, sensing, direct speech) vs. summarised (narration, exposition, filter verbs) ratio. Popular fiction usually sits with showing dominant; literary and reflective prose leans further toward telling.
  • Dialogues — what share of prose is dialogue and how it's structured (turn length, number of distinct speakers, tag-to-line ratio). Popular fiction usually sits 25–50%; literary fiction often less.
  • Vocabulary — richness metrics plus a top-N list of your most-used words. Useful for spotting unintended verbal tics.
  • Power keywords — the most emotionally or thematically loaded words in the manuscript, grouped into categories (violence, love, fear, achievement, movement). A quick way to see what your prose is really about beneath the plot machinery.

The endpoint caps at 50,000 words per run. Longer manuscripts get truncated with a marker; run per-chapter for longer books.

Why it's useful

Because it's the all-in-one snapshot. If you want one card that summarises the manuscript for yourself, a beta reader, an editor, or an agent — headline word/sentence counts, all seven readability formulas, sentiment shape, showing/telling ratio, dialogue share, most-used words, power keywords — this is the one to hand over. Every other report in the panel drills into one dimension; this one gives you the panoramic.

The cross-checking use is where it earns its place. If Style score tells you the prose is amber on complexity but green on readability, the Summary report's readability formulas can adjudicate — do all seven agree the prose is grade 8, or is Flesch-Kincaid saying 8 and Dale-Chall saying 12? If they disagree by more than 2 grades, your prose has an unusual mix of long words and short sentences (or vice versa), and the composite readability score is hiding that.

The power-keywords section is under-appreciated. Your plot may be about "an inheritance dispute", but if the top emotional keywords are "cold", "silence", "letter", "father", "return" — that's what your prose is actually about. Reading the power keywords out loud is one of the fastest ways to check whether the subtext of the book matches your intended theme.

How to read it

Skim the sub-sections top to bottom. Each block is a compact key-value list with the metric name on the left and the value (mono, tabular) on the right — nothing to click, everything to scan. The layout is deliberately dense; the report is meant to be read, not interacted with.

Anything that surprises you is worth drilling into via the specific report for that dimension. If the Sentiment block shows near-zero variance and your book is supposed to be a rollercoaster, cross-check against Pacing arc. If Vocabulary shows one word appearing hundreds of times, cross-check against Lexical diversity and Stylometry. If Showing vs telling shows a heavy telling skew and you wanted immersive prose, that's a targeted revision brief.

The Power keywords block renders as a wrap-flow of small chips — each chip a single loaded word from the top-N list. Read them together, not individually. The cluster is the signal, not any single chip.

When to ignore it

Never really — this report is nearly all descriptive statistics and hard to disagree with. If a number surprises you, that's the report doing its job. The only time to actively distrust it is on very short manuscripts (under ~5,000 words) where every metric is dominated by noise, or on documents that mix multiple registers (dialogue-heavy scenes stacked with reflection-heavy scenes) where the averages hide the interesting variance.

Sentiment specifically is worth interpreting cautiously. The analyzer sees words, not tone. Sarcasm, irony, deadpan comedy, and understatement all read as neutral or unexpectedly positive/negative. A dark novel written in a light voice will show as positive-sentiment; treat the number as descriptive of the lexical mood, not the emotional mood the reader experiences.

How to run it

  1. Click Write in the left sidebar and open the document you want to analyze.
  2. Open the right rail: click the Reports button (bar-chart icon) in the editor's top toolbar.
  3. In the rail header, click the Rust tab.
  4. All 8 Rust endpoints run in parallel on tab open — typically completes in 2–5 seconds. There is no card to click; everything renders in one composite panel.
  5. The summary report feeds the four stat tiles at the top of the panel (words, sentences, avg / sentence, long sentences) as the fallback source when /score or /sentencelength are unavailable. Its readability numbers are the fallback for the Readability block just below the stat tiles.
  6. The Summary report has no dedicated section of its own in the current panel — its role is to keep the headline numbers populated when the more specific endpoints fail. If a tile shows a dash and the top-of-panel banner is red, the Summary endpoint failed too.
  • Style score — the composite score view of the same territory
  • Reports — back to the panel overview