leoworks/amazon-review-classifier
leoworks--amazon-review-classifierThis tool calls the Actor "leoworks/amazon-review-classifier" and retrieves its output results. Actor description: Amazon review analysis: complaints, return reasons, praise, sentiment. products = ASINs/URLs (collected by junglee, billed to you ~$6/1k; Free plan 10 reviews/run; ~$6.5/1k all-in), or datasetId from any scraper, or texts. For AliExpress use leoworks/aliexpress-reviews-classifier.
Tool call template
This tool declares no required parameters. Add optional parameters from the input schema as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {},
"name": "leoworks--amazon-review-classifier"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| texts | No | Paste review texts directly (one per line). Use this OR “Reviews dataset”. Example values: ["Stopped charging after two weeks. Returning it.","Fits great but the fabric is see-through when I squat."] | |
| idFields | No | Fields copied unchanged from each input item to the output so you can join results back (e.g. reviewId, productAsin). Leave empty to auto-pick reviewId and the product ASIN. | |
| maxItems | No | Classify at most this many reviews (0 = all). | |
| maxStars | No | For example 3 = classify only 1–3 star reviews (where complaints and returns are). 0 = all ratings. Reviews without a star rating are kept. | |
| products | No | One product per line, up to 20: the 10-character ASIN (for example B0CWXNS552) or the product page URL (…/dp/ASIN; its marketplace is taken from the URL). The Actor then collects the reviews and classifies them in the same run. [Description truncated] | |
| waitSecs | No | Max seconds (0–45, default 30) to cap the wait for the Actor run to reach terminal state. For long-running Actors the response returns at the cap with the current run status; follow `nextStep` to poll via get-actor-run. Set to 0 to fire-and-forget. | |
| collector | No | The Amazon reviews scraper that collects for you. The default (junglee) returned and classified reviews in our test on amazon.com. The other is about six times cheaper but returned nothing for amazon.com in the same test — try it for other marketplaces. On the Free Apify plan, junglee returns at most 10 reviews per run (its own limit). Prices are the collectors' own and can change. Example values: "junglee/amazon-reviews-scraper" | junglee/amazon-reviews-scraper |
| datasetId | No | An Apify dataset containing Amazon reviews — for example the output of an Amazon reviews scraper. Use this OR “Products” (not both) OR “Review texts”. | |
| dateField | No | Leave empty to auto-detect (date, reviewDate, reviewedIn, …). Used for the monthly trend, the last-90-days comparison and the date filter. | |
| textField | No | Field holding the review text. Leave empty to auto-detect (reviewDescription, reviewText, text, body, …). | |
| outputMode | No | Minimal mode returns only label keys — smaller and easier to aggregate. Example values: "full" | full |
| photoField | No | Field with the picture addresses of each review. Auto-detected (reviewImages, reviewImageUrls, images, imageUrls, photos); set it only if your scraper uses another name. | |
| collectSort | No | Which reviews to collect first. Newest first is the least biased sample of what buyers say now. Example values: "recent" | recent |
| healthCheck | No | Internal: fail the run when results look degraded (used by the developer's scheduled checks). | |
| marketplace | No | Which Amazon site the ASINs in “Products” belong to. A product URL carries its own marketplace and overrides this. Example values: "com" | com |
| ratingField | No | Leave empty to auto-detect (ratingScore, rating, stars, …). | |
| customLabels | No | Up to 10 extra yes/no labels (each up to 200 characters) in plain language, e.g. “mentions the battery”, “compares with another brand”. Each gets a probability. | |
| photoJudging | No | Adds `photoEvidence` (supports_complaint_product, supports_complaint_packaging, damage_seen_text_silent, no_evidence, not_product_photo) and puts photo-backed product defects, then photo-backed packaging damage, at the top of ACTION_QUEUE.csv. "Only reviews with a complaint or 3 stars or fewer" judges far fewer photos and is the recommended setting. Example values: "off" | off |
| photoMaxSide | No | Longest side in pixels requested from the photo server (512, 768 or 1024; other values are rounded). Amazon review photos arrive as 88 px thumbnails, so a larger version is requested. Photos under 280 pixels on the short side are skipped. Example values: 768 | |
| variantField | No | Leave empty to auto-detect (variant, variantName, …). Used for the worst-variants breakdown. | |
| verifiedOnly | No | Skip reviews not marked as a verified purchase (needs a field such as `isVerified`). | |
| maxConcurrency | No | Reviews classified in parallel. Example values: 10 | |
| extraTextFields | No | Fields to prepend to the text (up to 5). Leave empty to add the review headline automatically (reviewTitle). | |
| reviewsNewerThan | No | Classify only reviews published after this date — a date like `2026-01-01` or a period like `90 days`, `6 months`, `1 year`. Needs a review date in the dataset (e.g. `date`). Leave empty for all. | |
| summaryGroupField | No | Field in your dataset that identifies the product (e.g. `productAsin`). The run then saves a REPORT record with, per product: complaint rate and complaint mix, sentiment shares, average stars, share of reviews with a return or refund and 3 example complaints. Leave empty to detect it automatically. No extra charge. | |
| complaintThreshold | No | Minimum probability (0–1) for the strongest complaint type, the return-intent flag or a custom label to be reported. Example values: 0.5 | |
| maxPhotosPerReview | No | The first photos of each review. With "All reviews that have photos", at most 2 photos per review are judged. Example values: 2 | |
| secondaryThreshold | No | Minimum probability (0–1) for any further complaint type after the strongest one. Product complaints overlap (a weak item can also feel uncomfortable), so side labels need to be surer. Lower it to see more labels per review. Example values: 0.8 | |
| maxReviewsPerProduct | No | Classify at most this many reviews per product (ASIN), in dataset order. 0 = no limit. | |
| collectReviewsPerProduct | No | How many reviews to collect for each product (1 to 100). You pay the collector for each one, so a lower number is cheaper; “Max reviews per product” below limits only how many are classified. Example values: 50 |
Output Schema
| Name | Required | Description | Default |
|---|---|---|---|
| tip | No | Advisory guidance RAG Web Browser wrote to its key-value store under the reserved "TIP" key | |
| runId | Yes | Actor run ID | |
| stats | No | Run statistics | |
| status | Yes | Run status: READY | RUNNING | TIMING-OUT | TIMED-OUT | ABORTING | ABORTED | SUCCEEDED | FAILED | |
| actorId | Yes | Stable Apify Actor ID from the run record | |
| summary | Yes | Past-tense summary of the run state | |
| exitCode | No | Actor process exit code; populated for terminal states (especially FAILED) | |
| nextStep | Yes | One primary follow-up action with identifiers interpolated | |
| storages | Yes | Dataset and key-value store metadata, keyed by alias. "default" is always the primary entry. | |
| actorName | No | "username/actor-name" | |
| startedAt | No | ISO timestamp when the run started | |
| finishedAt | No | ISO timestamp when the run finished (terminal states only) | |
| statusMessage | No | Pass-through from Apify run.statusMessage | |
| apifyConsoleUrl | No | Personalized Apify Console link to the run; present only for Console sessions |