Review Insight AI

Turn customer reviews into a clear, actionable business report without maintaining fragile website scrapers. Upload existing CSV, JSON, or text reviews and receive sentiment, recurring themes, complaints, requests, praise, and prioritized actions.

What it produces

• Positive, neutral, and negative review counts
• Executive summary
• Ranked complaint, request, praise, question, and other themes
• Theme frequency and prevalence
• Business-impact severity from 1 to 5
• Suggested action for every theme
• Prioritized action list
• Short anonymized signals instead of long customer quotes

Input

Use one or more of:

• reviews: up to 500 flexible JSON objects
• files: up to five CSV, JSON, TXT, or Markdown files
• reviewsText: pasted JSON, CSV, or one review per line

Field names can vary. Records may contain text, review, rating, title, date, product, or other useful fields.

{
  "reviews": [
    { "rating": 5, "text": "Beautiful necklace and luxury packaging." },
    { "rating": 2, "text": "Shipping took too long." },
    { "rating": 3, "text": "Please add more chain lengths." }
  ],
  "outputLanguage": "English",
  "instructions": "Focus on product, delivery, and new collection opportunities."
}

Reliable analysis flow

Reviews are processed in bounded batches and then synthesized into one report. Each stage uses strict JSON validation, exact sentiment totals, bounded occurrence counts, duplicate-theme checks, and correction retries. Review content is treated as untrusted data.

For privacy, the Actor asks for short paraphrased signals and excludes customer names, contact details, and long verbatim quotes from its output.

Output

• Dataset: one ranked row per theme
• OUTPUT.json: executive summary, sentiment, themes, actions, and processing metadata
• OUTPUT.csv: UTF-8, Excel-compatible theme table
• SUMMARY.json: counts, batch information, and metering event

Environment and local test

Set LLM_API_KEY or OPENAI_API_KEY in the environment. Optional settings are LLM_MODEL, LLM_BASE_URL, and LLM_TIMEOUT_MS.

pnpm install
pnpm test
export LLM_API_KEY="your-key"
export DISABLE_METERING=true
apify run --input-file samples/cloud-input.json

Never put API keys in input or source files. The deployed Actor uses an encrypted Apify secret.

Deploy to Apify

apify login
apify push

Run samples/cloud-input.json and verify that delivery complaints, product praise, sizing requests, and action items appear in the outputs.

Suggested pay-per-event pricing

One event is emitted only after the complete report has been saved:

Event	Reviews per run	Suggested price
insight-small	1–25	$0.06
insight-medium	26–100	$0.20
insight-large	101–500	$0.75

Review real token costs before later price changes.

Publish in Apify Store

Keep the Actor private until sample and real-file tests pass. Then open Actor → Publication, select Pay per event, add the three exact event names above, complete the Store metadata, and submit it for review.

Suggested Store title: Review Insight AI

Suggested description: Turn CSV, JSON, or text reviews into sentiment, recurring themes, complaints, feature requests, and prioritized actions.