jev-reviews
Pipeline that ingests Google Maps restaurant reviews and classifies sentiment, five rating pillars and flags with Jev or an LLM.
Overview
jev-reviews is the flagship experiment in an Astro and TypeScript workspace for AI experiments. It pulls Google Maps restaurant reviews from Apify storage, stores and deduplicates them in Supabase PostgreSQL, and classifies each review with Jev (or, for comparison, OpenAI, Gemini or Groq models). Jev answers a sentiment Choice (positive, neutral, negative), Score ratings from 1 to 5 for food quality, service, cleanliness, atmosphere and value, and Nouls for whether the reviewer recommends, complains or praises; the pipeline also extracts dishes, highlights and owner recommendations.
A dashboard shows review cards with sentiment badges and score strips, or a dense score-matrix table, plus live speed and cost KPIs. The README cites about 95-365 ms per Jev response (around 230 ms average) and roughly $0.00002-0.00004 per review classification.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What is the sentiment of this review? | positive · neutral · negative |
| Score | How does the review rate food quality / service / cleanliness / atmosphere / value (1-5)? | 1 · 2 · 3 · 4 · 5 |
| Yes / no | Does the reviewer recommend the restaurant? | yes · no |
How it works
- Ingest reviews from Apify Google Maps scrapes
- Store and deduplicate in Supabase
- Classify each review with Jev (or an LLM)
- Display cards, score matrix and cost KPIs
Strengths
- Side-by-side comparison with LLM backends
- Low per-review cost as reported
- Structured operational scores per review
Caveats
- Experimental workspace
- Depends on Apify and Supabase accounts
- Figures are the README's own
At a glance
- Type
- Service
- Status
- prototype
- Language
- TypeScript
- Runtime
- Astro web app with Supabase
- Has code
- yes
Use cases and integrations
How curated lists describe it
The flagship experiment in this workspace is \*\*\`jev-reviews\`\*\*: an automated pipeline that ingests Google Maps restaurant reviews from \*\*Apify Storage\*\*, stores and deduplicates them in \*\*Supabase PostgreSQL\*\*, and classifies them using \*\*Jev (TypeSafe System One)\*\* or traditional generative LLMs across 5 operational pillars, extracting dishes, customer highlights, and owner recommendations.
logicrw/awesome-jev-projectsTypeSafe AI - Jev - first look and experiment with its API + Google reviews experiments.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Review sentiment, 1-5 ratings across five pillars, and recommend/complaint/praise flags.
Can I compare with LLMs?
Yes. OpenAI, Gemini and Groq backends are supported.
Where do reviews come from?
Apify Google Maps scraper datasets.
Related projects
Codex plugin for batch labelling JSONL records with Jev under custom policies, with Codex reviewing uncertain labels.
MCP serverbetaTypeScriptMines timestamped driving scenes from vehicle logs by asking Jev lateral and longitudinal manoeuvre questions per moment.
ProjectbetaPythonLive race labelling 1,000 app reviews with four typed Jev questions versus Gemini 3.8 Flash JSON output.
AppprototypeJavaScriptPython library with a TypeScript port that turns Jev questions into functions over lists and DataFrames, with caching and escalation.
LibrarybetaPython, TypeScriptpandas and Polars .jev accessor for classifying, sentiment-scoring and grading DataFrame rows with Jev probabilities.
DemobetaPython- jev-arena★ 122
Side-by-side tool comparing Jev and DeepSeek on labelling the same comments for relevance, sentiment and intent, with replay and reports.
ProjectbetaTypeScript