SpecLens – Chrome Extension

Annotates CPU/GPU model names on shopping sites with PassMark performance, rank, value, and price.

Usage (Unpacked Install)

1. Download or clone the repo.
2. Open chrome://extensions and toggle Developer mode (top-right).
3. Click Load unpacked and select the project folder (unzipped).
4. Visit any page listing CPUs/GPUs (e.g., product pages or search results). Matches will be annotated inline.

Example Intel Core i5-13500 → Intel Core i5-13500 (31k, ranked 415, value 150.05, $208)

---

How SpecLens Works (Key Parts)

This section focuses only on the regexes and replace/remove commands that determine what gets annotated, so you can quickly see if your use case is covered.

1. Fetching Reference Data

• On first use, SpecLens requests two PassMark pages through its background service worker:
  • CPUs: https://www.cpubenchmark.net/cpu_list.php
  • GPUs: https://www.videocardbenchmark.net/gpu_list.php
• The background script performs fetch with credentials: "omit", bypassing CORS restrictions.
• The raw HTML is returned to the content script via chrome.runtime.sendMessage.

2. Parsing the Benchmark Tables

• The HTML is parsed with a DOMParser, scanning the <table><tbody><tr> rows.
• Each row typically has:
  1. Name (<td><a>Intel Core i5-13500</a></td>)
  2. Score (e.g., 31,236)
  3. Rank (e.g., 415)
  4. Value score (e.g., 150.05)
  5. Price (e.g., $208.17)
• For each entry, SpecLens stores a normalized object:

  {
    "name": "Intel Core i5-13500",
    "key": "intel core i5-13500",  // lowercased normalized key
    "score": 31236,
    "rank": 415,
    "value": 150.05,
    "price": "$208.17"
  }
  

3) Detection Regex

Used to decide if a text node might contain hardware names:

//annotate.js
/(Intel|AMD|Ryzen|Core|Xeon|Celeron|Pentium|Threadripper|GeForce|Radeon|RTX|GTX|RX)\b/i

This ensures we only process nodes that look like CPU/GPU names.

4) Candidate Extraction

Sliding window of up to 8 tokens; any substring that matches the detection regex above is considered a candidate.

5) Normalization (replace/remove steps)

These ensure that product names with trademarks, spacing, or vendor prefixes still match the benchmark table.

/[\u2122\u00AE]/g          // remove ™ and ®
/\s+/g                     // collapse whitespace
/\s*\(R\)|\s*\(TM\)/gi   // remove (R) and (TM)
/\s*-\s*/g                 // normalize spaces around dashes
/^Intel\s+CPU\s+/i         // tidy prefixed descriptors
/^AMD\s+CPU\s+/i
/^NVIDIA\s+GPU\s+/i

4) Alias Rules

• Remove redundant words like Processor or CPU from the name.
• Allow GPU names without vendor prefix (e.g., NVIDIA GeForce RTX 4070 → GeForce RTX 4070).

5) Annotation Injection

When a match is found, the name is annotated inline with:

(score, ranked X, value Y, $price)

---

Data Freshness

• Benchmark tables (CPU & GPU lists) are fetched and cached for 7 days.
• Clearing extension storage or reloading forces a refresh.

Limitations

• Matching is exact after normalization (not fuzzy beyond the rules above).
• If PassMark table markup changes, selectors will need updating.
• Price/value data is PassMark’s snapshot, not live retailer info.