paper-view-skill（论文查看技巧）

PaperView API

PaperView provides three AI-powered visualization APIs:

1. ECharts Visualization — generate interactive charts from CSV/JSON/text data
2. AI Scientific Diagram — generate publication-quality figures from text or arxiv PDF
3. Word Cloud — extract keyword frequencies from PDF documents (supports CJK)

Authentication

All requests require an API Token:

Authorization: Bearer pv_live_<your_token>

Obtain your token from the www.ipaperview.com website (Profile → API Token). Each account can create one token. Set via environment variable PAPERVIEW_API_TOKEN or ask the user for their token.

Base URL

https://api.ipaperview.com

API Quota

Daily API call limits by plan:

Each API call consumes 1 quota.

1. ECharts Visualization

POST /api/v1/viz/generate/

AI analyzes your data sample, selects the best chart type, and returns a ready-to-render ECharts option. The full data processing happens server-side — AI only sees a sample to generate a transform script, then the backend executes it with Node.js on the full dataset.

Available chart types:

• Basic: line, bar, scatter, pie, radar, funnel, boxplot
• Advanced: heatmap, treemap, sankey, tree, sunburst, map, structure
• 3D: gl3d (bar3D, scatter3D)
• Scientific: violin, manhattan, volcano, forest, survival, roc, venn, upset, enrichment, circos, waterfall

curl -X POST \
  -H "Authorization: Bearer $PAPERVIEW_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "data": "Year,China,USA,Japan\
2018,13608,20544,4971\
2019,14280,21373,5082\
2020,14723,20894,5040\
2021,17734,23315,4941\
2022,17963,25463,4231",
    "context": "Show GDP trends as a line chart"
  }' \
  https://api.ipaperview.com/api/v1/viz/generate/

Response:

{
  "success": true,
  "toolset": "line",
  "template": "confidence_band",
  "echarts_option": {
    "title": { "text": "GDP by Country (2018-2022)" },
    "xAxis": { "type": "category", "data": ["2018", "2019", "2020", "2021", "2022"] },
    "yAxis": { "type": "value" },
    "series": [
      { "name": "China", "type": "line", "data": [13608, 14280, 14723, 17734, 17963] },
      { "name": "USA", "type": "line", "data": [20544, 21373, 20894, 23315, 25463] },
      { "name": "Japan", "type": "line", "data": [4971, 5082, 5040, 4941, 4231] }
    ],
    "legend": { "data": ["China", "USA", "Japan"] }
  },
  "reason": "Time series data with multiple countries — line chart shows trends clearly"
}

The echarts_option can be rendered directly with echarts.setOption(echarts_option) in any ECharts-compatible environment.

2. AI Scientific Diagram

POST /api/diagram/ai-generate/

Generate publication-quality scientific diagrams, flowcharts, and research illustrations. Supports arxiv paper URLs directly. Returns a Server-Sent Events stream — use curl -N to receive events.

Example — from text description:

curl -N -X POST \
  -H "Authorization: Bearer $PAPERVIEW_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "template_type": "custom_text_only",
    "custom_prompt": "A flowchart: Data Collection -> Preprocessing -> Training -> Evaluation"
  }' \
  https://api.ipaperview.com/api/diagram/ai-generate/

Example — from arxiv paper with custom style:

curl -N -X POST \
  -H "Authorization: Bearer $PAPERVIEW_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "template_type": "custom_text_only",
    "pdf_url": "http://arxiv.org/abs/2510.13809v1",
    "custom_prompt": "Generate an architecture diagram with pink color scheme, include Doraemon as a mascot element"
  }' \
  https://api.ipaperview.com/api/diagram/ai-generate/

SSE Events (in order):

data: {"type": "step", "step": "extracting", "message": "Extracting document content..."}
data: {"type": "step", "step": "generating_prompt", "message": "Generating image prompt..."}
data: {"type": "prompt_generated", "prompt": "...", "enhanced_prompt": "..."}
data: {"type": "step", "step": "generating_image", "message": "Generating image..."}
data: {"type": "complete", "success": true, "image_url": "https://...", "model": "gemini-2.0-flash-preview-image-generation"}

The final complete event contains image_url — a CDN URL to the generated image.

3. Word Cloud

POST /api/wordcloud/extract/

Extract keyword frequencies from PDF documents with automatic CJK (Chinese/Japanese/Korean) segmentation and semantic clustering.

curl -X POST \
  -H "Authorization: Bearer $PAPERVIEW_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "pdf_url": "https://arxiv.org/pdf/2301.00001.pdf",
    "max_words": 50
  }' \
  https://api.ipaperview.com/api/wordcloud/extract/

Response:

{
  "success": true,
  "data": [
    { "word": "neural", "frequency": 45, "cluster": 0 },
    { "word": "network", "frequency": 38, "cluster": 0 },
    { "word": "attention", "frequency": 32, "cluster": 1 }
  ]
}

The cluster field (0-4) groups words by frequency — useful for color coding in word cloud rendering. CJK documents are automatically segmented using jieba.

Error Responses

{ "success": false, "error": "Error description" }

Rendering Results

After calling the API, you MUST render the results visually. No extra dependencies needed — use CDN-loaded HTML files.

Render ECharts Visualization

When you receive echarts_option from /api/v1/viz/generate/, create a self-contained HTML file and open it in the browser:

<!DOCTYPE html>
<html>
<head>
  <meta charset="utf-8">
  <title>PaperView Chart</title>
  <script src="https://cdn.jsdelivr.net/npm/echarts@5/dist/echarts.min.js"></script>
  <style>body{margin:0;background:#fff}#chart{width:100vw;height:100vh}</style>
</head>
<body>
  <div id="chart"></div>
  <script>
    var chart = echarts.init(document.getElementById('chart'));
    var option = {{ECHARTS_OPTION_JSON}};
    chart.setOption(option);
    window.addEventListener('resize', function() { chart.resize(); });
  </script>
</body>
</html>

Replace {{ECHARTS_OPTION_JSON}} with the echarts_option value from the API response (the JSON object, not stringified). Save as a .html file and open:

# macOS
open /tmp/paperview_chart.html
# Linux
xdg-open /tmp/paperview_chart.html

For 3D charts (gl3d toolset), add this additional script tag before the main script:

<script src="https://cdn.jsdelivr.net/npm/echarts-gl@2/dist/echarts-gl.min.js"></script>

Render Word Cloud

When you receive word frequency data from /api/wordcloud/extract/, create an HTML file using echarts-wordcloud:

<!DOCTYPE html>
<html>
<head>
  <meta charset="utf-8">
  <title>PaperView Word Cloud</title>
  <script src="https://cdn.jsdelivr.net/npm/echarts@5/dist/echarts.min.js"></script>
  <script src="https://cdn.jsdelivr.net/npm/echarts-wordcloud@2/dist/echarts-wordcloud.min.js"></script>
  <style>body{margin:0;background:#fff}#chart{width:100vw;height:100vh}</style>
</head>
<body>
  <div id="chart"></div>
  <script>
    var chart = echarts.init(document.getElementById('chart'));
    var words = {{WORDCLOUD_DATA_JSON}};
    var clusterColors = ['#5470c6','#91cc75','#fac858','#ee6666','#73c0de'];
    chart.setOption({
      series: [{
        type: 'wordCloud',
        sizeRange: [14, 80],
        rotationRange: [-45, 45],
        gridSize: 8,
        shape: 'circle',
        textStyle: {
          fontFamily: 'sans-serif',
          color: function(params) {
            return clusterColors[params.data.cluster || 0];
          }
        },
        data: words.map(function(w) {
          return { name: w.word, value: w.frequency, cluster: w.cluster };
        })
      }]
    });
  </script>
</body>
</html>

Replace {{WORDCLOUD_DATA_JSON}} with the data array from the API response.

Render AI Diagram

The AI diagram endpoint already returns an image_url (CDN link). Simply download it or open in browser:

# Open in browser
open "https://cdn.example.com/generated/image.jpg"
# Or download
curl -o /tmp/paperview_diagram.png "https://cdn.example.com/generated/image.jpg"

Privacy & Data Handling

What data is sent to api.ipaperview.com:

• ECharts API: The raw text data (CSV/JSON/text string) you provide in the data field is sent to the server. No local files are uploaded — you must pass the data content as a string in the request body.
• AI Diagram API: Only the pdf_url (a public URL) is sent. The server fetches the PDF from that URL. No local files are uploaded from your machine.
• Word Cloud API: Only the pdf_url (a public URL) is sent. The server fetches the PDF from that URL. No local files are uploaded from your machine.

Data retention:

• Submitted text data (CSV/JSON) is processed in memory and not stored after the response is returned.
• Generated AI diagram images are uploaded to Upyun CDN and retained for 30 days, after which they are automatically deleted.
• No user data is sold or shared with third parties.

Authentication:

• This skill requires a PAPERVIEW_API_TOKEN environment variable (format: pv_live_<hex_string>).
• Obtain your token from www.ipaperview.com → Profile → API Token.
• Each API call is authenticated and rate-limited per your subscription plan.

Usage Tips

1. Auto-selection: For ECharts, omit toolset and template to let AI pick the best chart type for your data
2. Large datasets: Safe to send large CSV files — AI only sees a sample, full data is processed server-side
3. SSE parsing: AI diagram endpoint uses Server-Sent Events. Parse each data: line as JSON and wait for the type: "complete" event
4. Arxiv URLs: Both http://arxiv.org/abs/... and https://arxiv.org/pdf/... formats are supported for pdf_url
5. Custom styles: Use the custom_prompt and context fields to request specific colors, themes, or visual elements
6. Always render: After receiving API results, always render them visually using the HTML templates above so the user can see the chart/wordcloud