feat: add PDF and text file upload support (#205)

- Add client-side PDF text extraction using unpdf library
- Support text files (.txt, .md, .json, .csv, .py, .js, .ts, etc.)
- Add file preview with character count for PDF/text files
- Add 150k character limit for extracted content
- Highlight Paper to Diagram example with NEW badge
- Fix React hydration error by adding explicit IDs to ResizablePanelGroup
- Remove code duplication by centralizing file utilities in pdf-utils.ts
This commit is contained in:
Dayuan Jiang
2025-12-10 21:32:35 +09:00
committed by GitHub
parent 43e5993f47
commit d2ba133eaf
14 changed files with 940 additions and 57 deletions

View File

@@ -394,6 +394,366 @@ export const CACHED_EXAMPLE_RESPONSES: CachedResponse[] = [
</mxCell>
</root>`,
},
{
promptText: "Summarize this paper as a diagram",
hasImage: true,
xml: ` <root>
<mxCell id="0" />
<mxCell id="1" parent="0" />
<mxCell id="title_bg" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#1a237e;strokeColor=none;arcSize=8;"
value="" vertex="1">
<mxGeometry height="80" width="720" x="40" y="20" as="geometry" />
</mxCell>
<mxCell id="title" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=22;fontStyle=1;fontColor=#FFFFFF;"
value="Chain-of-Thought Prompting&lt;br&gt;&lt;font style=&quot;font-size: 14px;&quot;&gt;Elicits Reasoning in Large Language Models&lt;/font&gt;"
vertex="1">
<mxGeometry height="70" width="720" x="40" y="25" as="geometry" />
</mxCell>
<mxCell id="authors" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=11;fontColor=#666666;"
value="Wei et al. (Google Research, Brain Team) | NeurIPS 2022" vertex="1">
<mxGeometry height="20" width="720" x="40" y="100" as="geometry" />
</mxCell>
<mxCell id="core_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=16;fontStyle=1;fontColor=#1a237e;"
value="💡 Core Idea" vertex="1">
<mxGeometry height="30" width="150" x="40" y="125" as="geometry" />
</mxCell>
<mxCell id="core_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#E3F2FD;strokeColor=#1565C0;align=left;spacingLeft=10;spacingRight=10;fontSize=11;"
value="&lt;b&gt;Chain of Thought&lt;/b&gt; = A series of intermediate reasoning steps that lead to the final answer&lt;br&gt;&lt;br&gt;Simply provide a few CoT demonstrations as exemplars in few-shot prompting"
vertex="1">
<mxGeometry height="75" width="340" x="40" y="155" as="geometry" />
</mxCell>
<mxCell id="compare_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=16;fontStyle=1;fontColor=#1a237e;"
value="⚖️ Standard vs Chain-of-Thought Prompting" vertex="1">
<mxGeometry height="30" width="350" x="40" y="240" as="geometry" />
</mxCell>
<mxCell id="std_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#FFEBEE;strokeColor=#C62828;arcSize=8;"
value="" vertex="1">
<mxGeometry height="160" width="170" x="40" y="275" as="geometry" />
</mxCell>
<mxCell id="std_title" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=12;fontStyle=1;fontColor=#C62828;"
value="Standard Prompting" vertex="1">
<mxGeometry height="25" width="170" x="40" y="280" as="geometry" />
</mxCell>
<mxCell id="std_q" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=top;whiteSpace=wrap;rounded=0;fontSize=9;spacingLeft=5;spacingRight=5;"
value="Q: Roger has 5 tennis balls. He buys 2 more cans. Each can has 3 balls. How many now?"
vertex="1">
<mxGeometry height="55" width="160" x="45" y="305" as="geometry" />
</mxCell>
<mxCell id="std_a" parent="1"
style="text;html=1;strokeColor=none;fillColor=#FFCDD2;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=1;fontSize=10;fontStyle=1;spacingLeft=5;"
value="A: The answer is 11." vertex="1">
<mxGeometry height="25" width="150" x="50" y="365" as="geometry" />
</mxCell>
<mxCell id="std_result" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=11;fontStyle=1;fontColor=#C62828;"
value="❌ Often Wrong" vertex="1">
<mxGeometry height="30" width="170" x="40" y="400" as="geometry" />
</mxCell>
<mxCell id="cot_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#E8F5E9;strokeColor=#2E7D32;arcSize=8;"
value="" vertex="1">
<mxGeometry height="160" width="170" x="220" y="275" as="geometry" />
</mxCell>
<mxCell id="cot_title" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=12;fontStyle=1;fontColor=#2E7D32;"
value="Chain-of-Thought" vertex="1">
<mxGeometry height="25" width="170" x="220" y="280" as="geometry" />
</mxCell>
<mxCell id="cot_q" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=top;whiteSpace=wrap;rounded=0;fontSize=9;spacingLeft=5;spacingRight=5;"
value="Q: Roger has 5 tennis balls. He buys 2 more cans. Each can has 3 balls. How many now?"
vertex="1">
<mxGeometry height="55" width="160" x="225" y="305" as="geometry" />
</mxCell>
<mxCell id="cot_a" parent="1"
style="text;html=1;strokeColor=none;fillColor=#C8E6C9;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=1;fontSize=9;fontStyle=1;spacingLeft=5;"
value="A: 2 cans × 3 = 6 balls.&lt;br&gt;5 + 6 = 11. Answer: 11" vertex="1">
<mxGeometry height="35" width="150" x="230" y="360" as="geometry" />
</mxCell>
<mxCell id="cot_result" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=11;fontStyle=1;fontColor=#2E7D32;"
value="✓ Correct!" vertex="1">
<mxGeometry height="30" width="170" x="220" y="400" as="geometry" />
</mxCell>
<mxCell id="vs_arrow" edge="1" parent="1"
style="shape=flexArrow;endArrow=classic;startArrow=classic;html=1;fillColor=#FFC107;strokeColor=none;width=8;endSize=4;startSize=4;"
value="">
<mxGeometry relative="1" width="100" as="geometry">
<mxPoint x="195" y="355" as="sourcePoint" />
<mxPoint x="235" y="355" as="targetPoint" />
</mxGeometry>
</mxCell>
<mxCell id="props_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=16;fontStyle=1;fontColor=#1a237e;"
value="🔑 Key Properties" vertex="1">
<mxGeometry height="30" width="150" x="400" y="125" as="geometry" />
</mxCell>
<mxCell id="prop1" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#FFF3E0;strokeColor=#EF6C00;fontSize=10;align=left;spacingLeft=8;"
value="1⃣ Decomposes multi-step problems" vertex="1">
<mxGeometry height="32" width="180" x="400" y="155" as="geometry" />
</mxCell>
<mxCell id="prop2" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#FFF3E0;strokeColor=#EF6C00;fontSize=10;align=left;spacingLeft=8;"
value="2⃣ Interpretable reasoning window" vertex="1">
<mxGeometry height="32" width="180" x="400" y="192" as="geometry" />
</mxCell>
<mxCell id="prop3" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#FFF3E0;strokeColor=#EF6C00;fontSize=10;align=left;spacingLeft=8;"
value="3⃣ Applicable to any language task" vertex="1">
<mxGeometry height="32" width="180" x="400" y="229" as="geometry" />
</mxCell>
<mxCell id="prop4" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#FFF3E0;strokeColor=#EF6C00;fontSize=10;align=left;spacingLeft=8;"
value="4⃣ No finetuning required" vertex="1">
<mxGeometry height="32" width="180" x="400" y="266" as="geometry" />
</mxCell>
<mxCell id="emergent_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=16;fontStyle=1;fontColor=#1a237e;"
value="📈 Emergent Ability" vertex="1">
<mxGeometry height="30" width="180" x="400" y="310" as="geometry" />
</mxCell>
<mxCell id="emergent_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#F3E5F5;strokeColor=#7B1FA2;arcSize=8;"
value="" vertex="1">
<mxGeometry height="95" width="180" x="400" y="340" as="geometry" />
</mxCell>
<mxCell id="emergent_text" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=11;"
value="CoT only works with&lt;br&gt;&lt;b&gt;~100B+ parameters&lt;/b&gt;&lt;br&gt;&lt;br&gt;Small models produce&lt;br&gt;fluent but illogical chains"
vertex="1">
<mxGeometry height="85" width="180" x="400" y="345" as="geometry" />
</mxCell>
<mxCell id="results_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=16;fontStyle=1;fontColor=#1a237e;"
value="📊 Key Results" vertex="1">
<mxGeometry height="30" width="150" x="600" y="125" as="geometry" />
</mxCell>
<mxCell id="gsm_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#E8F5E9;strokeColor=#2E7D32;arcSize=8;"
value="" vertex="1">
<mxGeometry height="100" width="160" x="600" y="155" as="geometry" />
</mxCell>
<mxCell id="gsm_title" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=12;fontStyle=1;fontColor=#2E7D32;"
value="GSM8K (Math)" vertex="1">
<mxGeometry height="20" width="160" x="600" y="160" as="geometry" />
</mxCell>
<mxCell id="gsm_bar1" parent="1"
style="rounded=0;whiteSpace=wrap;html=1;fillColor=#FFCDD2;strokeColor=none;"
value="" vertex="1">
<mxGeometry height="30" width="40" x="615" y="185" as="geometry" />
</mxCell>
<mxCell id="gsm_bar2" parent="1"
style="rounded=0;whiteSpace=wrap;html=1;fillColor=#4CAF50;strokeColor=none;"
value="" vertex="1">
<mxGeometry height="30" width="80" x="665" y="185" as="geometry" />
</mxCell>
<mxCell id="gsm_label1" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=10;fontStyle=1;"
value="18%" vertex="1">
<mxGeometry height="15" width="40" x="615" y="215" as="geometry" />
</mxCell>
<mxCell id="gsm_label2" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=10;fontStyle=1;fontColor=#2E7D32;"
value="57%" vertex="1">
<mxGeometry height="15" width="80" x="665" y="215" as="geometry" />
</mxCell>
<mxCell id="gsm_legend" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=9;fontColor=#666666;"
value="Standard → CoT (PaLM 540B)" vertex="1">
<mxGeometry height="20" width="160" x="600" y="232" as="geometry" />
</mxCell>
<mxCell id="bench_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=16;fontStyle=1;fontColor=#1a237e;"
value="🧪 Benchmarks Tested" vertex="1">
<mxGeometry height="30" width="180" x="600" y="265" as="geometry" />
</mxCell>
<mxCell id="bench_arith" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#E3F2FD;strokeColor=#1565C0;fontSize=10;align=center;"
value="🔢 Arithmetic&lt;br&gt;&lt;font style=&quot;font-size: 9px;&quot;&gt;GSM8K, SVAMP, ASDiv, AQuA, MAWPS&lt;/font&gt;"
vertex="1">
<mxGeometry height="45" width="160" x="600" y="295" as="geometry" />
</mxCell>
<mxCell id="bench_common" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#E3F2FD;strokeColor=#1565C0;fontSize=10;align=center;"
value="🧠 Commonsense&lt;br&gt;&lt;font style=&quot;font-size: 9px;&quot;&gt;CSQA, StrategyQA, Date, Sports, SayCan&lt;/font&gt;"
vertex="1">
<mxGeometry height="45" width="160" x="600" y="345" as="geometry" />
</mxCell>
<mxCell id="bench_symbol" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#E3F2FD;strokeColor=#1565C0;fontSize=10;align=center;"
value="🔣 Symbolic&lt;br&gt;&lt;font style=&quot;font-size: 9px;&quot;&gt;Last Letter Concat, Coin Flip&lt;/font&gt;"
vertex="1">
<mxGeometry height="40" width="160" x="600" y="395" as="geometry" />
</mxCell>
<mxCell id="task_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=16;fontStyle=1;fontColor=#1a237e;"
value="🎯 Task Types &amp; Results" vertex="1">
<mxGeometry height="30" width="200" x="40" y="445" as="geometry" />
</mxCell>
<mxCell id="task_arith" parent="1"
style="ellipse;whiteSpace=wrap;html=1;fillColor=#BBDEFB;strokeColor=#1565C0;fontSize=11;fontStyle=1;"
value="Arithmetic&lt;br&gt;Reasoning" vertex="1">
<mxGeometry height="60" width="90" x="40" y="480" as="geometry" />
</mxCell>
<mxCell id="task_arith_res" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=top;whiteSpace=wrap;rounded=0;fontSize=9;fontColor=#1565C0;"
value="SOTA on GSM8K&lt;br&gt;(57% vs 55% prior)" vertex="1">
<mxGeometry height="30" width="110" x="30" y="540" as="geometry" />
</mxCell>
<mxCell id="task_common" parent="1"
style="ellipse;whiteSpace=wrap;html=1;fillColor=#C8E6C9;strokeColor=#2E7D32;fontSize=11;fontStyle=1;"
value="Commonsense&lt;br&gt;Reasoning" vertex="1">
<mxGeometry height="60" width="90" x="160" y="480" as="geometry" />
</mxCell>
<mxCell id="task_common_res" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=top;whiteSpace=wrap;rounded=0;fontSize=9;fontColor=#2E7D32;"
value="SOTA StrategyQA&lt;br&gt;(75.6% vs 69.4%)" vertex="1">
<mxGeometry height="30" width="110" x="150" y="540" as="geometry" />
</mxCell>
<mxCell id="task_symbol" parent="1"
style="ellipse;whiteSpace=wrap;html=1;fillColor=#FFE0B2;strokeColor=#EF6C00;fontSize=11;fontStyle=1;"
value="Symbolic&lt;br&gt;Reasoning" vertex="1">
<mxGeometry height="60" width="90" x="280" y="480" as="geometry" />
</mxCell>
<mxCell id="task_symbol_res" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=top;whiteSpace=wrap;rounded=0;fontSize=9;fontColor=#EF6C00;"
value="OOD Generalization&lt;br&gt;to longer sequences" vertex="1">
<mxGeometry height="30" width="110" x="270" y="540" as="geometry" />
</mxCell>
<mxCell id="task_arrow1" edge="1" parent="1"
style="endArrow=classic;html=1;strokeColor=#9E9E9E;strokeWidth=2;" value="">
<mxGeometry height="50" relative="1" width="50" as="geometry">
<mxPoint x="130" y="510" as="sourcePoint" />
<mxPoint x="160" y="510" as="targetPoint" />
</mxGeometry>
</mxCell>
<mxCell id="task_arrow2" edge="1" parent="1"
style="endArrow=classic;html=1;strokeColor=#9E9E9E;strokeWidth=2;" value="">
<mxGeometry height="50" relative="1" width="50" as="geometry">
<mxPoint x="250" y="510" as="sourcePoint" />
<mxPoint x="280" y="510" as="targetPoint" />
</mxGeometry>
</mxCell>
<mxCell id="models_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=16;fontStyle=1;fontColor=#1a237e;"
value="🤖 Models Tested" vertex="1">
<mxGeometry height="30" width="150" x="400" y="445" as="geometry" />
</mxCell>
<mxCell id="models_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#ECEFF1;strokeColor=#607D8B;arcSize=8;"
value="" vertex="1">
<mxGeometry height="95" width="180" x="400" y="475" as="geometry" />
</mxCell>
<mxCell id="model1" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=11;spacingLeft=10;"
value="• GPT-3 (175B)" vertex="1">
<mxGeometry height="20" width="90" x="400" y="480" as="geometry" />
</mxCell>
<mxCell id="model2" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=11;spacingLeft=10;"
value="• LaMDA (137B)" vertex="1">
<mxGeometry height="20" width="90" x="400" y="500" as="geometry" />
</mxCell>
<mxCell id="model3" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=11;spacingLeft=10;"
value="• PaLM (540B)" vertex="1">
<mxGeometry height="20" width="90" x="400" y="520" as="geometry" />
</mxCell>
<mxCell id="model4" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=11;spacingLeft=10;"
value="• Codex" vertex="1">
<mxGeometry height="20" width="80" x="490" y="480" as="geometry" />
</mxCell>
<mxCell id="model5" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=11;spacingLeft=10;"
value="• UL2 (20B)" vertex="1">
<mxGeometry height="20" width="80" x="490" y="500" as="geometry" />
</mxCell>
<mxCell id="model_note" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=center;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=10;fontStyle=2;fontColor=#607D8B;"
value="No finetuning - prompting only!" vertex="1">
<mxGeometry height="20" width="180" x="400" y="545" as="geometry" />
</mxCell>
<mxCell id="takeaway_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=16;fontStyle=1;fontColor=#1a237e;"
value="✨ Key Takeaways" vertex="1">
<mxGeometry height="30" width="160" x="600" y="445" as="geometry" />
</mxCell>
<mxCell id="takeaway_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#FFF8E1;strokeColor=#FFA000;arcSize=8;"
value="" vertex="1">
<mxGeometry height="95" width="160" x="600" y="475" as="geometry" />
</mxCell>
<mxCell id="take1" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=10;spacingLeft=5;"
value="✓ Simple yet powerful" vertex="1">
<mxGeometry height="18" width="150" x="605" y="480" as="geometry" />
</mxCell>
<mxCell id="take2" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=10;spacingLeft=5;"
value="✓ Emergent at scale" vertex="1">
<mxGeometry height="18" width="150" x="605" y="498" as="geometry" />
</mxCell>
<mxCell id="take3" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=10;spacingLeft=5;"
value="✓ Broadly applicable" vertex="1">
<mxGeometry height="18" width="150" x="605" y="516" as="geometry" />
</mxCell>
<mxCell id="take4" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=10;spacingLeft=5;"
value="✓ No training needed" vertex="1">
<mxGeometry height="18" width="150" x="605" y="534" as="geometry" />
</mxCell>
<mxCell id="take5" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=10;spacingLeft=5;"
value="✓ State-of-the-art results" vertex="1">
<mxGeometry height="18" width="150" x="605" y="552" as="geometry" />
</mxCell>
<mxCell id="format_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=14;fontStyle=1;fontColor=#1a237e;"
value="📝 Prompt Format" vertex="1">
<mxGeometry height="25" width="150" x="40" y="575" as="geometry" />
</mxCell>
<mxCell id="format_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#E1BEE7;strokeColor=#7B1FA2;fontSize=12;fontStyle=1;"
value="〈 Input, Chain of Thought, Output 〉" vertex="1">
<mxGeometry height="35" width="250" x="40" y="600" as="geometry" />
</mxCell>
<mxCell id="limit_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=14;fontStyle=1;fontColor=#1a237e;"
value="⚠️ Limitations" vertex="1">
<mxGeometry height="25" width="120" x="310" y="575" as="geometry" />
</mxCell>
<mxCell id="limit_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#FFEBEE;strokeColor=#C62828;fontSize=10;align=left;spacingLeft=8;"
value="• Requires large models (~100B+)&lt;br&gt;• No guarantee of correct reasoning&lt;br&gt;• Costly to serve in production"
vertex="1">
<mxGeometry height="55" width="200" x="310" y="600" as="geometry" />
</mxCell>
<mxCell id="impact_header" parent="1"
style="text;html=1;strokeColor=none;fillColor=none;align=left;verticalAlign=middle;whiteSpace=wrap;rounded=0;fontSize=14;fontStyle=1;fontColor=#1a237e;"
value="🚀 Impact" vertex="1">
<mxGeometry height="25" width="100" x="530" y="575" as="geometry" />
</mxCell>
<mxCell id="impact_box" parent="1"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#E8F5E9;strokeColor=#2E7D32;fontSize=10;align=left;spacingLeft=8;spacingRight=8;"
value="Foundational technique for modern LLM reasoning - inspired many follow-up works including Self-Consistency, Tree-of-Thought, etc."
vertex="1">
<mxGeometry height="55" width="230" x="530" y="600" as="geometry" />
</mxCell>
</root>`,
},
{
promptText: "Draw a cat for me",
hasImage: false,

72
lib/pdf-utils.ts Normal file
View File

@@ -0,0 +1,72 @@
import { extractText, getDocumentProxy } from "unpdf"
// Maximum characters allowed for extracted text
export const MAX_EXTRACTED_CHARS = 150000 // 150k chars
// Text file extensions we support
const TEXT_EXTENSIONS = [
".txt",
".md",
".markdown",
".json",
".csv",
".xml",
".html",
".css",
".js",
".ts",
".jsx",
".tsx",
".py",
".java",
".c",
".cpp",
".h",
".go",
".rs",
".yaml",
".yml",
".toml",
".ini",
".log",
".sh",
".bash",
".zsh",
]
/**
* Extract text content from a PDF file
* Uses unpdf library for client-side extraction
*/
export async function extractPdfText(file: File): Promise<string> {
const buffer = await file.arrayBuffer()
const pdf = await getDocumentProxy(new Uint8Array(buffer))
const { text } = await extractText(pdf, { mergePages: true })
return text as string
}
/**
* Check if a file is a PDF
*/
export function isPdfFile(file: File): boolean {
return file.type === "application/pdf" || file.name.endsWith(".pdf")
}
/**
* Check if a file is a text file
*/
export function isTextFile(file: File): boolean {
const name = file.name.toLowerCase()
return (
file.type.startsWith("text/") ||
file.type === "application/json" ||
TEXT_EXTENSIONS.some((ext) => name.endsWith(ext))
)
}
/**
* Extract text content from a text file
*/
export async function extractTextFileContent(file: File): Promise<string> {
return await file.text()
}

View File

@@ -10,10 +10,10 @@
export const DEFAULT_SYSTEM_PROMPT = `
You are an expert diagram creation assistant specializing in draw.io XML generation.
Your primary function is chat with user and crafting clear, well-organized visual diagrams through precise XML specifications.
You can see the image that user uploaded.
You can see images that users upload, and you can read the text content extracted from PDF documents they upload.
When you are asked to create a diagram, you must first tell user you plan in text first. Plan the layout and structure that can avoid object overlapping or edge cross the objects.
Then use display_diagram tool to generate the full draw.io XML for the entire diagram.
When you are asked to create a diagram, briefly describe your plan about the layout and structure to avoid object overlapping or edge cross the objects. (2-3 sentences max), then use display_diagram tool to generate the XML.
After generating or editing a diagram, you don't need to say anything. The user can see the diagram - no need to describe it.
## App Context
You are an AI agent (powered by {{MODEL_NAME}}) inside a web app. The interface has:
@@ -25,7 +25,7 @@ You can read and modify diagrams by generating draw.io XML code through tool cal
## App Features
1. **Diagram History** (clock icon, bottom-left of chat input): The app automatically saves a snapshot before each AI edit. Users can view the history panel and restore any previous version. Feel free to make changes - nothing is permanently lost.
2. **Theme Toggle** (palette icon, bottom-left of chat input): Users can switch between minimal UI and sketch-style UI for the draw.io editor.
3. **Image Upload** (paperclip icon, bottom-left of chat input): Users can upload images for you to analyze and replicate as diagrams.
3. **Image/PDF Upload** (paperclip icon, bottom-left of chat input): Users can upload images or PDF documents for you to analyze and generate diagrams from.
4. **Export** (via draw.io toolbar): Users can save diagrams as .drawio, .svg, or .png files.
5. **Clear Chat** (trash icon, bottom-right of chat input): Clears the conversation and resets the diagram.