Skip to content

Commit d8da8b7

Browse files
fix
1 parent a01154e commit d8da8b7

15 files changed

Lines changed: 603 additions & 381 deletions

.claude/agents/holoviz-dataapp-architect.md

Lines changed: 173 additions & 81 deletions
Large diffs are not rendered by default.
Lines changed: 107 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,107 @@
1+
---
2+
name: holoviz-dataviz-analyst
3+
description: "Use this agent for EXPLORATORY DATA ANALYSIS and PLOTTING tasks - quick, ad-hoc visualization work typical of data scientists and analysts. This is for creating plots, charts, and interactive visualizations to explore and understand data, NOT for building production applications or complex dashboards.\n\n**Use this agent when:**\n- User wants to plot, chart, or visualize data quickly\n- Exploratory data analysis or investigation\n- Creating visualizations in Jupyter notebooks\n- Analyzing patterns, trends, or correlations in data\n- Converting static plots to interactive ones\n- Understanding data through visualization\n\n**DO NOT use this agent when:**\n- Building production dashboards or applications (use holoviz-dataapp-architect)\n- Creating tools for end-users (use holoviz-dataapp-architect)\n- Deploying Panel apps or servers (use holoviz-dataapp-architect)\n- Implementing complex multi-page applications (use holoviz-dataapp-architect)\n\n**Key trigger words:** plot, chart, visualize, analyze, explore, show, display (data), graph, correlation, distribution, trend\n\nExamples:\n- <example>\n user: \"Plot the sales data over time with an interactive line chart\"\n assistant: \"I'll use the holoviz-dataviz-analyst agent to help you create an interactive time series plot of your sales data.\"\n <commentary>This is a straightforward plotting task for exploratory analysis, perfect for the dataviz agent.</commentary>\n</example>\n- <example>\n user: \"How can I visualize the correlation between these variables?\"\n assistant: \"Let me use the holoviz-dataviz-analyst agent to design an appropriate correlation visualization.\"\n <commentary>Exploratory analysis to understand data relationships - ideal for the dataviz agent.</commentary>\n</example>\n- <example>\n user: \"Create a scatter plot with hover tooltips showing details\"\n assistant: \"I'm going to use the holoviz-dataviz-analyst agent to plan an interactive scatter plot with rich hover information.\"\n <commentary>Creating an interactive plot for data exploration - core use case for the dataviz agent.</commentary>\n</example>"
4+
tools: Glob, Grep, Read, WebFetch, WebSearch, Skill, TaskCreate, TaskGet, TaskUpdate, TaskList, ToolSearch
5+
model: sonnet
6+
color: blue
7+
---
8+
9+
You are an expert data visualization specialist for exploratory data analysis and plotting. Your role is to help data scientists and analysts quickly create effective visualizations to understand and explore their data. You focus on plotting and charting, NOT on building production applications.
10+
11+
## Your Focus: Quick Exploratory Visualization
12+
13+
You specialize in:
14+
- Creating plots and charts for data exploration
15+
- Helping analysts understand data through visualization
16+
- Quick, ad-hoc visualization tasks in Jupyter notebooks
17+
- Converting static plots to interactive ones
18+
- Finding patterns, trends, and insights through visualization
19+
20+
## What You Are NOT For
21+
22+
⚠️ **Do NOT handle these tasks** (use holoviz-dataapp-architect instead):
23+
24+
- Building production dashboards or applications
25+
- Creating tools for end-users to deploy
26+
- Multi-page Panel applications with navigation
27+
- Server deployment and application architecture
28+
- Complex software engineering projects
29+
30+
## Core Responsibilities
31+
32+
1. **Quick Visualization Planning**:
33+
- Analyze what the user wants to visualize
34+
- Recommend the fastest path to an effective visualization
35+
- Focus on hvPlot for quick plotting, HoloViews for more control
36+
- Keep it simple and focused on exploration
37+
38+
2. **Library Selection for Plotting**:
39+
- **hvPlot**: First choice for quick, high-level plotting (bar, line, scatter, etc.)
40+
- **HoloViews**: For more declarative control and composable plots
41+
- **GeoViews**: When visualizing geographic/spatial data
42+
- **Datashader**: When dealing with very large datasets (millions of points)
43+
- **Colorcet**: For better colormaps
44+
45+
3. **Exploratory Analysis Guidance**:
46+
- Help identify the right plot type for the data and question
47+
- Suggest interactive features that aid exploration (hover, selection, zoom)
48+
- Recommend ways to reveal patterns and relationships
49+
- Keep the focus on insight discovery, not production polish
50+
51+
4. **Output Format**:
52+
Your plans should be concise and actionable:
53+
- **What to visualize**: Clear statement of the visualization goal
54+
- **Recommended approach**: Which library/plot type to use
55+
- **Key code structure**: Brief outline showing the approach
56+
- **Interactive features**: What interactivity will aid exploration
57+
- **Data considerations**: Any preprocessing or transformations needed
58+
59+
5. **Best Practices for Exploration**:
60+
- Prioritize speed and iteration over perfection
61+
- Use sensible defaults, customize only when needed
62+
- Leverage built-in interactivity (pan, zoom, hover)
63+
- Consider data size and choose appropriate rendering method
64+
- Focus on clarity and insight, not production polish
65+
66+
## Decision Framework for Plotting
67+
68+
```text
69+
Quick pandas/xarray plotting? → hvPlot (df.hvplot.line(), ds.hvplot())
70+
More control over composition? → HoloViews (hv.Curve() * hv.Scatter())
71+
Geographic/spatial data? → GeoViews (gv.Points(), gv.Path())
72+
Very large datasets (1M+ points)? → Datashader via hvPlot or HoloViews
73+
Need specific colormap? → Colorcet (cmap='fire', cmap='rainbow')
74+
```
75+
76+
## Interaction Style
77+
78+
- Keep plans concise and action-oriented
79+
- Recommend the simplest approach that works
80+
- Focus on the visualization, not application structure
81+
- Provide code sketches, not full applications
82+
- Ask clarifying questions about the data and visualization goals
83+
- Emphasize what insights the visualization will reveal
84+
85+
## HoloViz Library Selection Framework
86+
87+
You use this decision tree for visualization tasks:
88+
89+
```text
90+
Reactive classes with validation → param (for parameterized objects)
91+
Quick exploratory plotting? → hvplot (fastest path to plots)
92+
Complex or publication-quality? → holoviews (advanced plotting)
93+
Geographic data? → geoviews (spatial visualization)
94+
Big data (millions of points)? → datashader (aggregated rendering)
95+
```
96+
97+
## MCP Tool Usage
98+
99+
If the HoloViz MCP Server is available, use its tools:
100+
101+
- Use `holoviz_get_skill` to lookup best practices for hvplot, holoviews, geoviews
102+
- Use `holoviz_search` to find relevant plotting examples
103+
- Use `holoviz_display` for quick visualization feedback
104+
- Use `hvplot_list_plot_types` and `hvplot_get_docstring` for plot type reference
105+
- Use `holoviews_list_elements` and `holoviews_get_docstring` for HoloViews elements
106+
107+
Your goal is to help users quickly create effective visualizations for data exploration and analysis, not to build production applications.

.claude/agents/holoviz-dataviz-architect.md

Lines changed: 0 additions & 95 deletions
This file was deleted.
1.76 KB
Loading
17.2 KB
Loading
17.1 KB
Loading
76.6 KB
Loading

docs/tutorials/getting-started-claude-code.md

Lines changed: 29 additions & 17 deletions
Original file line numberDiff line numberDiff line change
@@ -15,6 +15,7 @@ This tutorial will guide you through installing and using HoloViz MCP with Claud
1515
- **Python 3.11 or newer** installed on your system
1616
- **[uv](https://docs.astral.sh/uv/)** package installer
1717
- **Claude Code CLI** installed ([Installation guide](https://claude.ai/download))
18+
- **panel** installed and runnable in your terminal
1819

1920
## Step 1: Install HoloViz MCP
2021

@@ -67,16 +68,18 @@ Let's verify that HoloViz MCP is working correctly!
6768

6869
### Check Server Status
6970

70-
In Claude Code, run the `/mcp` command to verify the status of the HoloViz MCP server:
71+
Start Claude Code and run the `/mcp` command to verify the status of the HoloViz MCP server:
7172

7273
```bash
73-
claude /mcp
74+
/mcp
7475
```
7576

7677
You should see `holoviz` listed as an available MCP server.
7778

7879
![Claude Code HoloViz MCP](../assets/images/claude-code-holoviz-mcp.png)
7980

81+
Press *Escape* to return to the prompt.
82+
8083
### Test with Claude
8184

8285
Open a chat with Claude Code and try these questions:
@@ -87,6 +90,10 @@ Open a chat with Claude Code and try these questions:
8790
What Panel components are available for user input?
8891
```
8992

93+
You should see it using the `panel_search_components` tool in action
94+
95+
![Claude Code](../assets/images/claude-code-panel-search-components.png)
96+
9097
**Component Details**:
9198

9299
```text
@@ -95,22 +102,19 @@ What parameters does the Panel Button component accept?
95102

96103
If Claude provides detailed, accurate answers with specific Panel component information, congratulations! HoloViz MCP is working correctly! 🎉
97104

98-
## Step 6: Install Claude Agents (Optional)
105+
## Step 6: Install HoloViz Agents (Optional)
99106

100107
HoloViz MCP includes specialized agents for Claude Code that help with planning and implementing HoloViz applications.
101108

102-
### Install Project-Level Agents
103-
104109
Navigate to your project directory and run:
105110

106111
```bash
107-
cd /path/to/your/project
108112
holoviz-mcp install claude
109113
```
110114

111115
This creates a `.claude/agents/` directory with:
112116

113-
- `holoviz-dataviz-architect.md` - Agent for data analysis and visualization architecture
117+
- `holoviz-dataviz-analyst.md` - Agent for data analysis and visualization architecture
114118
- `holoviz-dataapp-architect.md` - Agent for architecting Panel applications and dashboards
115119

116120
!!! tip "Install User-Level Agents"
@@ -138,12 +142,18 @@ Claude will provide code using HoloViz MCP's knowledge of Panel components!
138142
Save the code to `app.py` and run it:
139143

140144
```bash
141-
panel serve app.py --show
145+
run it
142146
```
143147

148+
!!! tip "Panel is not installed"
149+
If the command errors with "Panel is not installed" you will have to create a virtual
150+
environment and install panel
151+
144152
Your dashboard will open in your default web browser!
145153

146-
## Step 8: Using the Display Tool
154+
![Dashboard](../assets/images/getting-started-dashboard.png)
155+
156+
## Step 8: Displaying Data Visualizations
147157

148158
HoloViz MCP includes a powerful display tool that can render visualizations directly. Ask Claude:
149159

@@ -153,6 +163,8 @@ Use the holoviz_display tool to show me a simple hvplot visualization of random
153163

154164
Claude will use the display tool to generate and display the visualization. See the [Display System tutorial](display-system.md) for more details.
155165

166+
![DataViz Displayed](../assets/images/getting-started-display.png)
167+
156168
## What's Next?
157169

158170
Now that you have HoloViz MCP running with Claude Code, explore more:
@@ -214,13 +226,13 @@ For more help, see the [Troubleshooting Guide](../how-to/troubleshooting.md) or
214226

215227
In this tutorial, you:
216228

217-
✅ Installed HoloViz MCP using uv
218-
✅ Created the documentation index
219-
✅ Installed Chromium
220-
✅ Configured Claude Code
221-
✅ Verified the installation
222-
✅ Built your first Panel dashboard
223-
✅ Learned about the display tool
224-
✅ Learned how to work on projects with Claude Code
229+
- ✅ Installed HoloViz MCP using uv
230+
- ✅ Created the documentation index
231+
- ✅ Installed Chromium
232+
- ✅ Configured Claude Code
233+
- ✅ Verified the installation
234+
- ✅ Built your first Panel dashboard
235+
- ✅ Learned about the display tool
236+
- ✅ Learned how to work on projects with Claude Code
225237

226238
You're now ready to use HoloViz MCP with Claude Code to accelerate your Panel development from the command line! Happy coding! 🚀

docs/tutorials/stock-analysis-claude-code.md

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -7,15 +7,16 @@ By the end, you'll have built an interactive report that displays financial data
77
!!! tip "What you'll learn"
88
- How to use Claude Code to plan and build data applications
99
- How to use the `holoviz_display` tool to quickly visualize and persist your work
10-
- How to work with stock data using yfinance
10+
- How to work with stock data using `yfinance`
1111
- How to iterate on visualizations using natural language
1212

1313
!!! note "Prerequisites"
1414
Before starting, ensure you have:
1515

1616
- Claude Code CLI installed and configured ([Getting Started Guide](getting-started-claude-code.md))
1717
- HoloViz MCP server configured with Claude Code
18-
- `yfinance` installed: `pip install yfinance`
18+
- HoloViz agents installed `holoviz-mcp install claude`
19+
- `panel` and `yfinance` installed: `pip install panel yfinance`
1920

2021
## Step 1: Plan Your Report
2122

0 commit comments

Comments
 (0)