​
Prerequisites

1

Create Relace Repo from Template

We start by creating a Relace Repo with our public template repository for data analysis.
import requests

# Create a new repository from the base template
url = "https://api.relace.run/v1/repo"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}

data = {
    "source": {
        "type": "git",
        "url": "https://github.com/squack-io/tailwind-vite-base",
        "branch": "main"
    }
}

response = requests.post(url, headers=headers, json=data)
repo = response.json()
repo_id = repo['repo_id']

print(f"Repository created with ID: {repo_id}")
This base repository is in Vite.js with support for React and Tailwind CSS. It can be forked and customized with custom themes & components.
2

Upload Structured Data

Next, we’ll upload our data files to the public folder of the Relace Repo.
# Upload your data file(s) to the repo
with open('./your_data_file.csv', 'rb') as f:
    file_data = f.read()

upload_url = f"https://api.relace.run/v1/repo/{repo_id}/file/public/data/your_data_file.csv"
upload_headers = {
    "Authorization": headers["Authorization"],
    "Content-Type": "application/octet-stream"
}

upload_response = requests.put(
    upload_url,
    headers=upload_headers,
    data=file_data
)

if upload_response.status_code in [200, 201]:
    print("Successfully uploaded your_data_file.csv")
else:
    print(f"Error uploading file: {upload_response.status_code}")
Relace agents are focused on codegen, so the files you upload should be structured data (i.e. csv, json, etc). Any unstructured data like txt, docx, or xlsx files should only be included if they are small and can be read in one shot.
3

Run Analysis Agent

Now we’ll execute the analysis agent with a custom query and system prompt. The agent will analyze your data and create an interactive dashboard.
import json

# Define agent parameters
agent_name = "analysis-agent"
query = "Analyze the uploaded data to identify key trends, patterns, and insights. Create interactive visualizations and a comprehensive dashboard."

system_prompt = """You are a data analysis expert. Analyze the uploaded data files to:
1. Identify key patterns and trends in the data
2. Calculate relevant statistics and metrics
3. Create interactive visualizations using Chart.js or similar libraries
4. Build a comprehensive dashboard with multiple views
5. Highlight any notable anomalies or insights
Present your findings in a professional, business-focused format with clear visualizations."""

# Call the agent API with streaming enabled
agent_url = f"https://api.relace.run/v1/repo/{repo_id}/agent"
agent_payload = {
    "agent_name": agent_name,
    "agent_inputs": {
        "query": query,
        "system_prompt": system_prompt
    },
}

response = requests.post(
    agent_url,
    headers=headers,
    json=agent_payload,
    stream=True
)
The agent will return an SSE stream with execution events and tool calls as a response. See the API reference for more details about how to make use of it.
4

Access Your Dashboard

Once the agent completes execution, your interactive dashboard will be deployed and accessible via a public URL.
# The deployed site URL follows this pattern:
dashboard_url = f"https://{repo_id}.dev-server.relace.run"
print(f"Your dashboard is live at: {dashboard_url}")
The generated dashboard can be embedded as an iframe in other applications to easily view results.
<iframe
  src="https://your-repo-id.dev-server.relace.run"
  width="100%"
  height="600px"
  frameborder="0"
>
</iframe>
5

Iterate and Refine Visualization

You can continue refining your analysis by sending additional requests to the same repository:
# Send follow-up analysis requests
followup_query = "Add a time-series analysis view and export functionality for the data insights"

followup_payload = {
    "agent_name": agent_name,
    "agent_inputs": {
        "query": followup_query,
        "system_prompt": "Enhance the existing dashboard with additional analysis features and export capabilities."
    },
    "overrides": {}
}

followup_response = requests.post(
    agent_url,
    headers=headers,
    json=followup_payload,
    stream=True
)