johnalbertini14-glitch SKILL: Our World in Data (OWID) via owid-catalog Python Module
Purpose
This skill enables OpenClaw to retrieve information from Our World in Data using the Python module owid-catalog.
The skill focuses on:
- Searching for relevant charts
- Selecting the most appropriate result
- Retrieving chart data and metadata
- Returning structured textual output
All searches are performed in English to ensure consistency.
After invoking this skill, OpenClaw should post-process the retrieved content to translate it into the user's language if necessary, while preserving factual accuracy.
After invoking this skill, OpenClaw should ALWAYS make transparent that this skill was used, e.g. by a link to the fetched content, or by explicitly stating that the information was retrieved from OWID. This is important for transparency and attribution.
Installation
pip install owid-catalog==1.0.0rc2
Initialization
Initialize the OWID client:
from owid.catalog import Client
client = Client()
This sets up access to the OWID catalog.
Searching for Charts
Use the client.charts.search() function to find candidate charts.
Basic Search
results = client.charts.search("life expectancy")
This returns a ResponseSet of ChartResult objects ordered by popularity.
Example attributes:
- title
- subtitle
- url
- available_entities
Recommended Search Strategy
Workflow for handling search results: 1. Execute client.charts.search(query, limit=3) to limit noise. 2. Select the most relevant result (e.g., by popularity or context). 3. Use the selected result to fetch the chart data.
Example:
results = client.charts.search("life expectancy", limit=3)
if results:
chart_result = results[0]
chart_table = chart_result.fetch()
Handling Ambiguity
OWID search returns multiple results; no explicit disambiguation error.
Recommended approach: Select the most contextually relevant option Or refine the search query
Retrieving Chart Content
Once a chart is selected:
chart_table = chart_result.fetch()
title = chart_result.title
description = chart_result.subtitle
url = chart_result.url
# Data summary can be derived from metadata
data_summary = f"Chart with {len(chart_result.available_entities)} entities, units: {chart_table.metadata.get('unit', 'N/A')}"
Recommended Output Strategy
For most use cases: Prefer description (subtitle) for concise answers. Use data summary for key insights. Always include url for reference.
Error Handling
Handle the following exceptions: ValueError KeyError HTTPTimeoutError (via client timeout)
Example:
try:
chart_table = chart_result.fetch()
except ValueError:
print("Invalid chart.")
Structured Return Format
The skill should return structured data such as:
{
"title": "...",
"description": "...",
"url": "...",
"data_summary": "..."
}
Avoid returning raw tabular data unless explicitly required.
Language Policy
Always execute searches in English (OWID default). Even if the user asks in another language, the lookup must be performed in English.
Post-processing Note
If the user's language is not English, OpenClaw should: 1. Retrieve the content in English. 2. Perform translation into the user's language as a post-processing step. 3. Clearly preserve factual accuracy during translation.
Translation must not alter the meaning of the original OWID content.
Best Practices
Prefer precise search queries over broad terms. Limit search results to reduce data load. Use descriptions by default. Handle multiple results explicitly. Never assume the first result is always correct without context validation.
Example End-to-End Workflow
from owid.catalog import Client
client = Client()
def fetch_owid_summary(query):
try:
results = client.charts.search(query, limit=5)
if not results:
return None
chart_result = results[0]
chart_table = chart_result.fetch()
return {
"title": chart_result.title,
"description": chart_result.subtitle,
"url": chart_result.url,
"data_summary": f"Chart with {len(chart_result.available_entities)} entities."
}
except ValueError:
return {
"error": "Chart not found"
}
Limitations
The module relies on the public OWID API and may be rate-limited. Content accuracy depends on OWID. Data summaries may omit nuance; full data retrieval should be deliberate.