Search Output

A living dataset of implementations, use cases, and impact signals for open-source tools built and maintained by Development Seed.

Why this exists

We build and maintain open-source geospatial libraries used by teams around the world — but we've never had a reliable way to understand the full scope of that impact. Who is using our tools? How are they deploying them? What problems are they solving?

This repo is an experiment to answer that question systematically. An automated agent monitors the web daily, surfacing implementations, integrations, blog posts, and other signals that show how our work is being used in the wild.

What's here

Folder	Tool	Description
data/titiler/	TiTiler	GitHub repos, PyPI dependents, and web results referencing TiTiler

More tools will be added over time.

How it works

An OpenClaw agent runs on a remote VM and searches three sources daily:

• GitHub Search API — repositories referencing the tool
• PyPI dependents — packages that list it as a dependency
• Brave Search API — broader web results including blog posts and deployments

New results are appended to the relevant CSV and pushed to this repo each morning. Duplicates are automatically filtered.

Technical details

The crawler runs as an isolated claw system user on a Hetzner CPX31 VM (Ubuntu 24), separate from the main development environment. It has no access to project repositories, SSH keys, or GitHub credentials.

Agent setup

• OpenClaw runs as a persistent gateway in a tmux session, controlled via Telegram
• Model: qwen2.5:3b via Ollama (local inference, no external API calls for the agent itself)
• Shell tool is disabled — the agent cannot execute arbitrary commands, which limits the blast radius of any prompt injection from scraped content

Crawler

The crawler is a standalone Python script registered as a custom OpenClaw skill. It runs inside a Python venv and has three source modules:

GitHub Search API (source: github)

• Unauthenticated requests to api.github.com/search/repositories
• Query: titiler, sorted by last updated
• Paginates up to 3 pages (90 results max per run)
• Rate limit: 10 req/min unauthenticated

PyPI dependents (source: pypi)

• Fetches package metadata from pypi.org/pypi/{package}/json
• Scrapes github.com/{org}/{repo}/network/dependents for downstream packages

Brave Search API (source: brave)

• Three queries per run: "titiler" site:github.com, titiler implementation example, titiler deployment
• 10 results per query, ~30 results per run
• Free tier: $5/month credit (~1,000 queries/month)

Dedup and output

Before appending, the script loads all existing URLs from the CSV into a set and skips any result already present. New results are written to the relevant data/{tool}/implementations.csv file in this repo.

Sync

A separate cron job runs as the anthony user at 6:05am UTC, 5 minutes after the crawler. It runs git add data/ && git commit && git push to sync new results to this repo. If there are no new results, it exits without committing.

Schedule

Time (UTC)	User	Action
6:00am	claw	Run crawler, append to CSV
6:05am	anthony	Commit and push to GitHub

CSV columns

Column	Description
url	Link to the implementation or resource
repo_name	Repository or page name
source	Where it was found (github, pypi, brave)
discovered_at	UTC timestamp when first found
description	Short description
stars	GitHub stars (if applicable)
last_updated	Last activity date (if applicable)

Contributing

Have a tool you want to monitor? Open an issue or PR with:

• Tool name
• Search queries to use
• Target sources (GitHub / PyPI / web)