https://github.com/ctxs-ai/ctxs.ai
an open-source, community-curated directory of contexts for use with LLMs
https://github.com/ctxs-ai/ctxs.ai
ai llm prompt-engineering
Last synced: 6 days ago
JSON representation
an open-source, community-curated directory of contexts for use with LLMs
- Host: GitHub
- URL: https://github.com/ctxs-ai/ctxs.ai
- Owner: ctxs-ai
- License: mit
- Created: 2025-01-28T11:03:13.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-04-08T22:05:53.000Z (7 days ago)
- Last Synced: 2025-04-08T23:20:25.319Z (7 days ago)
- Topics: ai, llm, prompt-engineering
- Language: TypeScript
- Homepage: https://ctxs.ai
- Size: 972 KB
- Stars: 20
- Watchers: 5
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
Awesome Lists containing this project
- awesome - ctxs-ai/ctxs.ai - an open-source, community-curated directory of contexts for use with LLMs (TypeScript)
README
Users can contribute context data as markdown files via GitHub pull requests.
[Rationale](#rationale) | [Use Cases](#use-cases) | [Contributing](CONTRIBUTING.md) | [Interface](#interface)
# Rationale
- LLMs are trained with huge datasets, constrained by a training cut off date
- Providing more precise context about a project, libraries used, available tools and frameworks improves model performance
- IDEs like Cursor are starting develop their own indexing system but this makes it hard to share what’s working
- Serve as an open-source “deterministic checkpoint” that can be used for evals
# Use Cases
Generally this exists to share context windows and prompts, see what works for others and learn from it.
Specifically for coding use cases:
- Help LLMs work more efficiently in a specific framework
- Define coding style, conventions and formatting
- Provide library documentation in a format suited for LLMs
# How to contribute
See [CONTRIBUTING.md](CONTRIBUTING.md)
# Interface
[`ctxs.ai`](https://ctxs.ai) serves as a public endpoint for the content of this repo.
The URL structure is as follows:
- `/gh/{username}/{context}` for HTML representations
- `/gh/{username}/{context}.txt` for plaintext
- `/r/gh/{username}/{context}.json` for shadcn-style registry
- `/r/index.json` contains all available contexts
## `ctxs` command line
`npx ctxs add` will allow you to add a selection of context windows to your project.
On the website you will also be able to copy `ctxs` commands that add files to your project. By default files will be written to `ctxs/` in your current directory. Some files will be written to a designated path (e.g. Cursor rules, [example](https://github.com/ctxs-ai/ctxs.ai/blob/main/contexts/danhollick/tailwind-css-v4.md?plain=1)).
```sh
npx ctxs add "https://ctxs.ai/r/gh/danhollick/tailwind-css-v4.json"
```
## HTTP endpoints
For any context on the site you can append `.txt` to the URL to get a plaintext version of the context. This URL can be used with LLMs as a token-efficient way of providing context.
```sh
curl https://ctxs.ai/gh/martinklepsch/babashka.txt
```
Note that currently there is no versioning and context may change. For regular, unsupervised use it is recommended to download contexts in order to preempt prompt injection.
# Inspirations
- Jim Lowin [The Qualified Self](https://www.jlowin.dev/blog/the-qualified-self)
- llms.txt and [llmstxt.directory](https://llmstxt.directory/)
- PatrickJS' [awesome-cursorrules](https://github.com/PatrickJS/awesome-cursorrules)
- Daniel Rosehill Prompt Library (read the [README](https://github.com/danielrosehill/Prompt-Library), it's great)
- `shadcn` CLI experience
# Like this project?
🌟 Star the repo and contribute! 🌟
# Available Contexts
Browse via [ctxs.ai](https://ctxs.ai)