https://github.com/writefreely/chatfreely
Generate brand new blog posts from the corpus of text on your own WriteFreely blog.
https://github.com/writefreely/chatfreely
ai generative-text writefreely
Last synced: 2 months ago
JSON representation
Generate brand new blog posts from the corpus of text on your own WriteFreely blog.
- Host: GitHub
- URL: https://github.com/writefreely/chatfreely
- Owner: writefreely
- License: gpl-3.0
- Created: 2023-04-01T16:48:14.000Z (about 2 years ago)
- Default Branch: main
- Last Pushed: 2023-04-01T20:25:40.000Z (about 2 years ago)
- Last Synced: 2024-10-29T20:22:08.691Z (8 months ago)
- Topics: ai, generative-text, writefreely
- Language: Go
- Homepage:
- Size: 23.4 KB
- Stars: 6
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# ChatFreely
ChatFreely is a super-intelligent, sentient and sweet AI. Just kidding. It's a program that generates brand new blog posts from the corpus of text on your own WriteFreely blog.
It uses a [Markov chain](https://en.wikipedia.org/wiki/Markov_chain) to generate often nonsensical but sometimes amusing new strings of words. It's functionally like ChatGPT, but more silly and without all the non-consensual training on vast amounts of intellectual property.
## Limitations
Like most modern "AI" tools:
* ChatFreely will confidently poop out language without any concern for truth or reality. It has no capacity for any kind of logical inference that would allow it to understand truth as humans do.
* It will produce instructions and content based on the data it was trained on, for better or worse.
* It has no knowledge of the world and events, in 2023 or in any other year. It does not have knowledge. It is a mathematical model coded into a computer program.
## Getting started
With Go installed, open a terminal and run:
```bash
go get github.com/writefreely/chatfreely
```
### Training
Next train the "AI" on your WriteFreely blog with the following command, replacing `[instance-url]` and `[blog-alias]` with your own.
The `-o` flag sets the order of your Markov model -- setting it to `2` makes it more coherent; setting it to `1` makes it more unhinged but more original.
```bash
chatfreely train -i [instance-url] -c [blog-alias] -o 2
```
For example:
```bash
chatfreely train -i write.as -c blog -o 2
```
#### Using with Write.as
Write.as implements rate-limiting on its post-retrieval API that [requires an application key](https://write.as/me/applications). Follow the instructions there to retrieve yours, and then set an environment variable, `WRITEAS_APP_KEY=your-app-key-here...`, before training your model.
### Generating
Finally, generate a brand new post, again specifying the order (`-o`) that you used to train the model:
```bash
chatfreely gen -i [instance-url] -c [blog-alias] -o 2
```
## Commands
```
NAME:
ChatFreely - Generative "AI" that learns from WriteFreely blogs.
USAGE:
chatfreely [global options] command [command options] [arguments...]
COMMANDS:
train Train the markov chain.
generate, gen Generate a blog post from training data.
help, h Shows a list of commands or help for one command
GLOBAL OPTIONS:
--help, -h show help
```
### train
Train the markov chain.
```
USAGE:
chatfreely train [command options] [arguments...]
OPTIONS:
--alias value, -c value Alias of the WriteFreely collection to train on
--instance value, -i value WriteFreely instance to train on
--order value, -o value Markov chain order (recommend 1 or 2)
--help, -h show help
```
### generate
Generate a blog post from training data.
```
USAGE:
chatfreely generate [command options] [arguments...]
OPTIONS:
--alias value, -c value Alias of the WriteFreely collection to train on
--instance value, -i value WriteFreely instance to train on
--order value, -o value Markov chain order (same as training)
--help, -h show help
```