Metadata-Version: 2.1
Name: flipgenic
Version: 2.1.1
Summary: High-speed conversational dialogue engine
Home-page: https://github.com/danth/flipgenic
Author: Daniel Thwaites
Author-email: danthwaites30@btinternet.com
License: UNKNOWN
Project-URL: Bug Reports, https://github.com/danth/flipgenic/issues
Project-URL: Source, https://github.com/danth/flipgenic
Keywords: conversation dialogue engine chat chatbot
Platform: UNKNOWN
Classifier: Development Status :: 3 - Alpha
Classifier: License :: OSI Approved :: GNU Affero General Public License v3
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.6
Classifier: Programming Language :: Python :: 3.7
Requires-Python: >=3.6,<4
Description-Content-Type: text/markdown
Provides-Extra: docs
License-File: LICENSE

<p align="center">
  <img
    src="images/header.png"
    alt="Flipgenic: High-speed conversational dialogue engine"
  />
</p>

---

## What is it?

Flipgenic is a library for creating simple chatbots. A Flipgenic bot starts
with an empty mind and learns how to communicate over time by reusing messages
it has seen in previous conversations. It does not write anything original.

Here is an example from its use in [Axyn](https://github.com/danth/axyn):

![Screenshot of example conversation](images/Screenshot_20200426_124703.png)

Flipgenic is focused on fast reply times and a simple API.
[Chatterbot](https://chatterbot.readthedocs.io/en/stable/), a more well-known
library, has additional features such as custom filters and logic adapters.

## How do I use it?

Here is a basic example:

```python
# python -m pip install flipgenic
# python -m spacy download en_core_web_md

from flipgenic import Responder

# Create and connect to database
# This can take a while to load the spaCy models
responder = Responder('/path/to/data/folder/')

# Initialize the database with a single response
responder.learn_response('Hello', 'Hi!')

response = None
while True:
    text = input('> ')

    if response:
        # Learn the input as a response to the previous output
        responder.learn_response(response, text)

    # Generate a response
    response, distance = responder.get_response(text)
    print(response.text, f'({distance})')
```

**For more, see [ReadTheDocs](https://flipgenic.readthedocs.io/en/latest/quickstart.html).**

## How does it work?

1. Upon calling `get_response`, the input is converted to a 300-dimensional
   vector. The vector embeddings for each word are provided by
   [SpaCy](https://spacy.io/api/token#vector); calculating the mean gives a
   single vector representing the whole text.
2. A closest match is found by searching the vectors of previously seen
   messages. An [NGT](https://github.com/yahoojapan/NGT) index improves
   performance.
3. The search result corresponds to one or more known responses. These
   responses are stored in an SQLite database.
4. We select a matching response at random. The same response might have been
   learned more than once, giving it a higher chance of selection.
5. Along with the selected response, the distance between the input vector and
   the search result is returned as a confidence heuristic.


