This is intended to serve as a proof of concept for features that I would like to see in other chatbots, such as:
- Conversation export/import to/from JSON
- Editing of responses
- The ability to edit old prompts in their old context
- Easy deletion
- No censorship
Even without the last point, this is a combination that is absent from all major LLM chat interfaces that I have tried, which include ChatGPT, Google Gemini, Mistral, Perplexity, Openrouter.ai Playground, Claude.ai and Poe. You often have a few of these features, but not all of them together.
This requires python and the flask and requests python libraries. If they are not installed on your machine, you may install them in a virtual environment from a POSIX shell:
python -m venv /path/to/venv
source /path/to/venv/bin/activate
pip install flask requests
Either export GEMINI_CHAT_API_KEY=YOUR_API_KEY or
echo YOUR_API_KEY > ${HOME}/.gemini-chat-api-key, replacing YOUR_API_KEY
with your gemini API key. The former will need to be done before future
sessions while the latter will persist across sessions. The environment
variable takes precedence over the key file.
If you do not have an API key, you may get one at:
https://makersuite.google.com/app/apikey
Then simply run:
python /path/to/app.py
You may then go to http://localhost:5000 in your web browser.
Future launches would require running source /path/to/venv/bin/activate to enter the virtual environment before running the python command if you install the prerequisite libraries through it.
- Google sometimes returns 500 errors.
As this is a proof of concept, I have no current plans for revisions, but I can think of a few things that I would do if I pursued them:
- Make the interface nicer.
- Add option to cancel in-progress LLM query.
- Add support for more LLMs, and dynamic LLM switching.
- Add support for modifying temperature, topP and topK.
- Make an extensible JSON format rather than the naive one currently implemented, plus make a conversion tool for old saves.