pl-sclai is an incidental ChRIS ds plugin that offers an interactive terminal based session with various LLMs.
The standard web-based interfaces to various (typically commercial) LLMs are convenient for simple interactions, but lack many more "power user" features. As the context of the "chat" becomes more and more complex (for instance multiple source files in a coding project) LLMs have been shown to suffer from several undesirable behaviors. For instance, "prompt directive decay" (where the LLM gradually ignores or "forgets" specific guidelines) and "consistency drift" (the LLM offers numerous slightly different versions of the same source file as the interaction continues).
For example, a prompt directive of "always use docstrings" will result in the next several code suggestions to adhere to this, but within a few iterations the model will revert (or "forget") to a default behavior. If a source file has been uploaded as a reference, the LLM will invariably suggest specific edits but will often when asked for the whole file, "forget" or "omit" large parts of simply refactor.
This application is an attempt to mitigate this: pl-sclai offers a text-based interface that easily allows the user to "re-inject" (reinforce) specific directives with each interaction (using convenient text markup shortcuts) and simplify the refreshing of source files with a base reference. Simply stated it is an attempt to keep the context as current as possible with minimal disruption to the end user.
While pl-sclai is developed as ChRIS plugin this is mostly incidental. The plugin framework was used more as a means to simply setup a python (dockerizable) project than to create an application to use in ChRIS. In fact, given that pl-sclai is a strongly interactive application, it is not really suitable to run in ChRIS (where plugins are typically and strongly non-interactive).
The two vectors for use are either as a PyPI application installed into a virtual environment or run from a container.
To run on the metal directly on your system, use/create a virtual python env (say with uv):
cd ~/some/dir
uv initTo get started with local command-line usage, use Apptainer (a.k.a. Singularity) to run pl-sclai as a container:
apptainer exec docker://fnndsc/pl-sclai sclai [--args values...] input/ output/To print its available options, run:
apptainer exec docker://fnndsc/pl-sclai sclai --helpsclai requires two positional arguments as artifacts of the application framework factory. These are not needed or used right now and can simply be set to /tmp.
apptainer exec docker://fnndsc/pl-sclai:latest sclai [--args] /tmp /tmpInstructions for developers.
Build a local container image:
docker build -t localhost/fnndsc/pl-sclai .Mount the source code sclai.py into a container to try out changes without rebuild.
docker run --rm -it --userns=host -u $(id -u):$(id -g) \
-v $PWD/sclai.py:/usr/local/lib/python3.13/site-packages/sclai.py:ro \
-v $PWD/in:/incoming:ro -v $PWD/out:/outgoing:rw -w /outgoing \
localhost/fnndsc/pl-sclai sclai /incoming /outgoingRun unit tests using pytest.
It's recommended to rebuild the image to ensure that sources are up-to-date.
Use the option --build-arg extras_require=dev to install extra dependencies for testing.
docker build -t localhost/fnndsc/pl-sclai:dev --build-arg extras_require=dev .
docker run --rm -it localhost/fnndsc/pl-sclai:dev pytestSteps for release can be automated by Github Actions. This section is about how to do those steps manually.
Increase the version number in setup.py and commit this file.
Build and push an image tagged by the version. For example, for version 1.2.3:
docker build -t docker.io/fnndsc/pl-sclai:1.2.3 .
docker push docker.io/fnndsc/pl-sclai:1.2.3
Run chris_plugin_info
to produce a JSON description of this plugin, which can be uploaded to ChRIS.
docker run --rm docker.io/fnndsc/pl-sclai:1.2.3 chris_plugin_info -d docker.io/fnndsc/pl-sclai:1.2.3 > chris_plugin_info.jsonIntructions on how to upload the plugin to ChRIS can be found here: https://chrisproject.org/docs/tutorials/upload_plugin