WIP

Author: johann-petrak

.github/workflows/static.yml .github/workflows/OLD-static.yml.bak.github/workflows/static.yml renamed to .github/workflows/OLD-static.yml.bak

@@ -0,0 +1 @@
+gem "github-pages", group: :jekyll_plugins

docs/Gemfile

@@ -0,0 +1,8 @@
+remote_theme: pages-themes/cayman@v0.2.0
+plugins:
+- jekyll-remote-theme # add this line to the plugins list if you already have one
+
+title: python-llms-wrapper
+description: A library for making it simple and straightforward to use many different LLMs, including tooling
+
+show_downloads: false

docs/_config.yml

@@ -0,0 +1,15 @@
+Github help:
+* https://docs.github.com/en/pages/setting-up-a-github-pages-site-with-jekyll/about-github-pages-and-jekyll
+* https://jekyll.github.io/github-metadata/site.github/
+* https://pages.github.com/themes/
+  * https://github.com/pages-themes/cayman
+  * https://github.com/pages-themes/midnight
+  * https://github.com/jekyll/minima
+  * https://github.com/pages-themes/minimal
+  * https://github.com/pages-themes/slate
+  * https://github.com/pages-themes/tactile
+* Customize theme: https://docs.github.com/en/pages/setting-up-a-github-pages-site-with-jekyll/adding-a-theme-to-your-github-pages-site-using-jekyll
+
+
+Test locally: see theme readme, https://github.com/pages-themes/cayman
+Then run: bundle exec jekyll serve

docs/_notes.txt

@@ -0,0 +1,3 @@
+#!/bin/bash
+
+bundle exec jekyll serve

docs/_run.sh

@@ -0,0 +1,73 @@
+For every LLM to be used in a project, there should be configuration data that is necessary for using the LLM.
+Configuration data is expected as a dictionary which contains the key `llms` associated with a list of per-LLM configurations
+and optionally the key `providers` associated with a dictionary mapping provider names to settings common to all LLMs of that 
+provider. 
+
+Here is an example configuration showing a subset of the settings:
+```
+{
+   "llms" : [
+      {
+         "api_key_env" : "OPENAI_KEY1",
+         "llm" : "openai/gpt-4o",
+         "temperature" : 0
+      },
+      {
+         "llm" : "gemini/gemini-1.5-flash",
+         "temperature" : 1,
+         "alias": "gemini1"
+      },
+      {
+         "llm" : "gemini/gemini-1.5-flash",
+         "temperature" : 0,
+         "alias": "gemini2"
+      }
+   ],
+   "providers" : {
+      "gemini" : {
+         "api_key_env" : "GEMINI_KEY1"
+      }
+   }
+}
+```
+
+Each llm is identified by the provider name, e.g. "openai" or "gemini" followed by a slash, followed by the model name or id. Provider 
+names must match the provider names known in the [litellm](https://docs.litellm.ai/docs/providers) package. 
+
+Parameters specified for each of the providers in the `providers`  section apply to every llm in the `llms`  section unless the same 
+parameter is also specified for the llm, in which case that value takes precedence.
+
+The following parameters are known and supported in the `llms` and/or `providers` sections:
+
+* `llm` (`llms` section only): specifies a specific model using the format `providername/modelid`. 
+* `api_key`: the literal API key to use
+* `api_key_env`:  the environment variable which contains the API key
+* `api_url`: the base URL to use for the model, e.g. for an ollama server. The URL may contain placeholders which will get replaced with
+   the model name (`${model}`), or the user and password for basic authentication (`${user}`, `${password}`), e.g.
+   `http://${user}:${password}@localhost:11434`
+* `user`, `password`: the user and password to use for basic authentication, this requires `api_url` to also be specified with the 
+   corresponding placeholders
+* `alias` (`llms` section only): an alias name for the model which will have to be used in the API. If no `alias` is specified, the name
+   specified for `llm` is used. 
+* `num_retries`: if present, can specify the number of retries to perform if an error occurs before giving up
+* `timeout`: if present, raise timeout error after that many seconds
+
+All other settings are passed as is to the model invocation function. Different providers or APIs may support different parameters, but 
+most will support `temperature`, `max_tokens` and `top_p` 
+
+IMPORTANT: the raw configuration as shown in the example above needs to get processed by the function `llms_wrapper.config.update_llm_config` 
+in order to perform all the necessary substitutions!
+
+#### Other top-leve config entries
+
+Currently the following other top level configuration fields in addition to `llms` and `providers` are recognized:
+
+* `use_phoenix`: if present, should be the URL of a local phoenix endpoint or a list containing the endpoint URL and the project name
+
+## Configuration files
+
+The configuration as described above can be created programmatically or read in from some config file. 
+The function `llms_wrapper.config.read_config_file`  allows reading the configuration from files in any of the following 
+formats: json, hsjson, yaml, toml. By default, reading the configuration that way will also perform the necessary substitutions by
+automatically invoking `llms_wrapper.config.update_llm_config` 
+

docs/configuration.md

@@ -0,0 +1,18 @@
+# python-llms-wrapper
+
+The `llms_wrapper` Python package is an opnionated library based on the excellen 
+[`litelmm`](https://github.com/BerriAI/litellm) package to simplify using diverse
+LLMs from many different providers. 
+
+This documentation contains the following sections:
+
+* [Installation](installation)
+* [Quick Start / Examples](quickstart)
+* [Configuration](configuration)
+* [Test Script](test-script)
+* [Usage](usage)
+* [Python Docs](pythondoc/llms_wrapper)
+
+LLMs currently supported: support is based on the [LiteLLM](https://github.com/BerriAI/litellm) and the supported LLMs are listed [here](https://docs.litellm.ai/docs/providers/)
+
+

docs/index.html

@@ -0,0 +1,15 @@
+# Installation
+
+### 1. Python environment 
+
+* If necessary create a new python environment
+  * e.g. using Ana/Miniconda: `conda create -y -n llms_wrapper python=3.11`
+
+### 2. Install package
+
+* From local source:
+  * clone the repository and change into the directory
+  * run `pip install -e .` 
+  * To install for development: `python -m pip  install -e .[dev]`
+* From source directly from github: `pip install -U git+https://github.com/OFAI/python-llms-wrapper.git`
+* To create a notebook kernel run `python -m ipykernel install --user --name=llms_wrapper`