Skip to content

Boo Documentation

Boo is a multi-provider AI assistant application built with Streamlit. It provides a single user interface for provider-backed text, image, audio, embedding, file, document, vector-store, and local data workflows.

Purpose

This documentation explains how to install, configure, run, maintain, and extend the Boo application. It also provides API reference pages generated from Google-style Python docstrings in the application source modules.

Application Scope

Boo is designed as a development and productivity workbench for multiple AI providers. The application centralizes common assistant workflows while preserving provider-specific capabilities where the underlying APIs differ.

The current documentation set covers:

Section Description
Getting Started Local setup, dependency installation, runtime configuration, and MkDocs commands.
Architecture Application structure, provider routing, module responsibilities, and data flow.
Application API Streamlit shell documentation and import-safety notes for app.py.
GPT API OpenAI GPT provider wrapper reference.
Gemini API Google Gemini provider wrapper reference.
Grok API xAI Grok provider wrapper reference.

Core Capabilities

Boo provides a common interface for the following workflows:

  • conversational text generation;
  • document question answering;
  • semantic search;
  • prompt engineering and prompt testing;
  • image generation or image analysis where supported by the selected provider;
  • audio generation, transcription, or translation where supported by the selected provider;
  • embeddings and vector-store workflows;
  • file upload, file retrieval, and file-based assistant operations;
  • local data utility workflows used during development and testing.

Provider Model

Boo uses provider wrapper modules to isolate provider-specific API syntax from the Streamlit user interface. This keeps the application shell easier to maintain because the UI can call stable wrapper methods rather than directly embedding provider SDK logic in the page rendering code.

The primary provider modules are:

Module Provider Typical Responsibilities
gpt.py OpenAI Text, images, audio, embeddings, files, vector stores, and assistant workflows.
gemini.py Google Gemini Text, multimodal generation, files, and Gemini-specific grounded workflows.
grok.py xAI Grok Text, image-capable workflows, embeddings where supported, files, and vector stores.
app.py Streamlit Page layout, session state, provider selection, workflow routing, and display logic.

Documentation Generation

The API reference pages use mkdocstrings directives such as:

::: gpt

The documentation build expects the source modules to be importable from the repository root. If a module executes UI code, network calls, or provider client initialization at import time, MkDocs may fail during API generation. In that case, move import-safe reusable logic into separate modules and document those modules instead.

Read the documentation in this order when onboarding to the project:

  1. Getting Started
  2. Architecture
  3. Application API
  4. GPT API
  5. Gemini API
  6. Grok API

This order starts with operational setup, then explains the design, and finally moves into the source-level API reference.