Documentation-related tasks

[ludwiktrammer]

Symbols

📃 - should become a ticket
☑️ - ticket created (with a link)

1. Minimal READMEs for All Subpackages ☑️

Each README should include:

• A short description
• A short "Quickstart" code snippet
• Installation instructions
• A link to the relevant documentation

These READMEs should be configured to display correctly on both GitHub and PyPI.

2. Table of Contents for the Main Documentation Page ☑️

The main documentation page should feature a table of contents with links and descriptions for each section.

3. New Quickstart Chapter: Evaluations with ragbits-evaluate ☑️

A new chapter should be added to the Quickstart guide, building on examples from previous chapters.

4. Improved Cross-Referencing and linking ☑️

• Ensure the Quickstart guide links to other relevant articles where necessary:
  • CLI commands should link to the CLI section.
  • Concepts such as prompts, ingestion, or document search should link to relevant How-Tos.
• Ensure that every mention of a specific class anywhere in the documentation is linked to the specific API documentation.

5. How-To: Default Configuration for Ragbits ☑️

This guide should cover:

• Using both YAML configuration and default factory functions
• A general section on configuring defaults for any type, with a table mapping class types to their configuration keys
• Dedicated sections for LLMs and Document Search due to their unique behavior

6. Additional How-Tos on Prompts

• Using images in prompts: defining and passing images to prompts. ☑️
• Creating custom output parsers for prompts: declaring a prompt with a custom output data type and providing a parser. 📃
• Using custom prompt sources: extending BasePrompt to enable dynamic prompts from external sources (e.g., YAML files or a database). 📃

7. How-Tos on LLMs

• Using LLMs with Ragbits: Configuring and using both local and remote LLMs, calling LLM classes with Prompts and raw strings/conversations. ☑️
• Writing support for a custom LLM. 📃

8. How-Tos on Embeddings

• Creating embeddings with Ragbits, covering both local and remote models (via LiteLLM). 📃
• Writing support for a custom embedding model. 📃

9. How-Tos on Vector Stores

• Using vector stores with Ragbits, including: 📃
  • Creating and configuring vector stores.
  • Adding vectors, listing, and querying.
  • Using both Python and CLI commands, when possible.
  • Links to sub-articles on configuring specific vector stores:
    • ChromaDB
    • Qdrant
    • Pgvector
• Writing support for a custom vector store. 📃

10. How-Tos on Document Search

• Rewriting the document search guide to focus exclusively on searching, linking to a separate ingestion article. It should show how to search for documents using Ragbits, both from Python and CLI, and provide an example of using retrieved documents in a RAG setup. 📃
• How to ingest documents, covering: 📃
  • Overview of the ingestion process.
  • Configuration via Python, factory functions and YAML.
  • Overview of all objects involved in Document Search (processors, vector stores, rerankers, etc.), with links to detailed guides.
  • Ingesting different document types from various sources.
  • Demonstrating the steps via both Python and CLI, whenever possible.
• Choosing ingestion sources, with sections on each available source type and their configurations. 📃
• Writing custom sources for document ingestion. 📃
• Using Unstructured for document ingestion. 📃
• Writing custom document processors 📃, including configuring them to work with processor routers.
• Ingesting images and PDFs 📃, covering:
  • Enabling/disabling OCR.
  • Enabling/disabling LLM-based image descriptions.
  • Choosing between embedding them as text, image, or doing both.
• Explaining Document Search terminology 📃 (Documents, DocumentMeta, Sources, VectorStore, Elements, Processors, etc.).
• Using query rephrasers 📃, including custom implementations.
• Using rerankers 📃, including defining custom ones.

11. Updating How-Tos on Evaluation

• Ensure all articles reflect the new high-level APIs. 📃
• Expand the current brief overview articles 📃 to include step-by-step guides with embedded code examples.

12. How-Tos on Guardrails

• Verifying messages with guardrails 📃
• Creating custom guardrails 📃

13. How-To: Using Tracing in Ragbits ☑️

• Configuring tracing with OpenTelemetry and CLI.

14. Add missing public classes to API Reference 📃

This includes all public classes from packages such as Evaluate, Guardrails, etc.