Grab your coffee. Here are this week’s highlights.

📅 Today’s Picks

kotaemon – Build Private Document QA with Citations

Problem

ChatGPT’s file upload is useful for quick questions, but it is not built for shared document workflows.

Teams often need page-level citations, shared document collections, user access control, and a way to keep sensitive files inside their own infrastructure.

Solution

kotaemon provides a self-hosted document QA interface for that.

It lets teams ask questions across documents, verify answers in a built-in PDF viewer, and organize files into private or public collections.

Key benefits:

• Finds answers with both keyword search and semantic search
• Reasons across multiple documents for more complex questions
• Runs on your own infrastructure with one Docker command
• Works with OpenAI, Anthropic, and local models

Graphify – Query a Codebase as a Knowledge Graph

Problem

When asking repo-level questions, the usual workflow is to search with grep, jump between definitions, read docs, and paste context back into the chat.

But this workflow takes time, makes indirect connections easy to miss, and forces the assistant to load more context than needed.

Solution

Graphify turns a project folder into a local knowledge graph that maps those relationships, so follow-up questions become cheaper and easier to answer accurately.

The following example shows how Graphify traces the shortest path between two scikit-learn components without manually opening files.