opensourceprojects.dev

A broadsheet for software that doesn't ask for your email

Self-hosted code search and codebase Q&A with inline citations
GitHub RepoImpressions2

Project Description

View on GitHub

Stop Grepping Through Ten Repos: Sourcebot Brings Search and Q&A to Your Whole Codebase

You know the drill. You need to find where a function is defined, but it's somewhere across a dozen repositories, three branches, and two code hosts. So you clone what you can, fire up ripgrep, and hope you remembered the right naming convention. Sourcebot is a self-hosted tool that wants to make that whole process go away—search across everything you've got, and ask questions in plain English when search alone won't cut it.

What It Does

Sourcebot is a self-hosted tool for understanding your codebase. It does two main things. First, it gives you code search across all your repos and branches, regardless of where they're hosted—so you're not bouncing between GitHub, GitLab, and whatever else your team accumulated over the years. Search supports regular expressions, repo and language filters, and boolean logic.

Second, there's "Ask Sourcebot," which lets you ask complex questions about your codebase in natural language. It works by leaning on Sourcebot's own search and navigation tools, letting reasoning models search your code, follow code navigation references, and produce an answer with inline citations and navigable code snippets. That last part matters: the answers are grounded in actual references you can click through, not just a confident-sounding paragraph.

On top of that, you get IDE-level code navigation (goto definition and find references) across all your repos, plus a built-in file explorer with syntax highlighting, a file tree, and code navigation. It deploys via Docker Compose, and the configuration is a JSON file where you specify what repositories to index, language model providers, auth providers, and more.

Why It's Cool

The citations are the whole point. Plenty of tools will let you chat with a codebase and hand you an answer that sounds right. Sourcebot's approach—using its own search and navigation tools as the retrieval layer, then citing inline—means you can verify the answer instead of trusting it. For code, that's not a nice-to-have. It's the difference between a useful tool and a liability.

It's not picky about where your code lives. The README is explicit that search works across any code host platform. If your organization has repos scattered across providers (and whose doesn't?), that's a real problem solved rather than a demo-friendly one.

Search and Q&A share the same foundation. Because Ask Sourcebot is built on the same search and code navigation machinery you use directly, you're not maintaining two separate mental models. The tool that answers your question is the tool you already use to find things.

Code navigation across repos is underrated. Goto definition and find references are things you take for granted inside a single project in your IDE. Getting that across every repo you own is a genuinely different capability, and it's the kind of thing that quietly saves you time every week.

Self-hosted, with a demo to kick the tires first. You can point a browser at the public demo before committing to anything, then deploy on your own infrastructure when you're ready. The Docker Compose path keeps the setup short.

How to Try It

Deployment is meant to take seconds with Docker Compose. Grab the compose file:

curl -o docker-compose.yml https://raw.githubusercontent.com/sourcebot-dev/sourcebot/main/docker-compose.yml

Then, in the same directory, create a configuration file. It's JSON and it controls Sourcebot's behavior—which repositories to index, language model providers, auth providers, and so on:

echo '{
    "$schema": "https://raw.githubusercontent.com/sourcebot-dev/sourcebot/main/schemas/v3/index.json",
    // Comments

The README's example is truncated, so head to the configuration docs for the full picture. There's also a Docker Compose deployment guide if you want the complete walkthrough.

If you'd rather look before you leap, the public demo is the fastest way to see whether the search and the Q&A behave the way you'd want. And if something's missing, the project has an open roadmap, plus templates for bug reports and feature requests.

The repo lives at github.com/sourcebot-dev/sourcebot.

Final Thoughts

Sourcebot is aimed at teams whose code sprawls—multiple repos, multiple branches, multiple hosts—and who've felt the pain of not being able to see all of it at once. The inline citations are what make the Q&A feature worth taking seriously rather than dismissing as another chatbot wrapper. It's early enough that the README itself is still filling in details, so expect to spend some time in the docs during setup. But if "where is this defined, and what calls it?" is a question you ask often, this is worth an afternoon.

Back to Projects
Project ID: 4e3eb7f0-a918-4486-ba62-cc5e3e4e79beLast updated: September 30, 2026 at 02:46 AM