Comments URL: https://news.ycombinator.com/item?id=49586878

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=49185355

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=49163679

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=49081594

Points: 41

# Comments: 76

Comments URL: https://news.ycombinator.com/item?id=49068796

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48779928

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48673117

Points: 2

# Comments: 0

tl;dr At work, I needed to find every repository that uses deprecated Node 16 and make a batch update. This process was exhausting, I had to find them, make a change and a PR for 30 repositories, and then follow up with the pipeline to make sure it's green. I built an MCP server that indexes all the repos, lets you query repositories, makes batch PRs, and gives you a summary of the workflow runs.

## Here's what it does:

1. Indexing, which happens in 2 forms: - Codebase level: runs an agent CLI (with proper context) over all repos to extract what each one does, how they relate, and what the system looks like as a whole.

- Repo level: Having the codebase context, it extracts logical info of each repo, and also the libraries, dependencies, etc for lexical search

2. Search, also in 2 forms:

- Natural language: where it answers search queries with respect to the codebase and targeted repository context - Structured search: where it returns the result based on actual dependencies (eg "find me repositories that are written with Python, have requirements.txt, and are using FastAPI)

3. Batch change:

Simply prompt "find my Python repositories and update library X from vY to vZ"; This will search and find the affected repos, clone them, run a CLI agent like CC on each with the context we already persisted, create and prepare PRs, and give you a report of the results.

## Tech stack `mongodb` To store the repository tree, dependencies, and workflows `redis` To store the user's session to track the ongoing batch job `claude-cli/Devin` Used as the main engine `docker-compose` to build `traefik` for routing

I would appreciate your feedback and thoughts on this

Demo video: https://infraas.ai/

PS: I reviewed all the code, so if it looks like slop, that's me ^^

Comments URL: https://news.ycombinator.com/item?id=48653659

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48619560

Points: 8

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48570549

Points: 5

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=48569562

Points: 5

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48500524

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48124411

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48111899

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48110865

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48099175

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=47952343

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=47936341

Points: 1

# Comments: 0

Before we launch, I'm trying to understand how teams actually solve this today.Also, a few questions for anyone dealing with this:

1. How many microservices does your team run in production?

2. Do you have true cross-service E2E tests, or mostly mocked integration tests?

3. If a tool could generate these tests from your Swagger + a plain-language description of your business flow, would that be useful enough to try?

Happy to share more about what we're building with anyone interested.

Comments URL: https://news.ycombinator.com/item?id=47774288

Points: 1

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=47663575

Points: 6

# Comments: 0