Clojurians Log v2

the holy grail of CI for me is builds I can REPL-in, inspect and debug and continue. I’ve done this on the usual CI software by intentionally not returning after failure, but I wonder if there is first class support for this in MonkeyCI?

Not yet. MonkeyCI supports unit tests on your build scripts already. But there are plans to add repl functionality in the future. I haven't really thought this through, but I'm thinking it should be possible to start a repl using the cli, or allow external repls to connect to a running build.

Seems mainly something for local builds though, not sure if this should be opened up when running builds via the web app (for security reasons). I'd have to investigate if this can be secured with a key somehow.

its very useful for deploy-steps as often the environment can’t be replicated perfectly locally

Indeed. I've created https://github.com/monkey-projects/monkeyci/issues/351 for further investigation.

I plan to record a video showing how to setup it properly, because there are some deps, like chroma which I use to store memories, instead of text-in-files etc - so it's light and LLMs can do semantic search - but, in general, tools yield safe degradation, and will use some fallback (like using org-kanban, instead of in-memory kanban, etc) if dependency is not provided. Everything infra that is needed is in docker-compose.yaml Then, I will proceed to give a demo and put in youtube. Current the hive-mcp comes with ~170 tools, it has builtin in-memory kanban for autonomous project-managements; tools that make possible bidirectional-communication among agents; multi-agent spawning; and much more. For using the swarm feature (multi-agent), you should setup bb-mcp which is the work-around I created to spawn lightweight JVM that attach to a full-blown JVM for lings (agents spawned by the "hivemind" e.g., the claude that is talking to you and coordinating the project with you). I have spawned up to 15 lings at once, with plenty of headroom. I will comment here to keep you posted 🙏😁.

Sounds really interesting! Tried it and start doesnt work:

Execution error (FileNotFoundException) at hive-mcp.server/eval228$loading (server.clj:1).
Could not locate io/modelcontext/clojure_sdk/stdio_server__init.class, io/modelcontext/clojure_sdk/stdio_server.clj or io/modelcontext/clojure_sdk/stdio_server.cljc on classpath. Please check that namespaces with dashes use underscores in the Clojure file name.

Full report at:
/tmp/clojure-3959416240507406413.edn

https://www.youtube.com/watch?v=eE8t9IWD94I

Install walk-through ✅

Hello @services799! It is. I'm really excited to share it open sourced. So, for starters, probably there is two things: Currently, the project uses forks of some projects with some changes on them (to enable things like the piggyback behavior - bidirectional communication - among the hive). I opened PRs, but some haven't been accepted yet etc. So, you will have to clone it recursively. That's probably why you are seeing that error. The second thing is that the project has an Emacs connection part. So, you will have to load the right config locally - although I'm working to put it in MELPA at some point. I'm currently updating the docs. It's a little lengthy process to set it up completely. So the hive is writing the multi-step install process in a wiki. As soon as it's in the remote, I will let you know, ok?

# 1. Install CLI and MCP server
go install 
go install 

# 2. Register MCP server with Claude
claude mcp add hive-setup --scope user -- hive-setup-mcp

Please check out the updates in https://github.com/hive-agi/hive-mcp Since it's more friendly now to install too. Cleaned up deps to use fixed versions (no need to clone recursively, as mentioned previously) and make it more reproducible setup by using env vars throughout the install process (and created a wiki with a lot of docs) You can feed it too claude, as well, and she will probably be able to figure everything up. docker-compose pulls most of infra needed to run locally too. I would go with a mix of cloning the repo and using hive-setup-mcp . Will be better guide than myself haha 🙂 Thanks for trying out hive-mcp . I hope you enjoy it, as much as I do