#fb5de6494c8b registry.dagger.io/engine
1 messages ยท Page 1 of 1 (latest)
Just ran the logs through chatgpt as well lol https://chatgpt.com/share/6860a0e4-cde0-800d-ba97-51e62c5de07a
have an idea about what could be happening
mind sharing your gemini settings.json file?
{
"theme": "Ayu Light",
"selectedAuthType": "oauth-personal",
"preferredEditor": "vscode",
"coreTools": [
"LSTool",
"ReadFileTool"
],
"mcpServers": {
"browsermcp": {
"command": "npx",
"args": ["@browsermcp/mcp@latest"]
},
"playwright": {
"command": "npx",
"args": [
"@playwright/mcp@latest"
]
},
"container-use": {
"command": "cu",
"args": ["stdio"],
"timeout": 60000,
"trust": true
},
"idea-to-production": {
"command": "node", "args": ["/mcp/mcp-idea-to-production/dist/index.js"] }
}
}
ok, can you try chaning your contrainer-use command to use the full path of your cu CLI?
just want to make sure gemini is calling the right cu binary. Since MacOS also has another cu binary that sometimes conflicts
hmm.. there's no tool to list environments currently ๐ค
you can run cu list but the MCP doesn't expose a tool for that
i forgot which one it was because I closed that prior chat, but i'll rerun and show ya
tried it with the actual route too
"container-use": {
"command": "/opt/homebrew/bin/cu",
"args": ["stdio"],
"timeout": 60000,
"trust": true
},
I assume it's because my application is too large.
that'd be strange.. does it work in a new empty directory?
I shall try!
seems to be trying to open en env that doesn't exist? does the environment exist with cu list?
strange.. it'd have to be quite big to bork it ๐ฌ
yeah so it works on a much smaller repository
the test repo is 149 MB
my actual repo is like 9.5 GB. It seems like it is not git ignoring my large data files
oh wow.. well that's large ๐
yeah.. that might be the case then
I thought it would git ignore it ๐
but alas
Why I guess isn't too crazy because if I added a transformer library it would come out to like 3-5 GB immediately
Even removing my large data files doesn't allow it to work, so could be something else. Anyways thank you
@cobalt bay cu prints its logs to /tmp/cu*. Maybe you can spot something there