#Getting Started with melvin
dry quiver
Tried this out with 15.3 and ran into an issue, made it further in 15.4 but looks like were missing the platform argument?
./dagger-llm | cli | export $HOME/bin/
Should be
./dagger-llm | cli darwin | export $HOME/bin/
darwin on MacOS and likely something else for others
This builds the llm-enabled version of the Dagger CLI, and installs it at `~/bin/dagger-llm`.
This is also not working for me, if I run the command as is then it installed the engine at $HOME/bin (i.e bin is the binary)
I am also confused on why we built the binary in step 2 because later on we are making calls with just plain dagger does this assume that dagger is the binary we built in step 2 instead of the latest stable version?
--
Happy to send a PR but wanted to rule out maybe something odd on my own machine.
bright axle
I think step 5 and onward should be the binary you built and not installed dagger
snow tide
yeah that's 3 different mistakes I made in the README ... sorry I'm seriously sleep deprived and not thinking straight
should be cli current (can't automatically select the client's platform but really really would like to...)
snow tide
@dry quiver @bright axle I pushed fixes for all the above, thank you for the report.
dry quiver
Hm, I am still running into issues with the updated quickstart.
setup.sh is a huge help and seems to be working fine, but trying to run the llm command results in an abrupt exit with no additional info. Verbose logging does not seem to help.
@bright axle or @compact thicket were you able to get this working?
bright axle
can you try llm | with-prompt "what is your purpose?" | loop | history?
dry quiver
Same issue, but feels like a me thing for some reason? :/
https://v3.dagger.cloud/levlaz/traces/475f20c457d7017d836418f9c256d2f8
bright axle
i'll read through the setup instructions again, one min
dry quiver
I was specifically testing this part out
To get started quickly:
1. Run `./setup.sh` in the root of the repo. Leave it running, and open a new terminal.
2. `export EXPERIMENTAL_DAGGER_RUNNER_HOST=tcp://localhost:1234`
3. `~/bin/dagger-llm shell -c 'llm | with-prompt "llm, are you there?" | last-reply'`
4. If you got a response from the LLM, congratulations! Setup is complete.
5. Continue to [running Melvin in the command-line](#run-melvin-from-the-command-line)
do you have a .env with stuff in it?
dry quiver
No, but OPENAI_API_KEY is set in the env
bright axle
ok. It has to be in the .env I believe. The setup.sh should create one but maybe it failed for some reason
dry quiver
Ill try that!
snow tide
@bright axle we support regular env variables too now
dry quiver
Ok so I am trying to do this step by step and it seems the issue is right in the front
-
open shell
~/bin/dagger-llm shell -
load llm module
✘ llm 0.0s
! function or module "llm" not found
│ ✔ looking for module 0.0s
ah. Did you get step 2?
dry quiver
is it because EXPERIMENTAL_DAGGER_RUNNER_HOST=tcp does not have a leading undescore
dry quiver
Oops 😂
bright axle
nice catch
dry quiver
Ok my stuff is finally working, huge thanks to @bright axle !