Description
I've been struggling to setup a project from scratch for over 10 days now so I've decided to open an issue here.
I've tried using all the possible options:
eliza-starterrepo: First I tried with this one. After many unsuccessful attempts, I managed to get this up and running. However, the repo is now archived so I've decided to find another option. There should at least be a notice somewhere in the README explaining why the repo was archived as it is a bit confusing to come back to the repo you used yesterday and find it archived.eliza(this) repo. It references an old video from a few months ago that doesn't help very much as a lot of stuff has changed in the meantime. I tried following the README:
git clone https://github.com/elizaos/eliza.git
git checkout $(git describe --tags --abbrev=0)
cp .env.example .env
Then I set the GOOGLE_GENERATIVE_AI_API_KEY variable in .env. This is just my best guess for the variable name using the eliza-starter repo as there's no word on how to set up an agent using Gemini in the .env itself or in the docs.
Then I execute the rest of the commands:
bun install
bun run build
bun start
Here the agent starts and I navigate to http://localhost:3000 to try and text my agent. However, the error is A context size of 8192 is too large for the available VRAM as the model is started locally and my Gemini API key is ignored.
- CLI tool: I create a new project by running
elizaos create my-agentand choose the standard options, including the OpenAI LLM. Long story short, this one fails me as well because I can't in God's name figure out how to start the model using my desired provider or change it afterwards.
TL;DR There's a lack of synchronisation between the official documentation, this repo, and the CLI tool. It is very difficult to figure out where to look up the issues you have and find an answer to them.
This is just a fraction of the issues I've had along the way, all while just trying to set up a simple project.