feat: Add memgpt quickstart command

[cpacker]

Please describe the purpose of this pull request

Add memgpt quickstart to allow users to immediately start using the hosted endpoint (or openai/azure) w/o copy-paste.

I've set things up so that the quickstart configs (.json) files are actually pulled from the internet (the repo main), that way we can update eg the memgpt hosted endpoint details on-the-fly with commits to main.

How to test

• memgpt quickstart
  • Simple, should just pull from remote url
• memgpt quickstart --type openai
  • To make this a true "one-click", we should check for the existence of OPENAI_API_KEY and throw a nice error here
• memgpt quickstart --type azure
  • Similarly, we should check for the existence of all the necessary Azure variables and throw a nice error

Have you tested this PR?

Yes, see examples below.

• memgpt quickstart
• memgpt quickstart --type openai
  - [ ] memgpt quickstart azure

Note: Azure is a little complicated, should put in separate PR.

Outstanding questions

• Should we pull configs from GitHub? Can also put the configs inside the gh-pages assets directory, so the wget goes to memgpt.ai/... instead of a GitHub link
• Should we store configs in memgpt/configs? Naming? (memgpt/example_configs?)
• Should we store JSON? YAML?
• Change command convention? eg memgpt quickstart --type openai could just be memgpt openai

---

Examples:

% memgpt quickstart --debug
JSON config file downloaded successfully.
Replacing config model: gpt-4 -> ehartford/dolphin-2.5-mixtral-8x7b
Replacing config model_endpoint_type: openai -> vllm
Replacing config model_endpoint: https://api.openai.com/v1 -> http://api.memgpt.ai
Replacing config model_wrapper: None -> airoboros-l2-70b-2.1
Replacing config context_window: 8192 -> 32768
Replacing config embedding_endpoint_type: openai -> hugging-face
Replacing config embedding_endpoint: https://api.openai.com/v1 -> http://embeddings.memgpt.ai
Replacing config embedding_model: None -> BAAI/bge-large-en-v1.5
Skipping new config embedding_dim: 1536 == 1536
Skipping new config embedding_chunk_size: 300 == 300
Saving new config file.

% memgpt quickstart --debug --type openai
? Enter your OpenAI API key (starts with 'sk..'): dummy key
JSON config file downloaded successfully.
Replacing config model: ehartford/dolphin-2.5-mixtral-8x7b -> gpt-4
Replacing config model_endpoint_type: vllm -> openai
Replacing config model_endpoint: http://api.memgpt.ai -> https://api.openai.com/v1
Replacing config model_wrapper: airoboros-l2-70b-2.1 -> None
Replacing config context_window: 32768 -> 8192
Replacing config openai_key: sk-... -> dummy key
Replacing config embedding_endpoint_type: hugging-face -> openai
Replacing config embedding_endpoint: http://embeddings.memgpt.ai -> https://api.openai.com/v1
Replacing config embedding_model: BAAI/bge-large-en-v1.5 -> None
Skipping new config embedding_dim: 1536 == 1536
Skipping new config embedding_chunk_size: 300 == 300
Saving new config file.

[sarahwooders]

What do you think of doing memgpt run --config openai ?

[sarahwooders]

lgtm!