Hands-on langchain

Requirements:

python = ">=3.10,<4.0"
langchain-core = "^1.0.0" 
langchain = "^1.0.0"
langchain-openai = "^1.1.0" 
langserve = "^0.3.0"
langsmith = "^0.4.0" 
langchain-community = "^0.4.1"

my-app

A simple prompt|model response chain

• langchain template new my-app
• needs OpenAI api key (get your own)
• ensure poetry dependencies are resolved

my-language

Setup a REST API with langchain serve. It serves a RAG based assistant, which roleplays any character from any book/play, with a personality cue, and has an context of George Orwell. It can say pessimistic stuff, if you ask it to!

• langchain app new my-language --package $PWD/my-assistant
• add_routes(app, my_app_chain, path="/assistant")
• ensure poetry dependencied are resolved
• cd my-language; langchain serve
• http://127.0.0.1:8000/docs to see all the API endpoints
• http://127.0.0.1:8000/assistant/playground to get a default chat interface
• interacting with the playground chat box UI, it will invoke the stream endpoint
• if curl is used then it uses the invoke endpoint

curl --location --request POST 'http://127.0.0.1:8000/assistant/invoke' \
--header 'Content-Type: application/json' \
--data-raw '{
   "input": {
       "question": "Hello, Anthony does the world feel like it is against you?",
       "character": "Mark Anthony",
       "title": "Julius Caesar",
       "personality": "sullen mood, rhetorical, sarcastic, and dramatic"
   }
}'

Reponse:

Oh, how astute of you to notice! Indeed, it feels as though the very world conspires against me, drowning me in a sea of treachery and deceit. The nightmare of statistics haunts my every step, twisting reality into a cruel game of numbers. But fear not, for I shall rise above this adversity with the fire of vengeance burning in my heart!

[Mood: Defiant]

Develop with a virtual environment

• Switch into top-level directory
• python3 -m venv .venv (create a venv)
• add export PIP_INDEX_URL=https://pypi.org/simple/ to the end of .venv/bin/activate
• add export PIP_EXTRA_INDEX_URL="" to the end of .venv/bin/activate
• activate venv: source .venv/bin/activate
• Then switch to my-app directory and do poetry install
• Similarly switch to my-language directory and do a poetry install

Packages to have in venv

• homebrew install envdir
• You can use envdir to load secrects like API_KEYS in .envrc
• add .venv, .envrc in .gitignore to ensure not uploading your open AI keys to public
• pip install requests
• you will need poetry

Installation & troubleshooting

• Create your virtual env inside the main cloned directory activate it
• update your langchan/langsmith API_KEYS in .envrc (with apt gitignore)
  • you need to set direnv allow when the system prompts you.
• navigate to my-language/packages/my-assistant
  • Do a poetry lock, poetry install
• Navigate back out to ../.. to my-language
  • do another poetry lock, and then poetry install
• From /my-language launch with langchain serve
  • If you might face random errors at langchain serve start
  • pip install --upgrade typer click, apparently this is because of some issue with typer.

Pretty straight forward python dependency debugging