new: langchain plugin

[aMahanna]

Bringing back the LangChain plugin as an optional feature

# pip install nx-arangodb[llm]
import nx_arangodb as nxadb

G_adb = nxadb.Graph(name="Karate")

G_adb.chat("who is connected to node 1?")
G_adb.chat("who is connected to node 1?", verbose=True)
G_adb.chat("what is the shortest path from node 1 to node 34?")
G_adb.chat("update node 1 with property 'foo' set to 'bar'")

The default LLM used is GPT-4 via the ChatOpenAI class of the langchain_openai module, but users are free to pass their own model via the llm parameter to the chat() function.

[hkernbach]

What about those two todos?

[hkernbach]

Can you share more information on those two methods. What are they for?

[aMahanna]

ya these are placeholders for a potential feature later on, completely unrelated to chat():

• pull proposes the idea of fetching the "latest" state of an ArangoDB Graph, which would introduce the concept of cross-collaboration across the same ArangoDB Graph. I don't have an implementation in-mind yet, so it's purely hypothetical.

• push would be more towards supporting transactions in an nxadb Graph. So something like this:

import nx_arangodb as nxadb

G = nxadb.Graph(name="MyGraph", use_transactions=True)

# Only modifies the cache:
G.nodes[1]['foo'] = 'bar'
G.add_node(2, foo='buz')
G.add_edge(1, 2)

G.push() # Sends the last 3 operations to the DB

[hkernbach]

LGTM to in general. Before we merge back, we need to discuss the two introduced TODOs.