remove llama index plugin and update examples

[longcw]

No description provided.

[github-actions]

❌ Invalid Changeset Format Detected

One or more changeset files in this PR have an invalid format. Please ensure they adhere to:

• Start with --- and include a closing --- on its own line.
• Each package line must be in the format:
  "package-name": patch|minor|major
• No duplicate package entries allowed.
• A non-empty change description must follow the front matter.

Error details:
.github/next-release/changeset-0ac9b53f.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-27d8268e.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-65b3292c.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-7116e2a3.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-73e5be2c.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-79b2e16a.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-7aed4d48.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-a4f99981.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-aee68111.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-bdfd50f8.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-be147682.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-c6cc5df2.md: Failed to read file from git branch 'pr_head'.
.github/next-release/changeset-f0294eb1.md: Failed to read file from git branch 'pr_head'.

[theomonnom]

I'm wondering if we should delete livekit-plugins-llama-index, but convert our example to directly use their API in our examples

[longcw]

I'm wondering if we should delete livekit-plugins-llama-index, but convert our example to directly use their API in our examples

I have no preference. Only the chat_engine example is using the llama index plugin, the other two are already using the llama index directly. But for chat_engine, it seems okay to have the llama index as a LLM plugin, otherwise how to specify the llm for the agent?

[theomonnom]

I have no preference. Only the chat_engine example is using the llama index plugin, the other two are already using the llama index directly. But for chat_engine, it seems okay to have the llama index as a LLM plugin, otherwise how to specify the llm for the agent?

Yes, but the tradeoffs are too big, with llama-index, if you use the ChatEngine, we don't support function calls. So IMO, we should maybe just do an example using llama-index inside the llm_node

[longcw]

Yes, but the tradeoffs are too big, with llama-index, if you use the ChatEngine, we don't support function calls. So IMO, we should maybe just do an example using llama-index inside the llm_node

Sounds good to me.

[bcherry]

@theomonnom send a ping if you do end up deleting the plugin (or any other niche/one-of-a-kind plugins like rag, nltk, etc) since we have links to them in docs (very happy to remove, they're kinda awkward to list out since they don't fit into a category

[longcw]

@theomonnom updated all examples to use original llama-index, feel free to remove the pulgin.

[longcw]

should we add a mock LLM in the api like llm=llm.NOT_REQUIRE to enable the pipeline reply task for ppl want to use a fully customized llm_node, for example for multiple LLMs inside the node, or other chat completion libraries like llama index chat engine or langchain.

[biouser-abiomix]

I have no preference. Only the chat_engine example is using the llama index plugin, the other two are already using the llama index directly. But for chat_engine, it seems okay to have the llama index as a LLM plugin, otherwise how to specify the llm for the agent?

Yes, but the tradeoffs are too big, with llama-index, if you use the ChatEngine, we don't support function calls. So IMO, we should maybe just do an example using llama-index inside the llm_node

Hi all,
if you may allow me to contribute to this discussion, i wanted to add here that for RAG applications, the llama-index chat engine produces much better results compared to just providing a manual RAG tool. This is important for (multi)-agent workflows where a rag component is important. This component is the main reason i am sticking to the 0.x branch now.
Best wishes

[longcw]

@biouser-abiomix does this example work for you, without a llamaindex plugin you can still use it in the llm_node of the agent.

[biouser-abiomix]

@biouser-abiomix does this example work for you, without a llamaindex plugin you can still use it in the llm_node of the agent.

Can confirm it works @longcw
Thank you very much !

[theomonnom]

this lg! feel free to remove the llama-index plugin directly inside this PR