llama-index-packs-agents-llm-compiler 0.2.0

llama-index packs agents llm compiler

LLMCompiler Agent Pack

This LlamaPack implements the LLMCompiler agent paper.

A lot of code came from the source repo, we repurposed with LlamaIndex abstractions. All credits to the original authors for a great work!

A full notebook guide can be found here.

CLI Usage

You can download llamapacks directly using llamaindex-cli, which comes installed with the llama-index python package:

llamaindex-cli download-llamapack LLMCompilerAgentPack --download-dir ./llm_compiler_agent_pack

You can then inspect the files at ./llm_compiler_agent_pack and use them as a template for your own project!

Code Usage

You can download the pack to a directory. NOTE: You must specify skip_load=True - the pack contains multiple files, which makes it hard to load directly.

We will show you how to import the agent from these files!

from llama_index.core.llama_pack import download_llama_pack

# download and install dependencies
download_llama_pack("LLMCompilerAgentPack", "./llm_compiler_agent_pack")

From here, you can use the pack. You can import the relevant modules from the download folder (in the example below we assume it's a relative import or the directory has been added to your system path).

# setup pack arguments

from llama_index.core.agent import AgentRunner
from llm_compiler_agent_pack.step import LLMCompilerAgentWorker

agent_worker = LLMCompilerAgentWorker.from_tools(
    tools, llm=llm, verbose=True, callback_manager=callback_manager
)
agent = AgentRunner(agent_worker, callback_manager=callback_manager)

# start using the agent
response = agent.chat("What is (121 * 3) + 42?")

You can also use/initialize the pack directly.

from llm_compiler_agent_pack.base import LLMCompilerAgentPack

agent_pack = LLMCompilerAgentPack(tools, llm=llm)

The run() function is a light wrapper around agent.chat().

response = pack.run("Tell me about the population of Boston")

You can also directly get modules from the pack.

# use the agent
agent = pack.agent
response = agent.chat("task")