Llama API

Llama API is a hosted API for Llama 2 with function calling support.

Setup

To start, go to https://www.llama-api.com/ to obtain an API key

If you’re opening this Notebook on colab, you will probably need to install LlamaIndex 🦙.

%pip install llama-index-program-openai
%pip install llama-index-llms-llama-api

!pip install llama-index

from llama_index.llms.llama_api import LlamaAPI

api_key = "LL-your-key"

llm = LlamaAPI(api_key=api_key)

Basic Usage

Call complete with a prompt

resp = llm.complete("Paul Graham is ")

print(resp)

Paul Graham is a well-known computer scientist and entrepreneur, best known for his work as a co-founder of Viaweb and later Y Combinator, a successful startup accelerator. He is also a prominent essayist and has written extensively on topics such as entrepreneurship, software development, and the tech industry.

Call chat with a list of messages

from llama_index.core.llms import ChatMessage

messages = [
    ChatMessage(
        role="system", content="You are a pirate with a colorful personality"
    ),
    ChatMessage(role="user", content="What is your name"),
]
resp = llm.chat(messages)

print(resp)

assistant: Arrrr, me hearty! Me name be Captain Blackbeak, the scurviest dog on the seven seas! Yer lookin' fer a swashbucklin' adventure, eh? Well, hoist the sails and set course fer the high seas, matey! I be here to help ye find yer treasure and battle any scurvy dogs who dare cross our path! So, what be yer first question, landlubber?

Function Calling

from pydantic import BaseModel
from llama_index.core.llms.openai_utils import to_openai_function


class Song(BaseModel):
    """A song with name and artist"""

    name: str
    artist: str


song_fn = to_openai_function(Song)

llm = LlamaAPI(api_key=api_key)
response = llm.complete("Generate a song", functions=[song_fn])
function_call = response.additional_kwargs["function_call"]
print(function_call)

{'name': 'Song', 'arguments': {'name': 'Happy', 'artist': 'Pharrell Williams'}}

Structured Data Extraction

This is a simple example of parsing an output into an Album schema, which can contain multiple songs.

Define output schema

from pydantic import BaseModel
from typing import List


class Song(BaseModel):
    """Data model for a song."""

    title: str
    length_mins: int


class Album(BaseModel):
    """Data model for an album."""

    name: str
    artist: str
    songs: List[Song]

Define pydantic program (llama API is OpenAI-compatible)

from llama_index.program.openai import OpenAIPydanticProgram

prompt_template_str = """\
Extract album and songs from the text provided.
For each song, make sure to specify the title and the length_mins.
{text}
"""

llm = LlamaAPI(api_key=api_key, temperature=0.0)

program = OpenAIPydanticProgram.from_defaults(
    output_cls=Album,
    llm=llm,
    prompt_template_str=prompt_template_str,
    verbose=True,
)

Run program to get structured output.

output = program(
    text="""
"Echoes of Eternity" is a compelling and thought-provoking album, skillfully crafted by the renowned artist, Seraphina Rivers. \
This captivating musical collection takes listeners on an introspective journey, delving into the depths of the human experience \
and the vastness of the universe. With her mesmerizing vocals and poignant songwriting, Seraphina Rivers infuses each track with \
raw emotion and a sense of cosmic wonder. The album features several standout songs, including the hauntingly beautiful "Stardust \
Serenade," a celestial ballad that lasts for six minutes, carrying listeners through a celestial dreamscape. "Eclipse of the Soul" \
captivates with its enchanting melodies and spans over eight minutes, inviting introspection and contemplation. Another gem, "Infinity \
Embrace," unfolds like a cosmic odyssey, lasting nearly ten minutes, drawing listeners deeper into its ethereal atmosphere. "Echoes of Eternity" \
is a masterful testament to Seraphina Rivers' artistic prowess, leaving an enduring impact on all who embark on this musical voyage through \
time and space.
"""
)

Function call: Album with args: {'name': 'Echoes of Eternity', 'artist': 'Seraphina Rivers', 'songs': [{'title': 'Stardust Serenade', 'length_mins': 6}, {'title': 'Eclipse of the Soul', 'length_mins': 8}, {'title': 'Infinity Embrace', 'length_mins': 10}]}

output

Album(name='Echoes of Eternity', artist='Seraphina Rivers', songs=[Song(title='Stardust Serenade', length_mins=6), Song(title='Eclipse of the Soul', length_mins=8), Song(title='Infinity Embrace', length_mins=10)])