Cohere Citations Chat Engine Pack

Creates and runs a custom VectorStoreIndexWithCitationsChat -- which provides the chat engine with documents/citation mode. See the documentation here and here.

CLI Usage

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

llamaindex-cli download-llamapack CohereCitationChatEnginePack --download-dir ./cohere_citation_chat_pack

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

You can also directly install it if you don't want to look at/inspect the source code:

pip install llama-index-packs-cohere-citation-chat

Code Usage

You can download the pack to the ./cohere_citation_chat_pack directory:

from llama_index.readers.web import SimpleWebPageReader
from llama_index.core.llama_pack import download_llama_pack

# download and install dependencies
CohereCitationChatEnginePack = download_llama_pack(
    "CohereCitationChatEnginePack", "./cohere_citation_chat_pack"
)
documents = SimpleWebPageReader().load_data(
    [
        "https://raw.githubusercontent.com/jerryjliu/llama_index/adb054429f642cc7bbfcb66d4c232e072325eeab/examples/paul_graham_essay/data/paul_graham_essay.txt"
    ]
)
cohere_citation_chat_pack = CohereCitationChatEnginePack(
    documents=documents, cohere_api_key="your-api-key"
)
chat_engine = cohere_citation_chat_pack.run()
response = chat_engine.chat("What can you tell me about LLMs?")

# print chat response
print(response)

# print documents
print(response.documents)

# print citations
print(response.citations)

See the notebook on llama for a full example.