Cohere

The Cohere platform builds natural language processing and generation into your product with a few lines of code. Cohere’s large language models (LLMs) can solve a broad spectrum of natural language use cases, including classification, semantic search, paraphrasing, summarization, and content generation. Use the Cohere Embed API endpoint to generate language embeddings, and then index those embeddings in Pinecone for fast and scalable vector search.

Setup guide

In this guide, you’ll learn how to use the Cohere Embed API endpoint to generate language embeddings, and then index those embeddings in Pinecone Database for fast and scalable vector search. This is a common combination for building semantic search, question-answering, threat-detection, and other applications that rely on NLP and search over a large corpus of text data. The basic workflow looks like this:

Embed and index
- Use the Cohere Embed API endpoint to generate vector embeddings of your documents (or any text data).
- Upload those vector embeddings into Pinecone, which can store and index millions or billions of these vector embeddings and search through them at low latency.
Search
- Pass your query text or document through the Cohere Embed API endpoint again.
- Take the resulting vector embedding and send it as a query to Pinecone.
- Get back semantically similar documents, even if they don’t share any keywords with the query.

Set up the environment

Start by installing the Cohere and Pinecone clients, along with Hugging Face Datasets for downloading the TREC dataset used in this guide:

Shell

pip install -U cohere pinecone datasets

Create embeddings

Python

import cohere

co = cohere.ClientV2(api_key="<YOUR_COHERE_API_KEY>")

Load the Text REtrieval Conference (TREC) question classification dataset, which contains 5.5K labeled questions. You’ll take only the first 1K samples for this walkthrough, but you can scale this to millions or even billions of samples.

Python

from datasets import load_dataset

# load the first 1K rows of the TREC dataset
trec = load_dataset('CogComp/trec', revision='refs/convert/parquet', split='train[:1000]')

Each sample in trec contains two label features and the text feature. Pass the questions from the text feature to Cohere to create embeddings. The Embed API accepts up to 96 texts per request, so send them in batches.

Python

embeds = []
batch_size = 96

for i in range(0, len(trec['text']), batch_size):
    res = co.embed(
        texts=trec['text'][i:i+batch_size],
        model='embed-english-v3.0',
        input_type='search_document',
        embedding_types=['float'],
        truncate='END'
    )
    embeds.extend(res.embeddings.float_)

Check the dimensionality of the returned vectors. Save the embedding dimensionality, because you need it when you create your Pinecone index later.

Python

import numpy as np

shape = np.array(embeds).shape
print(shape)

Response

(1000, 1024)

You can see the 1024 embedding dimensionality produced by Cohere’s embed-english-v3.0 model, and the 1000 samples you built embeddings for.

Store the embeddings

Now that you have your embeddings, you can move on to indexing them in Pinecone Database. For this, you need a Pinecone API key.First, initialize your connection to Pinecone, and then create a new index called cohere-pinecone-trec for storing the embeddings. When you create the index, specify the cosine similarity metric to align with Cohere’s embeddings, and pass the embedding dimensionality of 1024.

Python

from pinecone import Pinecone, ServerlessSpec

# initialize connection to pinecone (get API key at app.pinecone.io)
pc = Pinecone(api_key='<YOUR_PINECONE_API_KEY>')

index_name = 'cohere-pinecone-trec'

# if the index does not exist, we create it
if not pc.has_index(index_name):
    pc.create_index(
        name=index_name,
        dimension=shape[1],
        metric="cosine",
        spec=ServerlessSpec(
            cloud='aws', 
            region='us-east-1'
        ) 
    )

# connect to index
index = pc.Index(index_name)

Now you can begin populating the index with your embeddings. Pinecone expects you to provide a list of tuples in the format (id, vector, metadata), where the metadata field is an optional extra field where you can store anything you want in a dictionary format. For this example, you’ll store the original text of the embeddings.Upload the data in batches to avoid pushing too much data at once.

Python

batch_size = 128

ids = [str(i) for i in range(shape[0])]
# create list of metadata dictionaries
meta = [{'text': text} for text in trec['text']]

# create list of (id, vector, metadata) tuples to be upserted
to_upsert = list(zip(ids, embeds, meta))

for i in range(0, shape[0], batch_size):
    i_end = min(i+batch_size, shape[0])
    index.upsert(vectors=to_upsert[i:i_end])

# let's view the index statistics
print(index.describe_index_stats())

Response

DescribeIndexStatsResponse(dimension=1024, total_vector_count=1000, metric='cosine', namespaces=1)

You can see from index.describe_index_stats that you have a 1024-dimensional index populated with 1000 embeddings.

Run a semantic search

Now that you have your indexed vectors, you can perform a few search queries. To search, first embed your query with Cohere, and then search Pinecone with the returned vector.

Python

query = "What caused the 1929 Great Depression?"

# create the query embedding
xq = co.embed(
    texts=[query],
    model='embed-english-v3.0',
    input_type='search_query',
    embedding_types=['float'],
    truncate='END'
).embeddings.float_[0]

# query, returning the top 5 most similar results
res = index.query(vector=xq, top_k=5, include_metadata=True)

The response from Pinecone includes your original text in the metadata field. Print the top_k most similar questions and their similarity scores.

Python

for match in res['matches']:
    print(f"{match['score']:.2f}: {match['metadata']['text']}")

Response

62: Why did the world enter a global depression in 1929 ?
49: When was `` the Great Depression '' ?
38: What crop failure caused the Irish Famine ?
32: What caused Harry Houdini 's death ?
31: What causes pneumonia ?

The top results are relevant. To make the search harder, replace “depression” with the incorrect term “recession.”

Python

query = "What was the cause of the major recession in the early 20th century?"

# create the query embedding
xq = co.embed(
    texts=[query],
    model='embed-english-v3.0',
    input_type='search_query',
    embedding_types=['float'],
    truncate='END'
).embeddings.float_[0]

# query, returning the top 5 most similar results
res = index.query(vector=xq, top_k=5, include_metadata=True)

for match in res['matches']:
    print(f"{match['score']:.2f}: {match['metadata']['text']}")

Response

43: When was `` the Great Depression '' ?
40: Why did the world enter a global depression in 1929 ?
39: When did World War I start ?
35: What are some of the significant historical events of the 1990s ?
32: What crop failure caused the Irish Famine ?

Finally, search using the definition of depression rather than the word or related words.

Python

query = "Why was there a long-term economic downturn in the early 20th century?"

# create the query embedding
xq = co.embed(
    texts=[query],
    model='embed-english-v3.0',
    input_type='search_query',
    embedding_types=['float'],
    truncate='END'
).embeddings.float_[0]

# query, returning the top 10 most similar results
res = index.query(vector=xq, top_k=10, include_metadata=True)

for match in res['matches']:
    print(f"{match['score']:.2f}: {match['metadata']['text']}")

Response

40: When was `` the Great Depression '' ?
39: Why did the world enter a global depression in 1929 ?
35: When did World War I start ?
32: What are some of the significant historical events of the 1990s ?
31: What war did the Wanna-Go-Home Riots occur after ?
31: What do economists do ?
29: What historical event happened in Dogtown in 1899 ?
28: When did the Dow first reach ?
28: Who earns their money the hard way ?
28: What were popular songs and types of songs in the 1920s ?

This example shows that the semantic search pipeline can identify the meaning behind each of your queries. Using these embeddings with Pinecone lets you return the most semantically similar questions from the already indexed TREC dataset.

Clean up

When you’re finished with the index, delete it.

Python

pc.delete_index(name=index_name)

Connect an integration

Build an integration

Setup guide

Set up the environment

Create embeddings

Store the embeddings

Run a semantic search

Clean up

​Setup guide

Set up the environment

Create embeddings

Store the embeddings

Run a semantic search

Clean up

Setup guide