Pinecone Vector Database

Use Pinecone as a vector database for your Knowledge Base.

Setup

Follow the instructions in the Pinecone Setup Guide to get started quickly with Pinecone.

1uv pip install pinecone

Info

We do not yet support Pinecone v6.x.x. We are actively working to achieve compatibility. In the meantime, we recommend using Pinecone v5.4.2 for the best experience.

Example

1import os
2import typer
3from typing import Optional
4from rich.prompt import Prompt
5
6from kern.agent import Agent
7from kern.knowledge.knowledge import Knowledge
8from kern.vectordb.pineconedb import PineconeDb
9
10api_key = os.getenv("PINECONE_API_KEY")
11index_name = "thai-recipe-hybrid-search"
12
13vector_db = PineconeDb(
14    name=index_name,
15    dimension=1536,
16    metric="cosine",
17    spec={"serverless": {"cloud": "aws", "region": "us-east-1"}},
18    api_key=api_key,
19    use_hybrid_search=True,
20    hybrid_alpha=0.5,
21)
22
23knowledge_base = Knowledge(
24    vector_db=vector_db,
25)
26
27def pinecone_agent(user: str = "user"):
28    agent = Agent(
29        knowledge=knowledge_base,
30        debug_mode=True,
31    )
32
33    while True:
34        message = Prompt.ask(f"[bold] :sunglasses: {user} [/bold]")
35        if message in ("exit", "bye"):
36            break
37        agent.print_response(message)
38
39if __name__ == "__main__":
40    # Comment out after first run
41    knowledge_base.insert(
42        url="https://kern-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"
43    )
44
45    typer.run(pinecone_agent)

Async Support ⚡

Pinecone also supports asynchronous operations, enabling concurrency and leading to better performance.

1import asyncio
2from os import getenv
3
4from kern.agent import Agent
5from kern.knowledge.knowledge import Knowledge
6from kern.vectordb.pineconedb import PineconeDb
7
8api_key = getenv("PINECONE_API_KEY")
9index_name = "thai-recipe-index"
10
11vector_db = PineconeDb(
12    name=index_name,
13    dimension=1536,
14    metric="cosine",
15    spec={"serverless": {"cloud": "aws", "region": "us-east-1"}},
16    api_key=api_key,
17)
18
19knowledge_base = Knowledge(
20    vector_db=vector_db,
21)
22
23agent = Agent(
24    knowledge=knowledge_base,
25    # Enable the agent to search the knowledge base
26    search_knowledge=True,
27    # Enable the agent to read the chat history
28    read_chat_history=True,
29)
30
31if __name__ == "__main__":
32    # Load knowledge base asynchronously
33    asyncio.run(knowledge_base.ainsert(
34            url="https://kern-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"
35        )
36    )
37
38    # Create and use the agent asynchronously
39    asyncio.run(agent.aprint_response("How to make Tom Kha Gai", markdown=True))

Tip

Use aload() and aprint_response() methods with asyncio.run() for non-blocking operations in high-throughput applications.

PineconeDb Params

Parameter	Type	Default	Description
`name`	`str`	-	The name of the Pinecone index
`dimension`	`int`	-	The dimension of the embeddings
`spec`	`Union[Dict, ServerlessSpec, PodSpec]`	-	The index spec
`embedder`	`Optional[Embedder]`	`None`	Embedder instance for creating embeddings (defaults to OpenAIEmbedder if not provided)
`metric`	`Optional[str]`	`"cosine"`	The metric used for similarity search
`additional_headers`	`Optional[Dict[str, str]]`	`None`	Additional headers to pass to the Pinecone client
`pool_threads`	`Optional[int]`	`1`	The number of threads to use for the Pinecone client
`namespace`	`Optional[str]`	`None`	The namespace for the Pinecone index
`timeout`	`Optional[int]`	`None`	The timeout for Pinecone operations
`index_api`	`Optional[Any]`	`None`	The Index API object
`api_key`	`Optional[str]`	`None`	The Pinecone API key
`host`	`Optional[str]`	`None`	The Pinecone host
`config`	`Optional[Config]`	`None`	The Pinecone config
`use_hybrid_search`	`bool`	`False`	Whether to use hybrid search
`hybrid_alpha`	`float`	`0.5`	The alpha value for hybrid search