Klu Context

Context is a library of Documents. Klu automatically creates a vector index and performs retrieval augmented generation via similarity search. Klu appends the most relevant documents to the Action prompt at runtime.

Context and Document model

Create Context libraries in Klu Studio and via the API or SDK. Documents are automatically created via Klu Studio (crawler, file upload, or integrations) or programmatically via the API or SDK (app data or file upload).

Documents can be added to a Context library at any time. Klu automatically transcribes audio and video content. Klu supports the following file types:

Additionally, Klu supports non-file Context

All Context libraries and Documents are automatically associated with the connected Workspace via your API Key.

Context Examples

Here are common examples for how to use Klu Context in your applications.

Create Context from App Data

Create Context from app data in two steps. First, create a custom type Context. Then, add Documents to the Context library. Add specific metadata to each Document to improve retrieval augmented generation results.

from klu import Klu

klu = Klu("YOUR_TOKEN")

contect = await klu.context.create(
    name="A beautiful library",
    description="A library of secret documents",
    type="api"
)

print(context)

from klu import Klu

klu = Klu("YOUR_TOKEN")

docs = klu.context.add_doc(
    context=context.guid,
    content="John McAfee is a great guy. He works at Acme Inc. and lives in New York. He loves tacos.",
    meta_data={
      "organization_id": "3a38-223-8191",
      "organization_name": "Acme Inc.",
      "user_id": "1c877-069-4201",
      "user_name": "[email protected]"
    }
)

Create Context from Files

When creating Context from files, you will need to pre-sign the file URL and then upload the file to the pre-signed URL. This is a two-step process for each file. Once finished, simply add the file names to the Context creation request.

The Klu SDKs greatly simplify this process.

from klu import Klu

klu = Klu("YOUR_TOKEN")

presigned_url = await klu.context.presign(
    file_name="taco.csv"
)

print(presigned_url)

from klu import Klu
from klu.context.models import FileData

klu = Klu("YOUR_TOKEN")

file_upload = klu.context.upload_file(
  FileData(file_name="taco.csv, file_path="~/taco.csv")
)
print(file_upload.object_url)

from klu import Klu

klu = Klu("YOUR_TOKEN")

context = klu.context.create(
    name="A beautiful library",
    description="A library of secret documents",
    type="files",
    files=["taco.csv", "master_plan_final1.pdf", "unlimited_power.txt"]
)

print(context)

Create Context

from klu import Klu

klu = Klu("YOUR_TOKEN")

context = klu.context.create(
    name="A beautiful library",
    description="A library of secret documents",
    type="files",
    files=["taco.csv", "master_plan_final1.pdf", "unlimited_power.txt"]
)

print(context)

{
  "name":"A beautiful library",
  "description":"A library of secret documents",
  "typeId":4,
  "loaderId":12,
  "workspaceId":1,
  "createdAt":"2023-07-30T03:47:22.601000",
  "updatedAt":"2023-07-30T03:47:23.002000",
  "createdById":"zzz0a0bb694207z22zll69zphp",
  "processed":true,
  "meta_data":null,
  "guid":"11b4e91c-80be-4edc-96fd-a482c7f263bb"
}

Add App Data

from klu import Klu

klu = Klu("YOUR_TOKEN")

doc = klu.context.add_doc(
    context="context_guid",
    text="John McAfee is a great guy. He works at Acme Inc. and lives in New York. He loves tacos.",
    meta_data={
      "organization_id": "3a38-223-8191",
      "organization_name": "Acme Inc.",
      "user_id": "1c877-069-4201",
      "user_name": "[email protected]"
    }
)

{
  "guid": "123e4567-e89b-12d3-a456-426614174000",
  "status": "success"
}

Add Files

from klu import Klu

klu = Klu("YOUR_TOKEN")

context_updated = klu.context.add_files(
    context="context_guid",
    files=["taco.csv", "master_plan_final1.pdf", "unlimited_power.txt"]
)

print(context_updated)

{
  "name": "A beautiful library",
  "description": "A library of secret documents",
  "createdAt": "2023-01-01T12:00:00Z",
  "updatedAt": "2023-01-01T12:01:00Z",
  "processed": true,
  "meta_data": {
    "organization_id": "3a38-223-8191",
    "organization_name": "Acme Inc.",
    "user_id": "1c877-069-4201",
    "user_name": "[email protected]"
  },
  "guid": "123e4567-e89b-12d3-a456-426614174000"
}