Collections
Create a document collection that OpenIndex.ai will index and calculate embeddings for.
GET /api/collections HTTP/1.1
Host: www.openindex.ai
Authorization: Bearer JWT
Accept: */*
A list of collections.
{
"ok": true,
"collections": [
{
"id": "text",
"slug": "text",
"userId": "text",
"active": true,
"createdAt": 1,
"updatedAt": 1,
"name": "text",
"description": "text",
"image": "text",
"privacy": "public",
"namespace": "text",
"sources": [],
"status": "text"
}
]
}
A list of URLs (one per row) to add to the collection. YouTube URLs will be transcripted. Use links(url) to index the direct links at url.
POST /api/collections HTTP/1.1
Host: www.openindex.ai
Authorization: Bearer JWT
Content-Type: application/json
Accept: */*
Content-Length: 105
{
"id": "text",
"name": "text",
"description": "text",
"image": "text",
"privacy": "public",
"knowledgeBase": "text"
}
{
"ok": true,
"collectionId": "text"
}
List publicly-accessible URLs, one per line, in the knowledgeBase
parameter. Any YouTube URL will be transcribed and index.
Use links(<url>)
to index all direct links in url
.
The query to search for.
The number of results to return.
You can restrict the search to a specific collection by passing its ID.
GET /api/collections/search HTTP/1.1
Host: www.openindex.ai
Authorization: Bearer JWT
Accept: */*
A list of collections.
{
"ok": true,
"results": [
{
"snippet": "text",
"url": "text",
"sourceId": "text",
"source": "web",
"collection": {
"id": "text",
"slug": "text",
"userId": "text",
"active": true,
"createdAt": 1,
"updatedAt": 1,
"name": "text",
"description": "text",
"image": "text",
"privacy": "public",
"namespace": "text",
"sources": [],
"status": "text"
},
"score": 1
}
]
}
The /collections/search
endpoint will run a hybrid similarity search of your query passed as parameter q
and return the chunks and collections with the highest semantic similarity after a re-ranking step.
This endpoint can be used as an action in your custom OpenAI GPTs. Search the full OpenIndex.ai knowledge base by adding a new function to your custom GPT with this OpenAPI json schema.
If you want to only search a specific collection in your custom GPT, use the json schema URL https://www.openindex.ai/actions/gpt?collection_id=<collection_id>
. We now show that URL in the main collection page, see for example the collection Vitalik's Essays.
For now only public OpenIndex.ai collections can be retrieved. We are looking to implement OAuth login with OpenAI GPT to also allow for private and team collections search.
Last updated