FileGPT API Documentation

Still in Alpha version

About this

This is the official documentation of FileGPT. We will walk you through basics of using FileGPT API.

As a quick intro, FileGPT (https://filegpt.app) allows you to chat with all kinds of documents, including but not limited to, regular files, audios and videos, and youtube!

We deliver a consistent experience and integration for a range of documents, types, and embeddings, trying to bring the service to

Getting Started

1. First, log in on https://filegpt.app and go to your profile page. API access is available to our pro users and above.

2. Once you have access, you will be able to view your api usage on the profile page. You also have the options to generate api keys and invalidate old ones.

3. For all requests coming in at FileGPT, authentication is needed by setting Authorization: <YOUR_API_KEY>, for example Authorization: abcdefgh

4. Thats it! Easy peasy!

API Endpoints

The base endpoint is https://lb.filegpt.app

Below is a sample workflow of using our API. We first present the existing endpoints and then show some working end-to-end examples to get you started easily.

First, a collection is a set of documents. A document is the basic building block, and note that documents have different types. Generally, we categorize documents into three for now:

1. FILE: this is the most common one, it covers pdf, txt, pptx, jpg, etc.

2. YOUTUBE: this handles the youtube video links

3. WEB_URL: this handles web pages

POST /api/collection

create a collection with a list of documents. You should pass in a ApiCollectionCreateRequest object.

create a collection with a list of documents.

POST ENDPOINT/api/collection

Request Body

Name	Type	Description
title*	String
documents*	List	List of documents, see schema for detail reference

200: OK A collection object. will be returned

GET /api/collection/{collection_id}

Get the collection

GET ENDPONIT/api/collection/{collection_id}

This will get the collection and all documents in the collection

Path Parameters

Name	Type	Description
collection_id*	String

200: OK A collection object

Return value

A collection object.

DELETE /api/collection/{collection_id}

This will delete all files in your collection for privacy and security. Recommended to use as often as possible.

Get the collection

DELETE ENDPOINT/api/collection/{collection_id}

This will delete the collection and all documents in the collection

Path Parameters

Name	Type	Description
collection_id*	String

200: OK A collection object

Parameter

your collection id

Return Value

A collection object.

POST /api/build

Build and process the collection

POST http://ENDPOINT/api/api/build

Request Body

Name	Type	Description
collectionId*	String

200: OK A collection object

Parameter

A BuildRequest object.

Return Value

A success status of 200. Note that the build job is async. You should use the GET method to poll the status of the collection, and the status will change to READY if everything goes well, or it will be FAILED if error occured. The error message is in the StatusMessage field.

POST /api/query

Parameter

Send a JSON body of a QueryRequest Object.

Return Value

A plain text streaming response.

Non-streaming version will be out soon. We will provide examples on how to consume the stream if you need.

End-to-End Example

Example Building a Custom Chatbot with QA

In this example, we walk through how you can use filegpt as an add-on service, on top of your existing chatbot, to let users chat with their own documents.

Here's an outline of the workflow:

1. You manage your storage of users' uploaded file.

2. Generate a presigned URL with a 1-2 minute expiration window

3. Create a collection with the documents, you can create a collection with one or multiple documents.

4. Build the collection

5. Query the collection with the user input.

# s3, gcs, etc. Your file storage service.
file_url = await generate_presigned_url()
# create collection
collection = await requests.post(
    'ENDPOINT/api/collection', 
    headers={'Authorization':'YOUR_API_KEY'},
    data={
        # Here is the collection information    
        "title": "demo",
        "documents": [{
            "name": "abc",
            "document_type": "FILE",
            "url": "https://www.africau.edu/images/default/sample.pdf"
        }]
    })
# build collection
await requests.post(
    'ENDPOINT/api/collection',
    headers={'Authorization':'YOUR_API_KEY'},
    data={
        'collectionId': collection.id
    }
)
# query collection with Stream
query_data = {
    "collectionId": "457cd51f-662c-4ab5-9fb7-949ea031de2f",
    "query": "summarize the doc",
    "mode": "default",
    "messageHistory":[
        {
            "content": "my name is alex",
            "author": "bot"

        }
    ]
}
response = requests.post('ENDPOINT/api/query', data=query_data, stream=True)
result = ""
# Ensure the request was successful
if response.status_code == 200:
    # Consume the stream
    for chunk in response.iter_content(chunk_size=1024): 
        if chunk: # filter out keep-alive new chunks
            print(chunk)
            result += chunk
else:
    print("Error:", response.status_code, response.text)

Appendix

ApiCollectionCreateRequest

This is the object you need to create a a collection. You specify a list of documents of your own, with URL provided, and then you can create a collection. It holds all documents together.

class ApiCollectionCreateRequest():
    title: string
    documents: Array[ApiDocument]
class ApiDocument():
    """
    name: name of the document
    document_type: resource type
    url: link to the resource
    """
    name: string
    document_type: 'FILE' | 'YOUTUBE_URL' | 'WEB_URL'
    url: string
### Example
{
    "title": "demo",
    "documents": [{
        "name": "abc",
        "document_type": "FILE",
        "url": "https://www.africau.edu/images/default/sample.pdf"
    }]
}

BuildRequest

class BuildRequest():
    collectionId: str
### Example
{
    "collectionId": "abcd"
}

QueryRequest

class QueryRequest():
    query: str
    collectionId: str
    chatModel: Optional[str] = "gpt-3.5-turbo" | "gpt-4"
    messageHistory: Optional[List[ApiChatMessage]] = []
class ApiChatMessageAuthor(Enum):
    BOT = "bot"
    USER = "user"
class ApiChatMessage():
    content: str
    author: ApiChatMessageAuthor
### Example
{
    "query": "what is my name?",
    "collectionId": "abcd",
    "messageHistory": [
        {
            "author": "user",
            "content": "my name is alex."
        }
    ]
}

Collection Schema

class Collection():
    """Represents a Collection record"""
    id: _str
    title: _str
    status: 'NOT_STRATED' | 'BUILDING' | 'READY' | 'FAILED'
    createdAt: datetime.datetime
    updatedAt: datetime.datetime
    statusMessage: Optional[str]
    documents: Optional[List[Document]]
### Example
{
    "id": "abcdefghijklmn",
    "title": "demo",
    "status": "READY",
    "createdAt": "2023-05-13T01:39:01.106000+00:00",
    "updatedAt": "2023-05-13T01:39:40.452000+00:00",
    "userId": "abcdefghijklmn",
    "statusMessage": "collection is ready",
    "documents": [
        {
            "id": "abcdefghijklmn",
            "name": "abc",
            "URL": "https://www.africau.edu/images/default/sample.pdf",
            "createdAt": "2023-05-13T01:39:01.106000+00:00",
            "updatedAt": "2023-05-13T01:39:01.106000+00:00",
            "type": "FILE",
            "collections": null
        }
    ]
}