connectedpapers-py 0.1.4

The official python client for the connected papers API.

connectedpapers-py

The official python client for the connected papers API.

Installation

pip install connectedpapers-py

Usage

from connectedpapers import ConnectedPapersClient

DEEPFRUITS_PAPER_ID = "9397e7acd062245d37350f5c05faf56e9cfae0d6"

client = ConnectedPapersClient()
remaining_uses_count = client.get_remaining_usages_sync()
print(f"Remaining uses count: {remaining_uses_count}")
free_access_papers = client.get_free_access_papers_sync()
print(f"Free access papers: {free_access_papers}")
graph = client.get_graph_sync(DEEPFRUITS_PAPER_ID)
assert graph.graph_json.start_id == DEEPFRUITS_PAPER_ID

See more on the usage samples directory.

See graph structure at graph.py.

Configuring server address and API key

There are multiple ways to configure the server address and API key:

1. Set the environment variables CONNECTED_PAPERS_SERVER_ADDRESS and CONNECTED_PAPERS_API_KEY:

   export CONNECTED_PAPERS_REST_API="https://api.connectedpapers.com"
   export CONNECTED_PAPERS_API_KEY="YOUR_API_KEY"
   

   (You do not need to configure the server address, uses the correct one by default). Then you can use the client without parameters:

    from connectedpapers import ConnectedPapersClient
   
    client = ConnectedPapersClient()
   

2. Send parameters to the client's constructur:

   from connectedpapers import ConnectedPapersClient
   
   client = ConnectedPapersClient(access_token="YOUR_API_KEY", server_addr="https://api.connectedpapers.com")
   

Getting an access token

Contact us at [e-mail] to get an early-access access token.

API

Most functions offer a synchronous and an asynchronous version.

API structure

We have the following API calls available:

• Fetching a graph
• Getting the remaining usages count for your API key
• Getting the list of papers that are free to access

Free re-access to papers

If you have fetched a paper, it remains free to access for a month after the first access without counting towards your usages count.

Synchronous API

from connectedpapers import ConnectedPapersClient

client = ConnectedPapersClient()
client.get_graph_sync("YOUR_PAPER_ID")  # Fetch a graph for a single paper
client.get_remaining_usages_sync()  # Get the remaining usages count for your API key
client.get_free_access_papers_sync()  # Get the list of papers that are free to access

Asynchronous API

from connectedpapers import ConnectedPapersClient

client = ConnectedPapersClient()


async def usage_sample() -> None:
   await client.get_graph_async("YOUR_PAPER_ID")  # Fetch a graph for a single paper
   await client.get_remaining_usages_async()  # Get the remaining usages count for your API key
   await client.get_free_access_papers_async()  # Get the list of papers that are free to access

usage_sample()

Async iterator API

The client offers support for Python's asynchronous iterator access to the API, allowing for real-time monitoring of the progress of graph builds and retrieval of both current and rebuilt papers.

The method get_graph_async_iterator returns an asynchronous iterator, generating values of the GraphResponse type. Here's what the GraphResponse and related GraphResponseStatuses looks like in Python:

class GraphResponseStatuses(Enum):
    BAD_ID = "BAD_ID"
    ERROR = "ERROR"
    NOT_IN_DB = "NOT_IN_DB"
    OLD_GRAPH = "OLD_GRAPH"
    FRESH_GRAPH = "FRESH_GRAPH"
    IN_PROGRESS = "IN_PROGRESS"
    QUEUED = "QUEUED"
    BAD_TOKEN = "BAD_TOKEN"
    BAD_REQUEST = "BAD_REQUEST"
    OUT_OF_REQUESTS = "OUT_OF_REQUESTS"


@dataclasses.dataclass
class GraphResponse:
    status: GraphResponseStatuses
    graph_json: Optional[Graph] = None
    progress: Optional[float] = None
    remaining_requests: Optional[int] = None

Once the status falls into one of the terminal states (BAD_ID, ERROR, NOT_IN_DB, BAD_TOKEN, BAD_REQUEST, OUT_OF_REQUESTS), the iterator will cease to yield further values.

Here's the signature for invoking this method:

class ConnectedPapersClient:
    # ...
    async def get_graph_async_iterator(
        self, paper_id: str, fresh_only: bool = False, loop_until_fresh: bool = True
    ) -> AsyncIterator[GraphResponse]:
        # ...

Call this method with fresh_only=False and loop_until_fresh=True to request the existing graph and continue waiting for a rebuild if necessary.

The initial response will contain the status GraphResponseStatuses.OLD_GRAPH, then transition through GraphResponseStatuses.QUEUED and GraphResponseStatuses.IN_PROGRESS, with the progress field reflecting the percentage of the graph build completed. Upon completion of the rebuild, the status will change to GraphResponseStatuses.FRESH_GRAPH, and the iteration will end.

The graph_json field will contain the graph corresponding to each of these responses, and will remain a non-None as long as there is any version of the graph available.