Developer Portal

Service Reflection

Our RPC endpoints support server reflection. You can use grpcurl to inspect services such as DatasetService and EnrichService directly from your terminal.

Executing a Query

Once you have inspected the services, you can make direct RPC calls. Here is an example of querying DatasetService.Query using the same JSON payload structure as the REST API.

Native Streaming

For large-scale vector embeddings or multi-gigabyte dataset extractions, we highly recommend utilizing the DatasetService.Stream method via Python's grpcio library to maintain memory efficiency in your PyTorch or RAPIDS training loops.

Installation

Authentication

Set your API key as an environment variable. The SDK will pick it up automatically — no manual configuration needed.

Streaming Datasets

Use the built-in paginator to iterate over chunked legal documents. Cursor management is handled automatically.

from df import DFClient, AuthenticationError

try:
    with DFClient() as client:
        for doc in client.datasets.stream("legal", search_term="التأمين", limit=100):
            print(f"Doc category: {doc.category} | Title: {doc.title}")
            if doc.source_url:
                client.datasets.download_file(
                    doc.source_url,
                    destination=f"./archives/{doc.id}.pdf"
                )
except AuthenticationError:
    print("Invalid API Key.")

gRPC Client

Drop-in replacement for low-latency environments. Install the grpc extra and swap the client class.

from df import DFGRPCClient

with DFGRPCClient() as client:
    results, next_cursor = client.datasets.query(
        dataset="legal",
        limit=10
    )

Installation

Leverage native for await...of loops to consume large datasets without worrying about cursor management.

import { DFClient } from "dataflare-sdk";

const client = new DFClient();

async function main() {
    for await (const doc of client.datasets.stream("legal", { limit: 100 })) {
        console.log("Processing:", doc.id);
    }
}

Drop-in replacement using Server Reflection. No compiled stubs required.

import { DFGRPCClient } from "dataflare-sdk";

const rpc = new DFGRPCClient();

const [docs, next] = await rpc.datasets.query("legal", { limit: 10 });

Detailed Error Mapping

Installation

Implementation Examples

Choose between idiomatic channel-based streaming (REST) or high-performance binary protocols (gRPC).

import (
    "fmt"
    "github.com/dataflarelab/df-sdk/go"
)

client := dataflare.NewClient(&dataflare.ClientOptions{APIKey: "key"})

docChan, errChan := client.Datasets.Stream("legal", nil)

for doc := range docChan {
    fmt.Println(doc.ID)
}

import "github.com/dataflarelab/df-sdk/go"

client, _ := dataflare.NewGRPCClient("rpc.dataflare.com:443")
defer client.Close()

// Requests use binary protobuf encoding
err := client.Call(ctx, "GetDataset", req, resp)

Installation

Quick Start: Claude Desktop

Add the following configuration to your claude_desktop_config.json to start using Dataflare tools in Claude.

{
  "mcpServers": {
    "dataflare": {
      "command": "npx",
      "args": ["-y", "dataflare-mcp-server"],
      "env": {
        "DF_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Quick Start