SDK Overview - Graphor Docs

The Graphor SDKs provide convenient access to the Graphor REST API from Python and TypeScript/JavaScript applications. Both libraries include type definitions for all request params and response fields.

Python
TypeScript

The Python SDK supports Python 3.9+ and offers both synchronous and asynchronous clients.

This page provides a comprehensive overview of the SDK. It covers the full lifecycle:

Data Ingestion (Sources): Ingest, poll status, list, get elements, and manage documents by file_id
Document Chat: Ask questions about your documents with conversational memory
Data Extraction: Extract structured data using JSON Schema
Prebuilt RAG: Retrieve relevant document chunks for custom RAG pipelines

Python SDK Repository

View the Python SDK source code, report issues, and contribute.

TypeScript SDK Repository

View the TypeScript SDK source code, report issues, and contribute.

Installation

Python
TypeScript

Install the Graphor SDK from PyPI:

pip install graphor

Python 3.9 or higher is required.

Install the Graphor SDK from npm:

npm install graphor

Or with your preferred package manager:

yarn add graphor

pnpm add graphor

TypeScript 4.9+ and Node.js 20+ (LTS) are recommended. Also works in Deno v1.28+, Bun 1.0+, Cloudflare Workers, Vercel Edge Runtime, and modern browsers.

Data Ingestion (Sources)

The Sources methods cover the full ingestion lifecycle:

Ingest Source

Ingest documents from files, URLs, GitHub, and YouTube (returns build_id; poll for file_id)

Reprocess Source

Reprocess an existing source with a different partition method

List Sources

Retrieve all sources with status and metadata

List Source Elements

Retrieve structured elements/partitions from processed sources

Delete Source

Permanently remove sources from your project

Document Chat

Once your data is ingested, use the Chat method to ask questions:

Chat with Documents

Ask natural language questions about your documents with conversational memory and structured outputs

Data Extraction

Extract specific structured data from your documents using schemas:

Extract Structured Data

Extract structured information from documents using JSON Schema and natural language instructions

Prebuilt RAG

Build custom RAG pipelines with semantic document retrieval:

Retrieve Document Chunks

Retrieve relevant document chunks using semantic search for custom LLM integration

What “Data Ingestion” includes

Ingest: Create a new source (file, URL, GitHub, YouTube); returns build_id; poll get build status until ready, then use file_id
Reprocess: Reprocess an existing source with a different partition method (optional)
List: Monitor status and metadata; optionally filter by file_ids
Get elements: Retrieve structured elements/partitions by file_id after processing
Delete: Remove a source by file_id

Authentication

All SDK methods require authentication using API tokens. You can provide your API key in two ways:

Environment Variable (Recommended)

Set the GRAPHOR_API_KEY environment variable:

export GRAPHOR_API_KEY="grlm_your_api_key_here"

Python
TypeScript

from graphor import Graphor

# API key is automatically read from GRAPHOR_API_KEY
client = Graphor()

import Graphor from 'graphor';

// API key is automatically read from GRAPHOR_API_KEY
const client = new Graphor();

Direct Initialization

Python
TypeScript

from graphor import Graphor

client = Graphor(api_key="grlm_your_api_key_here")

import Graphor from 'graphor';

const client = new Graphor({ apiKey: 'grlm_your_api_key_here' });

Learn how to generate and manage API tokens in the API Tokens guide.

Token Security

Never expose tokens in client-side code or public repositories
Use environment variables to store tokens securely
Rotate tokens regularly for enhanced security
Use different tokens for different environments (dev/staging/prod)

Async Usage

Python
TypeScript

Simply import AsyncGraphor instead of Graphor and use await with each API call:

import asyncio
from graphor import AsyncGraphor

client = AsyncGraphor()

async def main():
    build_id = await client.sources.ingest_file(file=b"raw file contents")
    print(build_id)

asyncio.run(main())

All TypeScript SDK methods are async by default and return Promises:

import Graphor from 'graphor';
import fs from 'fs';

const client = new Graphor();

async function main() {
  const buildId = await client.sources.ingestFile({ file: fs.createReadStream('document.pdf') });
  console.log(buildId);
}

main();

Available Methods

Sources

Python
TypeScript

Method	Description
`client.sources.ingest_file()`	Ingest a local file (returns `build_id`)
`client.sources.ingest_url()`	Ingest from a web URL
`client.sources.ingest_github()`	Ingest from GitHub
`client.sources.ingest_youtube()`	Ingest from YouTube
`client.sources.get_build_status()`	Poll build status; returns `file_id` when ready
`client.sources.reprocess()`	Reprocess a source by `file_id` (returns `build_id`)
`client.sources.list()`	List all sources (optional `file_ids` filter)
`client.sources.get_elements()`	Get parsed elements by `file_id`
`client.sources.delete()`	Delete a source by `file_id`

Method	Description
`client.sources.ingestFile()`	Ingest a local file (returns `build_id`)
`client.sources.ingestUrl()`	Ingest from a web URL
`client.sources.ingestGitHub()`	Ingest from GitHub
`client.sources.ingestYoutube()`	Ingest from YouTube
`client.sources.getBuildStatus()`	Poll build status; returns `fileId` when ready
`client.sources.reprocess()`	Reprocess a source by `fileId` (returns `build_id`)
`client.sources.list()`	List all sources (optional `fileIds` filter)
`client.sources.getElements()`	Get parsed elements by `fileId`
`client.sources.delete()`	Delete a source by `fileId`

Chat & Extraction

Python
TypeScript

Method	Description
`client.sources.ask()`	Ask questions about your documents
`client.sources.extract()`	Extract structured data using JSON Schema
`client.sources.retrieve_chunks()`	Retrieve relevant chunks for custom RAG

Method	Description
`client.sources.ask()`	Ask questions about your documents
`client.sources.extract()`	Extract structured data using JSON Schema
`client.sources.retrieveChunks()`	Retrieve relevant chunks for custom RAG

Complete Workflow Example

Here’s the full “happy path”: ingest → get_build_status (poll) → list → get_elements → chat/extract/rag; optionally reprocess by file_id.

1. Ingest a source

Python
TypeScript

from pathlib import Path
from graphor import Graphor
import time

client = Graphor()

# Ingest returns build_id; poll get_build_status until ready
build_id = client.sources.ingest_file(file=Path("./document.pdf"))
while True:
    status = client.sources.get_build_status(build_id)
    if status.success and getattr(status, "file_id", None):
        file_id = status.file_id
        print(f"Ready. file_id: {file_id}")
        break
    time.sleep(2)

import Graphor from 'graphor';
import fs from 'fs';

const client = new Graphor();

const buildId = await client.sources.ingestFile({ file: fs.createReadStream('./document.pdf') });
let fileId: string;
while (true) {
  const status = await client.sources.getBuildStatus(buildId);
  if (status.success && status.fileId) {
    fileId = status.fileId;
    console.log('Ready. file_id:', fileId);
    break;
  }
  await new Promise((r) => setTimeout(r, 2000));
}

2. Reprocess (optional)

Python
TypeScript

build_id = client.sources.reprocess(file_id=file_id, method="balanced")
print(f"Reprocessing: {build_id}")

const reprocessBuildId = await client.sources.reprocess({ file_id: fileId, method: 'balanced' });
console.log('Reprocessing:', reprocessBuildId);

3. List sources

Python
TypeScript

sources = client.sources.list()
for s in sources:
    print(f"{s.file_id} {s.file_name}: {s.status}")
target = next((s for s in sources if s.file_id == file_id), None)

const sources = await client.sources.list();
for (const s of sources) console.log(`${s.fileId} ${s.fileName}: ${s.status}`);
const target = sources.find((s) => s.fileId === fileId);

4. Get elements (after processing)

Python
TypeScript

elements = client.sources.get_elements(file_id=file_id, page=1, page_size=50)
print(f"Total elements: {elements.total}")
for item in elements.items:
    print(f"  [{item.element_type}] {item.text[:100]}...")

const elements = await client.sources.getElements({ file_id: fileId, page: 1, page_size: 50 });
console.log(`Total elements: ${elements.total}`);
for (const item of elements.items) {
  console.log(`  [${item.element_type}] ${item.text.slice(0, 100)}...`);
}

5. Ask questions (Chat)

Python
TypeScript

response = client.sources.ask(
    question="What are the main topics in this document?",
    file_ids=[file_id]  # Optional: scope to specific sources
)
print(response.answer)
follow_up = client.sources.ask(
    question="Can you elaborate on the first topic?",
    conversation_id=response.conversation_id
)
print(follow_up.answer)

const response = await client.sources.ask({
  question: 'What are the main topics in this document?',
  fileIds: [fileId],
});
console.log(response.answer);
const followUp = await client.sources.ask({
  question: 'Can you elaborate on the first topic?',
  conversationId: response.conversation_id,
});
console.log(followUp.answer);

6. Extract data

Python
TypeScript

result = client.sources.extract(
    file_ids=[file_id],
    user_instruction="Extract the invoice number and total amount.",
    output_schema={
        "type": "object",
        "properties": {
            "invoice_number": {"type": "string"},
            "total_amount": {"type": "number"}
        },
        "required": ["invoice_number", "total_amount"]
    }
)
print(result.structured_output)

const result = await client.sources.extract({
  fileIds: [fileId],
  user_instruction: 'Extract the invoice number and total amount.',
  output_schema: {
    type: 'object',
    properties: {
      invoice_number: { type: 'string' },
      total_amount: { type: 'number' },
    },
    required: ['invoice_number', 'total_amount'],
  },
});
console.log(result.structured_output);

7. Retrieve chunks (Prebuilt RAG)

Python
TypeScript

chunks = client.sources.retrieve_chunks(
    query="What are the payment terms?",
    file_ids=[file_id]
)
for chunk in chunks.chunks:
    print(f"[{chunk.file_id}, Page {chunk.page_number}]", chunk.text[:80])

const chunks = await client.sources.retrieveChunks({
  query: 'What are the payment terms?',
  fileIds: [fileId],
});
for (const chunk of chunks.chunks) {
  console.log(`[${chunk.fileId}, Page ${chunk.page_number}]`, chunk.text.slice(0, 80));
}

Integration Patterns

Complete SDK Client Wrapper

Python
TypeScript

from graphor import Graphor, AsyncGraphor
import graphor
from pathlib import Path
from typing import Any
import os


class GraphorSDK:
    """Complete wrapper for common Graphor operations."""
    
    def __init__(self, api_key: str | None = None):
        self.client = Graphor(api_key=api_key) if api_key else Graphor()
    
    # ==================== Sources ====================
    
    def ingest_file(self, file_path: str | Path) -> str:
        """Ingest a file; returns build_id. Poll get_build_status for file_id."""
        return self.client.sources.ingest_file(file=Path(file_path))
    
    def get_build_status(self, build_id: str) -> Any:
        """Poll build status; when success, response has file_id."""
        return self.client.sources.get_build_status(build_id)
    
    def ingest_url(self, url: str, crawl: bool = False) -> str:
        """Ingest from a URL; returns build_id."""
        return self.client.sources.ingest_url(url=url, crawl_urls=crawl)
    
    def reprocess(self, file_id: str, method: str = "balanced") -> str:
        """Reprocess a source; returns build_id."""
        return self.client.sources.reprocess(file_id=file_id, method=method)
    
    def list_sources(self) -> list[dict[str, Any]]:
        """List all sources."""
        sources = self.client.sources.list()
        return [
            {"file_id": s.file_id, "file_name": s.file_name, "status": s.status}
            for s in sources
        ]
    
    def get_elements(
        self, file_id: str, page: int = 1, page_size: int = 50
    ) -> dict[str, Any]:
        """Get parsed elements from a source."""
        result = self.client.sources.get_elements(
            file_id=file_id, page=page, page_size=page_size
        )
        return {
            "total": result.total,
            "page": result.page,
            "total_pages": result.total_pages,
            "items": [
                {"type": item.element_type, "content": item.text, "page": item.page_number}
                for item in result.items
            ]
        }
    
    def delete(self, file_id: str) -> dict[str, Any]:
        """Delete a source by file_id."""
        result = self.client.sources.delete(file_id=file_id)
        return {"message": result.message}
    
    # ==================== Chat ====================
    
    def ask(
        self,
        question: str,
        file_ids: list[str] | None = None,
        conversation_id: str | None = None
    ) -> dict[str, Any]:
        """Ask a question about documents."""
        kwargs = {"question": question}
        if file_ids:
            kwargs["file_ids"] = file_ids
        if conversation_id:
            kwargs["conversation_id"] = conversation_id
        response = self.client.sources.ask(**kwargs)
        return {"answer": response.answer, "conversation_id": response.conversation_id}
    
    # ==================== Extraction ====================
    
    def extract(
        self, file_ids: list[str], instruction: str, schema: dict[str, Any]
    ) -> dict[str, Any]:
        """Extract structured data from documents."""
        result = self.client.sources.extract(
            file_ids=file_ids,
            user_instruction=instruction,
            output_schema=schema
        )
        return {"data": result.structured_output, "raw": result.raw_json}
    
    # ==================== RAG ====================
    
    def retrieve(
        self, query: str, file_ids: list[str] | None = None
    ) -> dict[str, Any]:
        """Retrieve relevant chunks for RAG."""
        kwargs = {"query": query}
        if file_ids:
            kwargs["file_ids"] = file_ids
        result = self.client.sources.retrieve_chunks(**kwargs)
        return {
            "query": result.query,
            "total": result.total,
            "chunks": [
                {"text": c.text, "file_id": c.file_id, "page": c.page_number, "score": c.score}
                for c in result.chunks or []
            ]
        }


# Usage example
sdk = GraphorSDK()
import time

def full_workflow(file_path: str):
    """Complete ingestion, chat, and extraction workflow."""
    try:
        build_id = sdk.ingest_file(file_path)
        while True:
            status = sdk.get_build_status(build_id)
            if status.success and getattr(status, "file_id", None):
                file_id = status.file_id
                break
            time.sleep(2)
        print(f"Ready: {file_id}")
        chat_result = sdk.ask("Summarize this document", [file_id])
        print(f"Summary: {chat_result['answer']}")
        extract_result = sdk.extract(
            [file_id],
            "Extract key information",
            {"type": "object", "properties": {"title": {"type": "string"}, "summary": {"type": "string"}}}
        )
        print(f"Extracted: {extract_result['data']}")
        return {"success": True, "file_id": file_id}
    except graphor.APIStatusError as e:
        print(f"Error: {e}")
        return {"success": False, "error": str(e)}

import Graphor from 'graphor';
import fs from 'fs';

class GraphorSDK {
  private client: Graphor;

  constructor(apiKey?: string) {
    this.client = apiKey ? new Graphor({ apiKey }) : new Graphor();
  }

  // ==================== Sources ====================

  async ingestFile(filePath: string) {
    return this.client.sources.ingestFile({ file: fs.createReadStream(filePath) });
  }

  async getBuildStatus(buildId: string) {
    return this.client.sources.getBuildStatus(buildId);
  }

  async ingestUrl(url: string, crawl = false) {
    return this.client.sources.ingestUrl({ url, crawlUrls: crawl });
  }

  async reprocess(fileId: string, method = 'balanced') {
    return this.client.sources.reprocess({ file_id: fileId, method });
  }

  async listSources() {
    const sources = await this.client.sources.list();
    return sources.map((s) => ({ fileId: s.file_id, fileName: s.file_name, status: s.status }));
  }

  async getElements(fileId: string, page = 1, page_size = 50) {
    const result = await this.client.sources.getElements({ file_id: fileId, page, page_size });
    return {
      total: result.total,
      page: result.page,
      totalPages: result.total_pages,
      items: result.items.map((item) => ({ type: item.element_type, content: item.text, page: item.page_number })),
    };
  }

  async delete(fileId: string) {
    const result = await this.client.sources.delete({ file_id: fileId });
    return { message: result.message };
  }

  // ==================== Chat ====================

  async ask(question: string, fileIds?: string[], conversationId?: string) {
    const response = await this.client.sources.ask({
      question,
      fileIds,
      conversationId,
    });
    return { answer: response.answer, conversationId: response.conversation_id };
  }

  // ==================== Extraction ====================

  async extract(fileIds: string[], instruction: string, schema: Record<string, unknown>) {
    const result = await this.client.sources.extract({
      fileIds,
      user_instruction: instruction,
      output_schema: schema,
    });
    return { data: result.structured_output, raw: result.raw_json };
  }

  // ==================== RAG ====================

  async retrieve(query: string, fileIds?: string[]) {
    const result = await this.client.sources.retrieveChunks({ query, fileIds });
    return {
      query: result.query,
      total: result.total,
      chunks: (result.chunks ?? []).map((c) => ({ text: c.text, fileId: c.file_id, page: c.page_number, score: c.score })),
    };
  }
}

// Usage example
const sdk = new GraphorSDK();

async function fullWorkflow(filePath: string) {
  try {
    const buildId = await sdk.ingestFile(filePath);
    let fileId: string;
    while (true) {
      const status = await sdk.getBuildStatus(buildId);
      if (status.success && status.fileId) {
        fileId = status.fileId;
        break;
      }
      await new Promise((r) => setTimeout(r, 2000));
    }
    console.log(`Ready: ${fileId}`);
    const chatResult = await sdk.ask('Summarize this document', [fileId]);
    console.log(`Summary: ${chatResult.answer}`);
    const extractResult = await sdk.extract(
      [fileId],
      'Extract key information',
      { type: 'object', properties: { title: { type: 'string' }, summary: { type: 'string' } } },
    );
    console.log(`Extracted: ${JSON.stringify(extractResult.data)}`);
    return { success: true, fileId };
  } catch (err) {
    if (err instanceof Graphor.APIError) {
      console.log(`Error: ${err.message}`);
      return { success: false, error: err.message };
    }
    throw err;
  }
}

Async Integration

Python
TypeScript

import asyncio
from graphor import AsyncGraphor
import graphor


class AsyncGraphorSDK:
    """Async wrapper for Graphor operations."""
    
    def __init__(self, api_key: str | None = None):
        self.client = AsyncGraphor(api_key=api_key) if api_key else AsyncGraphor()
    
    async def process_multiple(
        self, file_paths: list[str], method: str = "balanced"
    ) -> list[dict]:
        """Ingest multiple files concurrently; poll for file_id when needed."""
        from pathlib import Path
        import time

        async def ingest_one(file_path: str) -> dict:
            try:
                build_id = await self.client.sources.ingest_file(file=Path(file_path))
                while True:
                    status = await self.client.sources.get_build_status(build_id)
                    if status.success and getattr(status, "file_id", None):
                        return {"file": file_path, "status": "success", "file_id": status.file_id}
                    await asyncio.sleep(2)
            except graphor.APIStatusError as e:
                return {"file": file_path, "status": "failed", "error": str(e)}

        return await asyncio.gather(*[ingest_one(fp) for fp in file_paths])

    async def batch_ask(
        self, questions: list[str], file_ids: list[str] | None = None
    ) -> list[dict]:
        """Ask multiple questions concurrently."""
        async def ask_one(question: str) -> dict:
            response = await self.client.sources.ask(
                question=question, file_ids=file_ids
            )
            return {"question": question, "answer": response.answer}
        return await asyncio.gather(*[ask_one(q) for q in questions])


# Usage
async def main():
    sdk = AsyncGraphorSDK()
    
    # Process multiple files
    results = await sdk.process_multiple([
        "doc1.pdf",
        "doc2.pdf",
        "doc3.pdf"
    ])
    
    for r in results:
        status = "OK" if r["status"] == "success" else "FAIL"
        print(f"{status} {r['file']}")
    
    # Ask multiple questions
    answers = await sdk.batch_ask([
        "What is the main topic?",
        "Who are the key people mentioned?",
        "What are the conclusions?"
    ])
    
    for a in answers:
        print(f"Q: {a['question']}")
        print(f"A: {a['answer']}\n")

asyncio.run(main())

import Graphor from 'graphor';
import fs from 'fs';

class AsyncGraphorSDK {
  private client: Graphor;

  constructor(apiKey?: string) {
    this.client = apiKey ? new Graphor({ apiKey }) : new Graphor();
  }

  async processMultiple(filePaths: string[], method = 'balanced') {
    const ingestOne = async (filePath: string) => {
      try {
        const buildId = await this.client.sources.ingestFile({ file: fs.createReadStream(filePath) });
        let fileId: string;
        while (true) {
          const status = await this.client.sources.getBuildStatus(buildId);
          if (status.success && status.fileId) {
            fileId = status.fileId;
            break;
          }
          await new Promise((r) => setTimeout(r, 2000));
        }
        return { file: filePath, status: 'success' as const, fileId };
      } catch (err) {
        const message = err instanceof Graphor.APIError ? err.message : String(err);
        return { file: filePath, status: 'failed' as const, error: message };
      }
    };
    return Promise.all(filePaths.map(ingestOne));
  }

  async batchAsk(questions: string[], fileIds?: string[]) {
    const askOne = async (question: string) => {
      const response = await this.client.sources.ask({ question, fileIds });
      return { question, answer: response.answer };
    };
    return Promise.all(questions.map(askOne));
  }
}

// Usage
async function main() {
  const sdk = new AsyncGraphorSDK();

  // Process multiple files
  const results = await sdk.processMultiple(['doc1.pdf', 'doc2.pdf', 'doc3.pdf']);

  for (const r of results) {
    const status = r.status === 'success' ? 'OK' : 'FAIL';
    console.log(`${status} ${r.file}`);
  }

  // Ask multiple questions
  const answers = await sdk.batchAsk([
    'What is the main topic?',
    'Who are the key people mentioned?',
    'What are the conclusions?',
  ]);

  for (const a of answers) {
    console.log(`Q: ${a.question}`);
    console.log(`A: ${a.answer}\n`);
  }
}

main();

Error Handling

The SDK provides typed exceptions for different error scenarios:

Python
TypeScript

import graphor
from graphor import Graphor

client = Graphor()

try:
    build_id = client.sources.ingest_file(file=b"raw file contents")
except graphor.APIConnectionError as e:
    print("The server could not be reached")
    print(e.__cause__)
except graphor.RateLimitError as e:
    print("Rate limit exceeded. Back off and retry.")
except graphor.BadRequestError as e:
    print(f"Invalid request: {e}")
except graphor.AuthenticationError as e:
    print(f"Invalid API key: {e}")
except graphor.NotFoundError as e:
    print(f"Resource not found: {e}")
except graphor.APIStatusError as e:
    print(f"API error (status {e.status_code}): {e}")

import Graphor from 'graphor';
import fs from 'fs';

const client = new Graphor();

try {
  const buildId = await client.sources.ingestFile({ file: fs.createReadStream('file.pdf') });
} catch (err) {
  if (err instanceof Graphor.APIConnectionError) {
    console.log('The server could not be reached');
    console.log(err.cause);
  } else if (err instanceof Graphor.RateLimitError) {
    console.log('Rate limit exceeded. Back off and retry.');
  } else if (err instanceof Graphor.BadRequestError) {
    console.log(`Invalid request: ${err.message}`);
  } else if (err instanceof Graphor.AuthenticationError) {
    console.log(`Invalid API key: ${err.message}`);
  } else if (err instanceof Graphor.NotFoundError) {
    console.log(`Resource not found: ${err.message}`);
  } else if (err instanceof Graphor.APIError) {
    console.log(`API error (status ${err.status}): ${err.message}`);
  } else {
    throw err;
  }
}

Error Types

Status Code	Error Type	Description
400	`BadRequestError`	Invalid parameters or malformed request
401	`AuthenticationError`	Invalid or missing API key
403	`PermissionDeniedError`	Access denied to resource
404	`NotFoundError`	Resource doesn’t exist
422	`UnprocessableEntityError`	Validation error
429	`RateLimitError`	Too many requests
≥500	`InternalServerError`	Server-side error
N/A	`APIConnectionError`	Network connectivity issues
N/A	`APITimeoutError`	Request timed out

Configuration

Retries

Certain errors are automatically retried 2 times by default with exponential backoff:

Python
TypeScript

from graphor import Graphor

# Configure default retries
client = Graphor(max_retries=0)  # Disable retries

# Or per-request
client.with_options(max_retries=5).sources.ingest_file(file=b"...")

import Graphor from 'graphor';

// Configure default retries
const client = new Graphor({ maxRetries: 0 }); // Disable retries

// Or per-request
await client.sources.ask({ question: '...' }, { maxRetries: 5 });

Timeouts

Python
TypeScript

By default, requests time out after 1 minute:

from graphor import Graphor

# Configure default timeout (in seconds)
client = Graphor(timeout=120.0)  # 2 minutes

# Or per-request
client.with_options(timeout=300.0).sources.reprocess(
    file_id="file_abc123",
    method="agentic"
)

By default, requests time out after 60 seconds:

import Graphor from 'graphor';

// Configure default timeout (in milliseconds)
const client = new Graphor({ timeout: 120 * 1000 }); // 2 minutes

// Or per-request
await client.sources.reprocess(
  { file_id: 'file_abc123', method: 'agentic' },
  { timeout: 300 * 1000 },
);

Using aiohttp for Better Concurrency (Python only)

For high-concurrency async operations in Python, use the aiohttp client:

import asyncio
from graphor import AsyncGraphor, DefaultAioHttpClient

async def main():
    async with AsyncGraphor(
        http_client=DefaultAioHttpClient()
    ) as client:
        # Your async operations here
        sources = await client.sources.list()
        print(f"Found {len(sources)} sources")

# Install aiohttp first: pip install graphor[aiohttp]
asyncio.run(main())

Rate Limits and Best Practices

Performance Guidelines

Batch Operations: Process multiple files sequentially or with controlled concurrency
Async Processing: Use AsyncGraphor (Python) or Promise.all (TypeScript) for concurrent operations
Retry Logic: The SDK handles retries automatically; configure max_retries / maxRetries as needed
Timeout Handling: Increase timeouts for large documents or complex processing

Best Practices

Python
TypeScript

from graphor import Graphor
import graphor
import time

client = Graphor(
    max_retries=3,
    timeout=120.0  # 2 minutes for processing operations
)

def robust_upload(file_path: str, max_attempts: int = 3) -> dict | None:
    """Ingest with custom retry logic; poll until ready and return file_id."""
    from pathlib import Path
    import time
    for attempt in range(max_attempts):
        try:
            build_id = client.sources.ingest_file(file=Path(file_path))
            while True:
                status = client.sources.get_build_status(build_id)
                if status.success and getattr(status, "file_id", None):
                    return {"success": True, "file_id": status.file_id}
                time.sleep(2)
        except graphor.RateLimitError:
            wait_time = 2 ** attempt  # Exponential backoff
            print(f"Rate limited. Waiting {wait_time}s...")
            time.sleep(wait_time)
        except graphor.APIConnectionError as e:
            print(f"Connection error (attempt {attempt + 1}): {e}")
            time.sleep(1)
        except graphor.APIStatusError as e:
            print(f"API error: {e}")
            return {"success": False, "error": str(e)}
    
    return {"success": False, "error": "Max retries exceeded"}

import Graphor from 'graphor';
import fs from 'fs';

const client = new Graphor({
  maxRetries: 3,
  timeout: 120 * 1000, // 2 minutes for processing operations
});

async function robustUpload(filePath: string, maxAttempts = 3) {
  for (let attempt = 0; attempt < maxAttempts; attempt++) {
    try {
      const buildId = await client.sources.ingestFile({ file: fs.createReadStream(filePath) });
      while (true) {
        const status = await client.sources.getBuildStatus(buildId);
        if (status.success && status.fileId) return { success: true, fileId: status.fileId };
        await new Promise((r) => setTimeout(r, 2000));
      }
    } catch (err) {
      if (err instanceof Graphor.RateLimitError) {
        const waitTime = 2 ** attempt;
        console.log(`Rate limited. Waiting ${waitTime}s...`);
        await new Promise((r) => setTimeout(r, waitTime * 1000));
      } else if (err instanceof Graphor.APIConnectionError) {
        console.log(`Connection error (attempt ${attempt + 1}): ${err.message}`);
        await new Promise((r) => setTimeout(r, 1000));
      } else if (err instanceof Graphor.APIError) {
        console.log(`API error: ${err.message}`);
        return { success: false, error: err.message };
      } else {
        throw err;
      }
    }
  }
  return { success: false, error: 'Max retries exceeded' };
}

Common Use Cases

Document Processing Pipeline

Python
TypeScript

from graphor import Graphor
from pathlib import Path
import graphor

client = Graphor(timeout=300.0)

import time

def document_pipeline(directory: str, partition_method: str = "balanced"):
    """Ingest all PDFs in a directory; poll until ready."""
    results = []
    for file_path in Path(directory).glob("*.pdf"):
        try:
            build_id = client.sources.ingest_file(file=file_path)
            while True:
                status = client.sources.get_build_status(build_id)
                if status.success and getattr(status, "file_id", None):
                    results.append({"file": str(file_path), "status": "success", "file_id": status.file_id})
                    print(f"Ready: {status.file_id}")
                    break
                time.sleep(2)
        except graphor.APIStatusError as e:
            results.append({"file": str(file_path), "status": "failed", "error": str(e)})
    print(f"\nProcessed {sum(1 for r in results if r['status'] == 'success')}/{len(results)} files")
    return results

import Graphor from 'graphor';
import fs from 'fs';
import path from 'path';

const client = new Graphor({ timeout: 300 * 1000 });

async function documentPipeline(directory: string, partitionMethod = 'balanced') {
  const files = fs.readdirSync(directory).filter((f) => f.endsWith('.pdf'));
  const results = [];
  for (const file of files) {
    const filePath = path.join(directory, file);
    try {
      const buildId = await client.sources.ingestFile({ file: fs.createReadStream(filePath) });
      let fileId: string;
      while (true) {
        const status = await client.sources.getBuildStatus(buildId);
        if (status.success && status.fileId) {
          fileId = status.fileId;
          console.log(`Ready: ${fileId}`);
          break;
        }
        await new Promise((r) => setTimeout(r, 2000));
      }

      results.push({ file: filePath, status: 'success', fileId });
    } catch (err) {
      const message = err instanceof Graphor.APIError ? err.message : String(err);
      results.push({ file: filePath, status: 'failed', error: message });
    }
  }

  const successful = results.filter((r) => r.status === 'success').length;
  console.log(`\nProcessed ${successful}/${results.length} files`);

  return results;
}

Q&A System

Python
TypeScript

from graphor import Graphor

client = Graphor()

class DocumentQA:
    """Simple Q&A system with conversation history."""
    
    def __init__(self, file_ids: list[str] | None = None):
        self.file_ids = file_ids
        self.conversation_id = None
    
    def ask(self, question: str) -> str:
        """Ask a question, maintaining conversation history."""
        response = client.sources.ask(
            question=question,
            file_ids=self.file_ids,
            conversation_id=self.conversation_id
        )
        
        # Store conversation ID for follow-up questions
        self.conversation_id = response.conversation_id
        
        return response.answer
    
    def reset(self):
        """Reset conversation history."""
        self.conversation_id = None


# Usage (file_ids from list())
qa = DocumentQA(file_ids=["file_abc123"])

print(qa.ask("What is this document about?"))
print(qa.ask("What are the main findings?"))  # Follow-up
print(qa.ask("Can you summarize the conclusions?"))  # Follow-up

qa.reset()  # Start new conversation

import Graphor from 'graphor';

const client = new Graphor();

class DocumentQA {
  private fileIds?: string[];
  private conversationId?: string;

  constructor(fileIds?: string[]) {
    this.fileIds = fileIds;
  }

  async ask(question: string): Promise<string> {
    const response = await client.sources.ask({
      question,
      fileIds: this.fileIds,
      conversationId: this.conversationId,
    });

    // Store conversation ID for follow-up questions
    this.conversationId = response.conversation_id ?? undefined;

    return response.answer;
  }

  reset() {
    this.conversationId = undefined;
  }
}

// Usage (fileIds from list())
const qa = new DocumentQA(['file_abc123']);

console.log(await qa.ask('What is this document about?'));
console.log(await qa.ask('What are the main findings?')); // Follow-up
console.log(await qa.ask('Can you summarize the conclusions?')); // Follow-up

qa.reset(); // Start new conversation

Custom RAG with Your LLM

Python
TypeScript

from graphor import Graphor
from openai import OpenAI

graphor_client = Graphor()
openai_client = OpenAI()

def custom_rag(question: str, file_ids: list[str] | None = None) -> dict:
    """Custom RAG pipeline with OpenAI."""
    chunks = graphor_client.sources.retrieve_chunks(
        query=question, file_ids=file_ids
    )
    context = "\n\n".join([
        f"[{c.file_id}, Page {c.page_number}]\n{c.text}"
        for c in chunks.chunks or []
    ])
    response = openai_client.chat.completions.create(
        model="gpt-4",
        messages=[
            {"role": "system", "content": "Answer questions based on the provided context. Cite sources."},
            {"role": "user", "content": f"Context:\n{context}\n\nQuestion: {question}"}
        ]
    )
    return {
        "answer": response.choices[0].message.content,
        "sources": [{"file_id": c.file_id, "page": c.page_number} for c in chunks.chunks or []]
    }

# Usage (pass file_ids from list())
result = custom_rag("What are the payment terms?", ["file_abc123"])
print(result["answer"])
print("Sources:", result["sources"])

import Graphor from 'graphor';
import OpenAI from 'openai';

const graphorClient = new Graphor();
const openaiClient = new OpenAI();

async function customRag(question: string, fileIds?: string[]) {
  const chunks = await graphorClient.sources.retrieveChunks({
    query: question,
    fileIds,
  });
  const context = (chunks.chunks ?? [])
    .map((c) => `[${c.fileId}, Page ${c.page_number}]\n${c.text}`)
    .join('\n\n');
  const response = await openaiClient.chat.completions.create({
    model: 'gpt-4',
    messages: [
      { role: 'system', content: 'Answer questions based on the provided context. Cite sources.' },
      { role: 'user', content: `Context:\n${context}\n\nQuestion: ${question}` },
    ],
  });
  return {
    answer: response.choices[0].message.content,
    sources: (chunks.chunks ?? []).map((c) => ({ fileId: c.file_id, page: c.page_number })),
  };
}

// Usage (pass fileIds from list())
const result = await customRag('What are the payment terms?', ['file_abc123']);
console.log(result.answer);
console.log('Sources:', result.sources);

Support and Resources

Getting Help

Contact Support

Direct support for technical questions and issues

API Tokens Guide

Learn how to generate and manage authentication tokens

Data Ingestion Guide

Best practices for document upload and processing

REST API Reference

Full REST API documentation for advanced use cases

Next Steps

Ready to start building with the Graphor SDK? Choose your path:

For Beginners

Ingest Sources

Ingest documents from files, URLs, GitHub, and YouTube; poll for file_id

Chat with Documents

Ask natural language questions about your documents

API Tokens

Set up authentication for API access

For Advanced Users

Data Extraction

Extract structured data using JSON Schema

Prebuilt RAG

Build custom RAG pipelines with semantic search

Reprocess Source

Reprocess sources with different partition methods

List Elements

Access structured document elements and metadata

The Graphor SDKs provide a powerful foundation for building intelligent, document-driven applications. With comprehensive support for document ingestion, conversational AI, structured extraction, and custom RAG pipelines, both the Python and TypeScript SDKs give you the flexibility to build sophisticated AI workflows that scale from simple document search to complex analysis systems.

Get Started

Data SDK Options

Python SDK Repository

TypeScript SDK Repository

​Installation

​Data Ingestion (Sources)

Ingest Source

Reprocess Source

List Sources

List Source Elements

Delete Source

​Document Chat

Chat with Documents

​Data Extraction

Extract Structured Data

​Prebuilt RAG

Retrieve Document Chunks

​What “Data Ingestion” includes

​Authentication

​Environment Variable (Recommended)

​Direct Initialization

​Token Security

​Async Usage

​Available Methods

​Sources

​Chat & Extraction

​Complete Workflow Example

​1. Ingest a source

​2. Reprocess (optional)

​3. List sources

​4. Get elements (after processing)

​5. Ask questions (Chat)

​6. Extract data

​7. Retrieve chunks (Prebuilt RAG)

​Integration Patterns

​Complete SDK Client Wrapper

​Async Integration

​Error Handling

​Error Types

​Configuration

​Retries

​Timeouts

​Using aiohttp for Better Concurrency (Python only)

​Rate Limits and Best Practices

​Performance Guidelines

​Best Practices

​Common Use Cases

​Document Processing Pipeline

​Q&A System

​Custom RAG with Your LLM

​Support and Resources

​Getting Help

Contact Support

API Tokens Guide

Data Ingestion Guide

REST API Reference

​Next Steps

​For Beginners

Ingest Sources

Chat with Documents

API Tokens

​For Advanced Users

Data Extraction

Prebuilt RAG

Reprocess Source

List Elements

Installation

Data Ingestion (Sources)

Document Chat

Data Extraction

Prebuilt RAG

What “Data Ingestion” includes

Authentication

Environment Variable (Recommended)

Direct Initialization

Token Security

Async Usage

Available Methods

Sources

Chat & Extraction

Complete Workflow Example

1. Ingest a source

2. Reprocess (optional)

3. List sources

4. Get elements (after processing)

5. Ask questions (Chat)

6. Extract data

7. Retrieve chunks (Prebuilt RAG)

Integration Patterns

Complete SDK Client Wrapper

Async Integration

Error Handling

Error Types

Configuration

Retries

Timeouts

Using aiohttp for Better Concurrency (Python only)

Rate Limits and Best Practices

Performance Guidelines

Best Practices

Common Use Cases

Document Processing Pipeline

Q&A System

Custom RAG with Your LLM

Support and Resources

Getting Help

Next Steps

For Beginners

For Advanced Users