Document Ingestion

Upload and process your documents with state-of-the-art AI. Supports PDFs, DOCX, XLSX, PPTX, HTML, CSV, images (PNG/JPG/GIF/BMP/TIFF/WEBP), and audio files (MP3/WAV/M4A/MP4/WebM).

New to the platform? Check out our comprehensive documentation to learn how to get API keys and optimize your ingestion pipeline.

1. Select Files

Upload documents, images, or audio

Upload Your Documents

Upload documents, images, or audio files to create your vector database

Drag & drop files here, or click to browse

Supports PDF, DOCX, XLSX, PPTX, TXT, MD, HTML, CSV, images (PNG/JPG/GIF/etc), and audio (MP3/WAV/M4A/etc) files (max 50MB each)

2. Configure

Customize processing options

Pinecone Configuration

Index Name *

Lowercase letters, numbers, and hyphens only

Namespace

Organize data by namespace (e.g., "public", "internal", "confidential")

Pinecone API Key *

OpenAI API Key (Required)

OpenAI API Key *

Used for embeddings and, in enhanced mode, image descriptions and audio transcription.

💡 Pro Tip: API-powered pipeline uses token-aware ~1000 char chunks with GPT-4o Vision for images and Whisper for audio. All embeddings and AI generation handled automatically!

Enhanced Pipeline Options

Use enhanced multimodal pipeline

Enabled by default for documents, images, and audio. Disable for the lightweight document-only fallback.

Enable audio transcription

Uses Whisper in enhanced mode. The lightweight pipeline does not process audio or images.

⚠️ Advanced Options

Recreate Index (⚠️ DELETES EXISTING DATA)

WARNING: This will delete the existing index and all its data before ingesting.

⚡ Retrieval compatibility: Hybrid ingestion is controlled by the server configuration. It requires a dotproduct index and reingestion to add sparse values to existing dense-only records.

🚀 Pipeline Summary

• Index: Not set
• Namespace: default
• Mode: Enhanced multimodal
• Audio: Whisper transcription enabled
• Hybrid search: Server-controlled; requires dotproduct and sparse values
• Architecture: CPU-only local parsing with hosted provider APIs

Deployment Access Key (optional)

Stored only for this browser tab and sent as X-API-Key.

Select files above to continue

Multimodal Support

Process PDFs, Word docs, images (PNG/JPG), and audio files (MP3/WAV) with GPT-4o Vision and Whisper

API-Powered Reliability

100% OpenAI API processing ensures consistent, reliable results without local model complexity

Production Ready

Retry logic, error handling, parallel processing, and comprehensive progress tracking