Retrieval-augmented generation, now in your browser

Chat with your PDFs & web like never before

Powered by Pinecone 768-dim vector search and Google Gemini 2.5 Flash for sub-second retrieval, hybrid sparse-dense rank fusion, and zero-hallucination document synthesis.

pdf-ai / workspace
COLLECTIONS6
DOCUMENTS142
PAGES CRAWLED318
What is the aggregate liability cap under Section 9.2 of the Master Services Agreement?

Under Section 9.2 (Limitation of Liability), the total aggregate liability of either party for all claims arising out of this Agreement is strictly capped at 2.0× the total fees paid in the preceding 12-month period, excluding instances of gross negligence or willful misconduct.

[Contract-Q3.pdf, p.12]✓ 98.4% Grounded Match
Gemini-Powered
Your Keys Local Only
Pinecone Vector Search
PDF + Web Ingestion
Features

Everything you need to ground AI in your data

Built from scratch for deep document extraction, parallel vector indexing, and zero-hallucination verification.

High-Throughput

Multi-PDF Vector Collections

Upload research papers, financial audits, or contracts. Documents are parsed, split into optimal semantic chunks, and embedded in parallel batches into isolated Pinecone namespaces.

Open workspace
Live Ingestion

Real-Time Web Crawler

Point the crawler at any documentation URL or website. Content is scraped with BeautifulSoup & Trafilatura, converted to clean markdown, and vector indexed alongside your PDFs.

Open web crawler
Zero Hallucination

SOTA 6-Stage RAG Chat

Queries trigger HyDE expansion, Hybrid Dense (768-dim) + Sparse BM25 scoring, Reciprocal Rank Fusion (k=60), Cross-Encoder re-ranking, and CRAG grading before Google Gemini answers.

Start research session
100% Private

Bring Your Own Keys (BYOK)

Connect your personal Pinecone (pcsk_...) and Google Gemini API keys. Keys reside strictly in your browser localStorage and are sent via request headers — never stored on external servers.

Manage API vault
Capabilities

Complete Toolkit for Production RAG

Multi-PDF Collections
Web Page Crawler
Smart Semantic Chunking
Pinecone Namespaces
HyDE Query Expansion
Gemini 2.5 Flash Synthesis
BM25 Lexical Scoring
Reciprocal Rank Fusion (k=60)
Sub-Second Retrieval
CRAG Document Grader
Verbatim In-Line Citations
768-Dim Vector Embeddings
Voice-to-Text Dictation
Split-Screen Workspace
1-Click Markdown Export (.md)
100% Client-Side BYOK
Workflow

From document to answer in three steps

No complex infrastructure setup. Connect your keys and analyze documents instantly.

01

Ingest

Upload multiple PDF documents or crawl any live documentation URL to extract clean, formatted markdown content.

02

Index

Generate 768-dimensional Google embeddings and index semantic vectors in parallel into isolated serverless Pinecone namespaces.

03

Retrieve & Answer

Execute hybrid dense + sparse BM25 search, re-rank with reciprocal rank fusion, and generate cited answers with Google Gemini.

Ready to turn your documents into answers?

Jump straight in, connect your Pinecone and Gemini API keys, upload a PDF or crawl a URL, and start chatting in under two minutes.