> ## Documentation Index
> Fetch the complete documentation index at: https://docs.trynia.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# PDF Indexing
> Index and chat with any complex PDF using tree-guided hybrid search
Index any PDF — research papers, SEC filings, legal documents, technical manuals — and have intelligent conversations with accurate, section-aware retrieval.
***
## Tree-Guided Hybrid Search
Traditional RAG systems split documents into flat chunks, losing structure and context. Nia invented a different approach:
PDFs are parsed into hierarchical structures — sections, subsections, figures, tables — preserving the document's logical organization.
Each section is indexed with its position in the hierarchy, enabling precise retrieval that knows *where* information lives.
Combines vector embeddings with non-vector signals (headers, page numbers, cross-references) for high recall with minimal context bloat.
Agents traverse documents as trees instead of scanning flat chunks, finding relevant sections faster and more accurately.
**Why it matters:** High recall across large PDF corpora without stuffing context windows with irrelevant chunks. [Read the technical deep-dive →](https://www.trynia.ai/blog/tree-reasoning-pdf-search)
***
## Interactive Papers Playground
Experience PDFs like never before at [app.trynia.ai/playground/papers](https://app.trynia.ai/playground/papers):
View documents directly in the browser with a full-featured PDF viewer. Resizable panes let you read and chat side-by-side.
Navigate using an intelligent document tree. Jump to specific sections, figures, or equations instantly.
Your conversations are saved. Return anytime to continue where you left off, or review past discussions.
Full LaTeX support for mathematical expressions. Complex equations render beautifully in chat responses.
***
## Document Agent & Data Extraction
Once a PDF is indexed, you can go beyond search with two powerful capabilities:
Deploy an autonomous AI agent into your PDF. It plans its own research strategy, uses tools to search, read, and navigate — then delivers cited answers or structured output via JSON schemas.
Extract structured records from PDFs using JSON schemas. Pull tables, financial data, specifications, and more into clean structured formats.
***
## How It Works
Provide an arXiv URL, paper ID, or upload directly:
* *"Index [https://arxiv.org/abs/2401.12345](https://arxiv.org/abs/2401.12345)"*
* *"Index this SEC filing"*
Nia parses the PDF structure — sections, headers, figures, tables — building a navigable document tree.
Ask questions and get answers grounded in specific sections, with citations pointing to exact locations.
***
## Use Cases
Index arXiv papers for literature reviews. Search across dozens of papers at once, compare methodologies, understand dense sections.
Navigate 10-Ks, 10-Qs, and proxy statements. Find specific disclosures, compare across years, extract financial data.
Index contracts, briefs, and regulations. Search for specific clauses, understand obligations, compare versions.
Index product documentation, API specs, or internal wikis. Get precise answers with section references.
***
## Quick Start
```bash theme={null}
# In your AI coding agent, just say:
"Index https://arxiv.org/abs/2401.12345"
# Then ask questions:
"What problem does this paper solve?"
"Summarize section 3.2"
"What are the limitations mentioned by the authors?"
```
Or use the [Papers Playground](https://app.trynia.ai/playground/papers) for a visual experience.
***
## Supported Sources
| Source | Format | Example |
| ---------- | ------------------ | -------------------------------------------------- |
| arXiv | URL or ID | `https://arxiv.org/abs/2401.12345` or `2401.12345` |
| arXiv PDF | Direct PDF link | `https://arxiv.org/pdf/2401.12345` |
| PDF Upload | Direct file upload | Upload via the Papers Playground |
You can upload PDFs directly in the [Papers Playground](https://app.trynia.ai/playground/papers) — just drag and drop or use the upload button.