wdoc
Summarize and query massive heterogeneous document corpora with advanced RAG.
Query tens of thousands of files, audio, and flashcards with custom LLM pipelines.
wdoc is a specialized RAG system built to summarize and search massive collections of diverse document formats. It suits researchers, students, and professionals processing complex information from multiple varied sources.
What it is
It parses 15+ file types including PDFs, audio, video lectures, and Anki flashcards, using multi-query generation and semantic batching. This creates sourced, indented markdown answers that point directly to exact portions of your source documents.
Who it's for
It suits professionals and researchers drowning in varied media who need definitive, sourced answers from large document sets. It is not ideal for casual users looking for a simple point-and-click cloud app without handling CLI tools or API keys.
Key features
Heterogeneous File Support
Processes over 15 distinct file types including audio, video, EPUBs, and Anki flashcards simultaneously.
High-Recall Search Pipeline
Employs multi-query retrieval and dual LLM setups to fetch up to 500 relevant chunks per query.
Semantic Batching
Aggregates intermediate answers using hierarchical clustering to synthesize a single coherent final response.
Opinionated Summaries
Generates detailed markdown bullet-point summaries capturing the author's exact thought process.
Docker Web UI
Includes an optional Gradio-based web interface and Open-WebUI tool integration for simpler deployment.
Up and running in 3 steps
Our honest verdict
What's great
- Native handling of difficult formats like audio recordings and flashcards alongside standard documents
- Deeply configurable search pipeline scaling up to hundreds of retrieved documents per query
- Provides precise source citations pointing directly to the exact portion of the source material
- Extensible codebase usable as both a command-line tool and a Python library
Where it falls short
- Requires technical familiarity with CLI arguments and Python environment setups
- Running heavy queries across hundreds of documents consumes significant LLM API tokens and costs
- Requires your own LLM provider API keys or local model configuration to function
Pricing & license
Free under its open-source AGPL-3.0 licence. The only operational cost comes from running your own hardware or paying for external LLM API providers like OpenRouter or DeepSeek.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
What file types can wdoc process?
It supports over 15 file types including PDFs, online URLs, audio files, video lectures, EPUBs, and Anki flashcards.
Can I use local LLMs with wdoc?
Yes, it supports virtually any LLM provider including local models through LangChain and LiteLLM backends.
How does the summary feature work?
It splits text into chunks, passes them through an LLM for low-level detailed markdown summaries while maintaining context, and can recursively combine them for large books.
