Show HN: I built a deep research tool for local file system
8 days ago
- #research-tool
- #markdown-report
- #local-resources
- DeepDoc is a tool for deep research on local resources, not the internet.
- It uses a research-style workflow to explore documents and generate markdown reports.
- Supports various file formats like PDF, DOCX, JPG, TXT.
- Text is extracted and split into page-wise chunks stored in a vector database.
- Generates content structure based on user queries and allows feedback for refinement.
- Research agents generate knowledge, create queries, and refine results for each section.
- Final report is compiled into a structured markdown format.
- Setup involves cloning the repository, creating a virtual environment, and installing dependencies.
- Requires API keys for Mistral, Tavily, and OpenAI to function.
- Docker and Docker Compose are needed to start services like Qdrant.
- Customizable via configuration.py for LLM and thread settings.
- Project is licensed under MIT License.