jonzo97/mchp-fpga-mcp
If you are the rightful owner of mchp-fpga-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.
The FPGA Document RAG Stack is a local-first retrieval-augmented generation system designed to manage and process Microchip PolarFire FPGA collateral efficiently.
FPGA Document RAG Stack
Scaffold for a local-first retrieval-augmented generation system focused on Microchip PolarFire FPGA collateral.
Components
incoming/– staging area for vendor PDFs before ingest.content/– normalized text, tables, figures, and derived artifacts.src/fpga_rag/– Python package housing ingestion workers, storage adapters, and MCP server.scripts/– CLI utilities for ingestion and maintenance.tests/– pytest-based checks for pipeline units.
Getting Started
1. Install mchp-mcp-core (Development Dependency)
This project depends on mchp-mcp-core for PDF extraction, embeddings, and vector storage.
# Install mchp-mcp-core in editable mode for development
cd ~/mchp-mcp-core
pip install -e .
Installing in editable mode (-e) means changes to mchp-mcp-core are immediately available without reinstalling.
Note: As of 2025-11-14, fpga_mcp uses mchp-mcp-core for enhanced PDF extraction including:
- Multi-strategy table extraction - 3-way consensus (pdfplumber + Camelot + PyMuPDF)
- Structure-aware chunking - Preserves section hierarchy
- Specification extraction - Extracts electrical parameters (voltage, current, timing)
- ~2x more content extracted compared to previous pdftotext-only approach
This migration maintains full backward compatibility with existing ingestion pipeline and ExtractionWorker API.
2. Install fpga_mcp
cd ~/fpga_mcp
python -m venv .venv
source .venv/bin/activate
pip install -e .[dev]
3. Ingest FPGA PDFs
# Populate incoming/ with FPGA PDFs
python scripts/ingest.py
4. Run the MCP Server
uvicorn fpga_rag.server.app:app --reload
Refer to SPEC.md for the full system design and roadmap.