cyper-rag-server

# ─────────────────────────────────────────────────────────────────────────────
# MCP RAG Server — Python dependencies
# Install: pip install -r requirements.txt
# ─────────────────────────────────────────────────────────────────────────────

# ── Core MCP ──────────────────────────────────────────────────────────────────
mcp

# ── Vector store ──────────────────────────────────────────────────────────────
chromadb

# ── Embeddings + reranking ────────────────────────────────────────────────────
sentence-transformers          # embedding models + CrossEncoder reranker

# ── Document extraction ───────────────────────────────────────────────────────
pypdf                          # PDF
python-frontmatter             # Markdown YAML front matter
beautifulsoup4                 # HTML (graceful fallback if missing)
python-docx                    # DOCX (graceful fallback if missing)

# ── Chunking ─────────────────────────────────────────────────────────────────
nltk                           # sentence tokenization (sentence + semantic strategies)

# ── HTTP client ───────────────────────────────────────────────────────────────
httpx                          # async HTTP for web search + HyDE

# ── Web search providers ──────────────────────────────────────────────────────
duckduckgo-search              # DuckDuckGo (no API key needed)
# Uncomment as needed:
# brave-search                 # Brave (needs --brave-api-key)
# tavily-python                # Tavily (needs --tavily-api-key)
# (Serper uses httpx directly — no extra package needed)

# ── Optional: EPUB support ────────────────────────────────────────────────────
# ebooklib                     # EPUB extraction (graceful fallback if missing)

# ── Optional: GPU acceleration ────────────────────────────────────────────────
# torch                        # Install the CUDA build manually for GPU embeddings
# onnxruntime-gpu              # ONNX GPU inference (faster embedding on NVIDIA)