Quickstart

pip install ragforge             # core (zero dependencies)
pip install ragforge[api]        # add HTTP API server
pip install ragforge[pdf]        # PDF support (pypdf)
pip install ragforge[pipeline]   # sentence-transformers embeddings + reranking
pip install ragforge[openai]     # OpenAI embeddings + LLM
pip install ragforge[anthropic]  # Anthropic LLM
pip install ragforge[docling]    # Docling backend (best for complex docs)
pip install ragforge[ui]         # local web dashboard
pip install ragforge[all]        # everything

ragforge info                                  # see registered components
ragforge knowledge build my-kb ./docs/         # build a knowledge base
ragforge query my-kb "How do refunds work?"    # query it
ragforge query my-kb "refunds?" --generate --llm ollama  # with grounded answer

import ragforge as rf

doc = rf.parse_file("notes.md")
chunks = rf.chunk_document(doc, strategy="structure")

for c in chunks:
    print(f"[{c.metadata.get('section')}] ~{c.token_count} tokens")

from ragforge.pipeline import build_knowledge_base, query_knowledge_base

build_knowledge_base(name="my-kb", sources=["./docs/"])

result = query_knowledge_base(knowledge="my-kb", question="Refund policy?")
for chunk in result["chunks"]:
    print(f"  [{chunk['score']:.3f}] {chunk['text'][:80]}")

from ragforge.pipeline import KnowledgeBase

kb = KnowledgeBase.build(
    name="my-kb",
    sources=["docs/", "faq.md"],
    embedder="default",
    chunk_strategy="structure",
)

for chunk, score in kb.query("How do refunds work?", top_k=5, mode="hybrid"):
    print(f"  [{score:.4f}] {chunk.text[:80]}...")

pip install ragforge[api]
ragforge serve
# Interactive docs at http://localhost:8000/docs

curl -X POST http://localhost:8000/query \
  -H "Content-Type: application/json" \
  -d '{
    "knowledge": "my-kb",
    "question":  "How do refunds work?",
    "generate":  true,
    "llm":       "ollama"
  }'

docker build -t ragforge .
docker run -p 8000:8000 ragforge