llamaindex-omega 0.3.0

LlamaIndex integration with OmegaEngine governance - RAG safety and compliance

llamaindex-omega

LlamaIndex Integration for OmegaEngine

Documentation • Playground • Issues

---

✨ Features

• 🔗 Query Engine Wrapper — Add governance to any LlamaIndex query
• 📊 Retrieval Guard — Validate retrieved context before synthesis
• 🛡️ Response Validator — Check LLM outputs against policies
• 📈 Observability — Full audit trail for RAG pipelines
• ⚡ Async Support — Native async/await compatibility

---

📦 Installation

pip install llamaindex-omega

---

🚀 Quick Start

Governed Query Engine

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
from llamaindex_omega import GovernedQueryEngine

# Load your documents
documents = SimpleDirectoryReader("data").load_data()
index = VectorStoreIndex.from_documents(documents)

# Wrap with OmegaEngine governance
query_engine = GovernedQueryEngine(
    index.as_query_engine(),
    api_key="your-omega-api-key",
    policy_id="rag_policy",
)

# All queries are now governed
response = query_engine.query("What are the quarterly results?")
print(response)

Response Callback

from llamaindex_omega import OmegaCallbackHandler

callback = OmegaCallbackHandler(
    api_key="your-omega-api-key",
    log_retrievals=True,
    log_responses=True,
)

# Attach to your service context
from llama_index.core import Settings
Settings.callback_manager.add_handler(callback)

---

🛡️ Retrieval Guard

Validate retrieved context before it's sent to the LLM:

from llamaindex_omega import GovernedRetriever

guard = GovernedRetriever(
    api_key="your-omega-api-key",
    block_pii=True,
    block_sensitive_topics=["financial", "medical"],
)

# Wrap your retriever
safe_retriever = guard.wrap(index.as_retriever())

# Retrieved nodes are validated before synthesis
nodes = safe_retriever.retrieve("Tell me about the patient")

---

📊 Audit Trail

Every query is logged with:

• ✅ Original query
• ✅ Retrieved context (optionally)
• ✅ Generated response
• ✅ Policy evaluation result
• ✅ Risk score
• ✅ Latency metrics

# Access audit data
print(response.metadata["omega_audit_id"])
print(response.metadata["omega_decision"])

---

🔌 Works With All LlamaIndex Features

• ✅ VectorStoreIndex
• ✅ KnowledgeGraphIndex
• ✅ DocumentSummaryIndex
• ✅ Chat Engine
• ✅ Agents
• ✅ SubQuestionQueryEngine

---

📄 License

Licensed under the Apache License 2.0.

---

Built with ❤️ by the OmegaEngine team