The latest articles on DEV Community by Philip Solobay (@maximus242). https://dev.to/maximus242 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3709152%2Ffd1a0f1a-3e29-4e1a-a2fa-c5f033125d6c.jpeg https://dev.to/maximus242 en Philip Solobay Tue, 13 Jan 2026 13:55:39 +0000 https://dev.to/maximus242/ragguard-filter-during-vector-search-not-after-retrieval-3ni0 https://dev.to/maximus242/ragguard-filter-during-vector-search-not-after-retrieval-3ni0 <p>If you're building a RAG application with document-level permissions, you've probably implemented something like this:</p> <ol> <li>User makes a query</li> <li>Retrieve top-k documents from vector DB</li> <li>Filter out documents the user shouldn't see</li> <li>Send remaining docs to LLM</li> </ol> <p><strong>The problem?</strong> By step 3, unauthorized documents have already been retrieved. They've hit your retrieval layer, been processed, and potentially logged.</p> <h2> Enter RAGGuard </h2> <p>I built <a href="https://github.com/maximus242/ragguard" rel="noopener noreferrer">RAGGuard</a> to fix this. Instead of post-retrieval filtering, it translates permission policies into native vector database filters. Unauthorized documents are never retrieved in the first place.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight python"><code><span class="kn">from</span> <span class="n">ragguard.langchain</span> <span class="kn">import</span> <span class="n">SecureRetriever</span> <span class="n">retriever</span> <span class="o">=</span> <span class="nc">SecureRetriever</span><span class="p">(</span> <span class="n">base_retriever</span><span class="o">=</span><span class="n">your_retriever</span><span class="p">,</span> <span class="n">policy</span><span class="o">=</span><span class="n">your_policy</span> <span class="p">)</span> <span class="c1"># Filtered at the DB level - zero unauthorized exposure </span><span class="n">docs</span> <span class="o">=</span> <span class="n">retriever</span><span class="p">.</span><span class="nf">get_relevant_documents</span><span class="p">(</span><span class="n">query</span><span class="p">,</span> <span class="n">user_context</span><span class="o">=</span><span class="n">user</span><span class="p">)</span> </code></pre> </div> <h2> What it supports </h2> <ul> <li> <strong>14 Vector DBs:</strong> Qdrant, ChromaDB, Pinecone, pgvector, Weaviate, Milvus, and more</li> <li> <strong>Any Auth:</strong> OPA, Cerbos, OpenFGA, or custom RBAC</li> <li> <strong>Frameworks:</strong> LangChain, LlamaIndex, LangGraph</li> </ul> <h2> Why it matters </h2> <ul> <li>Compliance (HIPAA, SOC2, GDPR)</li> <li>Multi-tenant isolation</li> <li>Blocks 19/19 tested attack patterns</li> </ul> <h2> Get started </h2> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>pip <span class="nb">install </span>ragguard </code></pre> </div> <ul> <li>GitHub: <a href="https://github.com/maximus242/ragguard" rel="noopener noreferrer">https://github.com/maximus242/ragguard</a> </li> <li>PyPI: <a href="https://pypi.org/project/ragguard/" rel="noopener noreferrer">https://pypi.org/project/ragguard/</a> </li> </ul> <p>Open source (Apache 2.0). Feedback welcome!</p> python ai security opensource