Skip to content
Veristamp
Type to search the engineering journal.Full-text index · ⌘K to open
Topic

#AI-Agents

6 articles on this topic.

An elegant architectural diagram showing QQL query transformation through JWT validation, rate limiting, and AST-level filter injection before executing on Qdrant.
LatestQdrantSecurityEngineering

QQL Gateway: Zero-Trust Multi-Tenancy and AST Filter Injection for Qdrant

Vector databases ignore permissions, risking RAG leaks. Compare RAG post-filtering with Zanzibar-style pre-filtering, and learn how the QQL Gateway injects zero-trust AST filters before Qdrant queries.

13 min readRead
A side-by-side comparison diagram showing a locked, rigid hardcoded pipeline versus a dynamic agent loop querying Qdrant through the QQL gateway.
QdrantSecurityAI Agents

Secure Runtime Autonomy: QQL as the Dynamic Query Interface for Recursive Language Models

Hardcoding retrieval limits Recursive Language Models (RLMs). Discover how QQL acts as a secure, dynamic query layer via compile-time AST injection.

13 min readRead
Terminal-driven retrieval operations pipeline with QQL scripts, JSON artifacts, CI checks, and vector search diagnostics
RAGEngineeringAI Agents

qql-go and QQL Retrieval Operations for Qdrant

Vector search breaks quietly. Learn how qql-go turns QQL into a single-binary CLI for Qdrant retrieval operations, CI checks, and debug runbooks.

12 min readRead
Schema-first concept harvesting: documents to entities, relations, and an LLM wiki graph
AI AgentsKnowledge GraphsInformation Extraction

Schema-First Extraction for LLM Wikis with GLiNER2

Most RAG stacks fail at ingestion. Learn how schema-first extraction and GLiNER2 compile messy documents into a persistent LLM wiki and knowledge graph.

20 min readRead
Diagram showing an Oracle Always Free VPS secured with Tailscale running a PicoClaw AI agent
AI-AgentsSelf-Hosted AIOracle Cloud

Oracle Cloud Always Free Telegram Bot + AI Agent

Build an Oracle Cloud Always Free Telegram bot and private AI agent. Learn how to secure your VPS with Tailscale, PicoClaw, and UFW for zero public exposure.

12 min readRead
A contained sandbox with glowing code and policy gates, symbolizing governed execution
AI-AgentsMCPcode-mode

Governed Code Mode: From Tool-Calling to Zero-Trust Execution

Cloudflare, Anthropic, Google, and others are converging on a new pattern: let agents write code, but only inside a governed, zero-trust runtime.

10 min readRead
All posts