Posts
All the articles I've posted.
-
Qwen3.5 Cross-Platform Benchmark: 4 Hardware Targets × 5 Engines Compared
Updated:Apples-to-apples Qwen3.5 numbers across Mac M5 Max, RTX 3090×2, DGX Spark, and Ryzen AI MAX 395+. Cold prefill, cache disabled, randomized run order.
-
Qwen3.5 Local Inference Benchmark Results: 4 Machines × 5 Engines
Updated:Generation and prefill throughput for Qwen3.5 (9B, 27B, 35B-A3B MoE, 122B-A10B MoE) on M5 Max, RTX 3090×2, DGX Spark GB10, and Ryzen AI MAX 395 — measured with llama.cpp, MLX, Ollama, vLLM, and Lemonade.
-
Building a GraphRAG Pipeline — From Vector Search to Graph Expansion
Solve multi-hop questions that plain vector RAG can't answer. Vectorize graph nodes with from_existing_graph in one line, auto-convert natural language to Cypher with CypherQAChain.
-
Mastering Vector Search with langchain-age — Hybrid Search, MMR, and Metadata Filtering
Why Hybrid Search matters for pgvector, when to use each strategy, and real recall benchmarks. Includes HNSW vs IVFFlat selection criteria and MongoDB-style metadata filtering.
-
Full AI Agent Stack on One PostgreSQL — LangGraph + langchain-age
Can you replace Neo4j+Redis+Pinecone with just PostgreSQL for an AI Agent? A real architecture that unifies graph, vectors, checkpoints, and long-term memory in one database.
-
GraphRAG with Just PostgreSQL — No Neo4j Required
Build a production-ready Graph + Vector RAG pipeline using only PostgreSQL with Apache AGE and pgvector. One database, one connection string, zero licensing fees.
-
Neo4j vs Apache AGE Benchmark — Same Cypher, Same Data, Different Results
A fair benchmark running identical Cypher queries on Neo4j 5 and Apache AGE 1.7.0. Eight tests covering point lookups, 1–6 hop traversals, CREATE operations, and schema introspection.