Software developer & AWS consultant.
Hello. I'm Martin, a software developer and entrepreneur based in Manchester, England.
I'm a co-founder of Si Novi, where we build cloud-native applications and consult on AWS architecture. I also build tools for DynamoDB - including Dynoxide and Nubo.
Latest posts
DynamoDB changed how it bills a vector read
On 9th October a DynamoDB read of an item with a vector in it cost 2 read units. On 22nd October the same read cost 1. I can't find anything from AWS about it, and the 400KB limit still counts the vector the old way.
Read article
Dynoxide 1.4.0: clean shutdowns and JSON Lines import
Dynoxide 1.4.0 is a smaller one. JSON Lines import thanks to Sam Chung, a clean way to stop an embedded server, and updates and scans that behave more like DynamoDB.
Read articleDynoxide 1.3.0: security fixes and fussier pagination
Dynoxide 1.3.0 ended up a bit bigger than I'd planned. A few of the DynamoDB quirks I worked through, two security fixes, and quieter test output thanks to an external contribution.
Read articleParity Suite 3.6 and a month of emulator fixes
Floci went from 337 failing tests to one in five weeks, and LocalStack and MiniStack both moved up a grade. Meanwhile AWS changed how transactions validate halfway through measuring 3.6.0, so that one never reached the board.
Read article
DynamoDB vector search without embeddings
Serverless Advocate #85 had Lee Harding making the case that DynamoDB vector indexes are a nearest-neighbour tool, not an embeddings one, so I tried the geography case. A single geohash cell misses the nearest shop entirely. Read nine and it does not, which is the update at the end.
Read article
Running DynamoDB vector search locally
Dynoxide has done DynamoDB vector search since 1.0.0. What I measured before building it, including the bit where AWS says the index is ready about fifteen minutes before it actually is, and the end of a rollout I have been watching since June.
Read article