Ward Peeters
@wardpeet
Followers
1K
Following
1K
Media
15
Statuses
2K
OSS lead at @mastra_ai. Web Performance and building stuff is my JAM. Herder of 🐐🐐🐐. (he/him)
Antwerp, Belgium
Joined May 2016
RT @ShlomesLew: Some lessons learned while building out agent input processors for Mastra. 1. Every token matters.2….
mastra.ai
How we built a suite of out-of-the-box input processors and optimized them from 5000ms to under 500ms per request.
0
8
0
A big issue with nested agents in tools or workflows is the nested usage count. We now automatically sum it up for you.
we now have support for nested streaming in mastra. previously, if you have an agent as part of a tool, you would lose the nested stream (and the same thing was true of streaming within a workflow step & within a long tool step). but now, with nested streaming support, you can do.
0
0
2
LongMemEval has given us a lot of insights, check out this thread to find what!.
At @mastra_ai we systematically tested agent memory with LongMemEval and achieved 80% accuracy using RAG alone - beating published benchmarks at comparable latency. Here's what we learned:
0
0
4
Quality content.
Did I accidentally spend $8k on OpenAI credits in 3 days running memory benchmarks? Yes. Did I burn through 3.8 TRILLION tokens from a single laptop? Also yes. really @mastra_ai should be thanking me though. another team would've taken weeks to waste this much money.
0
0
0
You heard it, folks! @smthomas3 THE LEGEND!.
Watching @smthomas3 from @mastra_ai right now in the @recallnet hackathon @encodeclub live stream (recordings will be available). So so sooo good. He totally customized the demo to use OUR MCP tools, and other relevant tools?! Legend.
0
1
6
I'm developing a new streaming protocol for Mastra that will allow you to stream any type of content into the @mastra_ai output stream. It will also support nested token usage, you will be able to see the total number of tokens used, including those from tools or workflows.
2
5
21
RT @rachelnabors: Tomorrow I'm going to migrate my house hunting flow off of Zapier with @mastra_ai and @TryArcade live with every fan in t….
twitch.tv
Twitch is the world's leading video platform and community for gamers.
0
11
0
deploy your coding agent to mastra cloud. you can just deploy things
1
4
7
If I were you, I would join because of that impressive marketing page. 🚀.
every cloud has a silver lining. yours does too. welcome to mastra cloud.
0
1
6
RT @tylbar: 🧠 Big update at @mastra_ai: We're rolling out cross-conversation memory that persists for each user!. ✅ Semantic recall is live….
0
17
0
So true !!!.
LLMs forget everything. Vector DBs are complex. @mastra_ai + Turso lets you build AI agents that remember. Link in reply 👇
0
0
1
Not only do we ship features, we also do deep research and patch models so you don’t have to!.
My first blog post for @mastra_ai exploring how we improved MCP tool calling success rate by 12% to 97% across various models. If you’ve ever tried running third-party MCP tools in your agent and they error, silently fail, or hallucinate, this ones for you
0
0
9
RT @abhiaiyer: We’re gonna do some impromptu meetups this weekend at coffee shops or bars. I’ll post where we’ll be come meet us!!.
0
2
0