Joined September 2017
- Tweets 583
- Following 3
- Followers 3,593
- Likes 866
44 Photos and videos
vespa.ai retweeted
Jun 5
Perplexity taking this to its logical conclusion and then some, but the core idea here is important and not yet widely understood: Agents are able to tailor queries in detail to what they are trying to do.
This is makes a huge difference in cost and quality. For example:
- When researching case law, search for the names of those involved near each other in text.
- When seeking an overview of a broader topic, do a pure vector search prioritizing authoritative sources.
- When making timeline of some historic development, limit by year range and group by month.
And so on. Models already know YQL, all you need is to tell them what they have to work with.
We’re moving away from search as a web fetch tool call to search as codegen to be future proof in a world where code execution inside agent harnesses is the way to do almost all of our knowledge work.
Doing this lets you compose multi-step primitives far more naturally and be much more adaptable to changes made to the agent harness, as well as benefit from improvements in coding capabilities that are guaranteed to come from the next generation of frontier models.
3
2
9
717
vespa.ai retweeted
May 29
Vespa skills for coding agents: Claude Code and competitors are very useful for working on Vespa applications. Everything is always completely specified in application packages, which are in their favorite form factor.
Still, they do better with a good collection of skills. Skills are a dime a dozen though, what you want are skills that are proven by evals to carry more than their own weight.
We're maintaining a collection of optimized and evaluated skills you can grab here: github.com/vespaai-playgroun…
2
6
210
vespa.ai retweeted
Jun 1
Vespa is now available as a Document Store in Haystack. Use VespaDocumentStore for hybrid and semantic search with a powerful, production-ready engine, and pair it with VespaEmbeddingRetriever to index and retrieve documents directly in your pipelines. Metadata filtering included.
@vespaengine excels at large-scale information retrieval with advanced features like real-time indexing, multi-modal search, and distributed document management - ideal for applications that demand both speed and sophistication.
🐍 pip install vespa-haystack
🔗 haystack.deepset.ai/integrat…
1
9
341
vespa.ai retweeted
May 29
Detailed metric dashboards: Some people like to dig deep, and using a cloud solution shouldn't prevent them. We've made all the dashboards user by Vespa operators available in the Vespa Cloud console out of the box.
1
1
252
May 29
Vespa isn't only powering American AI, now it's also @MistralAI's retrieval solution
1
2
6
491
vespa.ai retweeted
May 29
bm25 is nice and all, but you won't believe how easy it is to improve upon it with and how much more you can squeeze from lexical features in @vespaengine
3
5
26
1,960
May 28
The May Vespa.ai Newsletter is out!
This month we’re announcing updates focused on retrieval quality, ranking flexibility, and developer productivity (agents: try out the skills and let us know).
- Vespa Cloud: Detailed metric dashboards
- Vespa Cloud: Index backup
- Vespa Cloud: Fine-grained maintenance controls
- Vespa Cloud: Custom resource tags
- Vespa skills for agents
- @VoyageAI , @OpenAI , and @MistralAI embedders
- A new query operator for text matching
- Cluster-size independent config of relevance effort
- Boolean array fields
- Match specific array elements
- In-memory document ids
- Search group pinning
- Near matching aware ranking
- Detect ignored write operations
- Accessing the max first phase score in re-ranking
- Geo distance in grouping
Read it here: blog.vespa.ai/vespa-newslett…
8
2
9
838
May 29
Cluster-size independent config of relevance effort: Vespa has several parameters that let you influence how much effort should be spent on relevance quality for a query, such as ANN and WeakAnd targetHits, and second-phase rerank-coint.
These have been set in terms of amounts per node, which is impractical when the number of nodes change, and doesn't work well with autoscaling.
Now we have added variants prefixed by "total" which sets the effort over the whole cluster instead.
1
282
May 29
One skill that will particularly accelerate the move to modern AI search is our ElasticSearch migration skill: github.com/vespaai-playgroun…
We're already getting feedback saying this enabled people to migrate complex ES applications to Vespa in less than a day.
7
221
May 29
Fine-grained maintenance controls: One consequence of Mythos-like capabilities soon being broadly available is that we'll need to get used to upgrade all parts of the software stack much more frequently. You're already working on a plan for this, right?
Those running on Vespa Cloud are already in great shape here since platform and OS upgrades just happen automatically. We'll need to make OS upgrades more frequent though, which means they'll be more intrusive. That's why we have added controls that allows you to specify at what time windows they can happen.
1
58
May 29
Index backup: Since indexes on Vespa are realtime, backing them up is not straightforward. However, this may make disaster recovery much faster. You can now turn on automated index backups on Vespa Cloud
1
184
vespa.ai retweeted
Radu Gheorghe is a software engineer at @vespaengine. He joins @seanfalconer to discuss vector search limits, tensor based retrieval, ranking tradeoffs, and where AI search is headed next.
@radu0gheorghe
softwareengineeringdaily.com…
3
7
600
vespa.ai retweeted
Apr 29
Amazing blog post by @vinted team on scaling personalized search autocomplete to 125M suggestions, 4.7k QPS and 31ms p99 latency 🤯
3
2
9
317
vespa.ai retweeted
Apr 27
I really look forward to our first @vespaengine full-day community event in London, September 10! Don't forget to submit a talk, or just hang out with other practitioners. Secure your Early-bird ticket now!
This is an event to learn from each other and get inspiration for how to build state-of-the-art applications: content.vespa.ai/vespa-live
3
3
187
Apr 13
Vinted explains how to get stable p99 latency of 3ms when using Vespa as a feature store. 👇
2
7
656