Open source AI engineering platform. Now part of @clickhousedb. We're hiring: langfuse.com/join-us
github.com/langfuse/langfuse
- Tweets 1,004
- Following 679
- Followers 6,405
- Likes 3,928
Pinned Tweet
@marcklingen sat down to demo all langfuse features in a single video.
Agent observability, monitoring, evals, prompt management, and what is special about Langfuse.
Great place to start if you are new, or catch up on latest changes.
langfuse.com/watch-demo
4
8
23
5,386
Great conversation featuring @lotte_verheyden on production evals, from online vs. offline evaluation to sampling costs and better LLM judges 👇
Evals are the most important part of building AI systems, but there’s not a lot out there about building them well.
So I hosted a fireside chat on evals with Liam Bush (@LangChain), @lotte_verheyden (@langfuse), Braden Holstege (@mercor), @neutralino1 (@CoreWeave), and @soumyamohan (Galileo) to walk me through the tricks of the trade - how they actually work in production, where teams get them wrong, and where they're headed over the next 12 months.
We get into:
-online vs offline evals
-why running evals can cost you more than running the actual agent
-what nobody can see after the agent makes a tool call
-why understanding accounting rules might be harder than AGI
-and much, much, more…
Some of my favorite parts and link to full episode in the comments:
1
8
838
Conversation signals are the richest source for investigating how your agent performs in production
👉Catch frustrated users, follow-up questions, corrections from your users, requests to be handed off to a human, and more now with a single, Jev-based eval pass
👉Save costs over LLM-only approaches, use LLMs where more reasoning is necessary
2
23
2,599
langfuse.com retweeted
You can also access jev via @OpenRouter now on @langfuse to power your evals
Jev by @typesafeai is now on OpenRouter, in beta.
Jev is a System One model. Instead of generating text, it takes your app's state plus a typed question and returns a typed decision with a probability attached. There is no JSON prompting, parsing layer, and nothing to validate against.
5
9
975
langfuse.com retweeted
with Jev
it is less
eval -> eval -> eval
and more
-> eval
-> eval
-> eval
2
3
32
8,398
Jev-powered evals are now available in Langfuse
👉️ Pay 40-400x less than with frontier models
👉️ Run Jev (by @typesafeai) on all production traces without sampling
👉️ Check multiple criteria at minimal extra cost
langfuse.com/blog/2026-09-22…
11
21
86
14,330
We now made it natively available in Langfuse Evals/Monitoring. Set up Jev model connection, and define "decision model" evaluators.
Langfuse then executes these evals on all new agent traces.
1
1
252
Learn more about Jev in Langfuse Evals on our blog
langfuse.com/blog/2026-09-22…
209
langfuse.com retweeted
organizing a chill 5k run with Outrun and our friends at @langfuse before AI Engineer Paris next week.
> Thursday, September 24, 7am
> Harvest (11e)
> Breakfast and coffee after (on us 👀)
Sign up here:
luma.com/langfuse-umkb
@saishrane_
3
6
318
Jev from @typesafeai cannot write sentences but is great at deciding and classifying
This makes it a direct alternative to LLM-as-a-judge setups.
Jev returns a choice between predefined categories, a score, or a true/false (noul) decision. And it will indicate how certain it is about it.
Learn how Jev can help you in your core LLM pipeline and especially with evals 👇
6
9
973
langfuse.com retweeted
Props to @langfuse for their excellent docs and learning materials! 👏
If you’re trying to wrap your head around concepts like LLM observability, monitoring, and evaluations, these resources are well worth checking out:
🎓 Academy: langfuse.com/academy
🛠️ Workshop: langfuse.com/workshop
📚 Guides: langfuse.com/guides
1
5
14
337
first-party langfuse observability for @pidotdev.
one trace per prompt, all turns of a session under one session id, subagent processes nested under the turn that spawned them.
shoutout to @mmilanagm for this ship!
langfuse.com/integrations/de…
17
2
38
2,110
langfuse.com retweeted
New to Langfuse?
Try langfuse.com/workshop to explore all capabilities based on a funny example application
It is really good now thanks to @annabellschfr!
1
6
721
langfuse.com retweeted
Singapore, our next meetup is headlined by Max Deichmann, Co-founder & CTO of Langfuse 🎤
Building AI Products
🗓️ Friday, Sept 25 · 6:00 PM · AWS Office | Singapore
Langfuse is the open-source LLM engineering platform used by thousands of teams to trace, evaluate, and debug AI applications — and it runs on ClickHouse under the hood.
Max's talk, "Trace It, Store It, Ship It," goes past the demo stage into what production actually demands.
Seats are limited and registration requires approval: buff.ly/zLpT5ZQ
Want to speak at a future APJ meetup? buff.ly/ak4tJO8
#ClickHouse #Langfuse #Singapore #LLM #AIEngineering #Observability
2
7
1,270
Why does @tryramp use Langfuse?
- API-first / hackable / extendable -> they built Ramp Reflect on top of Langfuse
- Built on a mature database (@ClickHouseDB)
- OSS, they send PRs that get merged
- Based on OpenTelemetry, works with any harness/framework
Yesterday, the @tryramp team shared the story of how/why they use @langfuse at OpenHouse New York.
When creating software gets cheap, collaborating with tasteful teams at the frontier is what makes Langfuse great.
7
2
22
2,069
langfuse.com retweeted
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
20,621
167,423
802,958
173,964,604
Late night podcast usage made the Rest team realize that 70M Americans suffer from symptoms of insomnia.
That's why they built Rest. A sleep coach built on the principles of CBT-I. Learn how they use Langfuse to cut the coach's memory issues in half.
langfuse.com/users/evolve
7
3
22
1,190
llm-as-a-judge evaluators now support multi-modal inputs. score the images, audio, video, PDFs, and text files captured in your observations.
prompts can be multi-message too: system for the criteria, user for the content.
langfuse.com/changelog/2026-…
11
10
56
8,368