Pinned
Stop manually triaging agent failures.
Let LangSmith Engine fix it.
00:00
LangChain
11.1K posts
LangChain
@LangChain
Joined November 2022
- .@mondaydotcom had one agent trying to handle 200+ tools. Context pollution everywhere. The LLM was confused, costs rising, and it still wasn't working. @omribruchim on rebuilding Sidekick with Deep Agents: youtu.be/c2fLLS7np3Y
00:00 - Before we shipped LangSmith LLM Gateway, we rolled it out internally. ✅We don’t have to wait until the end of the month to understand spend ✅We have been able to set budgets set by org, workspace, user, or API key ✅Our teams can flexibly use coding agents without creating
- Rubrics for Deep Agents, programmatic subagents, and everything new in LangSmith ⤵️
- Fine-tuned models match frontier performance In our research with @FireworksAI_HQ, a fine-tuned @Alibaba_Qwen outperformed all model sizes. They’re also cheaper to run at scale 10-100x depending on trace volume and model choice As trace volumes grow, so will cost-savings on
We partnered with @FireworksAI_HQ to fine-tune an @Alibaba_Qwen judge model to detect “Perceived Error” from user interactions. See our findings ⤵️ - LangChain repostedHow do you construct, compact, and query a full-text search index when all your durable data lives in object storage? And where every GET costs 50–100ms? @ankush_gola11 + team on how we built and run SmithDB's inverted index in production.
- LangChain reposted
Technical blog on how we built SmithDB (our database purpose built for agent traces)How do you construct, compact, and query a full-text search index when all your durable data lives in object storage? And where every GET costs 50–100ms? @ankush_gola11 + team on how we built and run SmithDB's inverted index in production. langchain.com/blog/full-text… - LangChain reposted
if you get 1% better every day then you're 37x better in a year (from @JamesClear's atomic habits book) my theory is you can do this for agents if you build a good continual learning loop which is what i'm working on at @LangChain with deepagents
- LangChain reposted
Stoked to be speaking at @aiDotEngineer next week. Come catch my session at 11.40am PT. Will be doing a deep dive into agent memory - 'sleep-time compute', 'dreaming' and the critical role traces play in the entire processIf you're headed to @aiDotEngineer World's Fair next week, don't miss @jakebroekhuizen and @Vtrivedy10's stage sessions.
- If you're headed to @aiDotEngineer World's Fair next week, don't miss @jakebroekhuizen and @Vtrivedy10's stage sessions.
excited to be speaking at @aiDotEngineer World Fair next week on Improving Agents, Continual Learning, and why we think a large part of it is...Data Mining! Trace Mining is how we understand agent behavior at scale so we can: - build evals/environments to hill-climb - gather - LangChain reposted
Supporting sub-second full-text search on object storage is hard, especially when dealing with large agent observability workloads. Here is part 2 of our blog post that outlines how we accomplished this in SmithDB!How do you construct, compact, and query a full-text search index when all your durable data lives in object storage? And where every GET costs 50–100ms? @ankush_gola11 + team on how we built and run SmithDB's inverted index in production. langchain.com/blog/full-text… - Time to cook.Agents are easy to demo locally. The hard part is shipping them inside a real app. We published a deployment cookbook for @LangChain agents: full-stack examples with streaming UI, subagents, thread history, and production persistence notes across common JS frameworks 🚀 🧵👇
- How do you construct, compact, and query a full-text search index when all your durable data lives in object storage? And where every GET costs 50–100ms? @ankush_gola11 + team on how we built and run SmithDB's inverted index in production.Missed Part 1? Here’s how we built a custom inverted index from scratch for SmithDB.
