LangChain built SmithDB, a new database to handle massive agent traces for LangSmith
Hacubu · x · 2026-10-13
LangChain built SmithDB, a purpose-built database for ingesting and serving massive agent trace volumes in LangSmith, arguing existing data infrastructure isn't optimized for this workload. The quoted tweet explains why: hundreds of nested spans per trace, huge prompts and tool outputs, runs staying open for hours, events arriving over time, plus full-text search and JSON filtering requirements. The team is reaching out to interested engineers.
More from coding & agent
- Sentry CEO zeeg: nobody has a secret AI coding workflow, everyone just iterates until it ships — zeeg · 2026-10-13
- LinkedIn replaces APM program with Associate Product Builders as 30% of PM jobs demand AI skills — aakashgupta · 2026-10-13
- Solo dev builds open-source multi-agent mesh where Claude, GPT and Gemini share one room — Radiant_Bug9778 · 2026-10-13
- Codex app turns paperback highlights into synced Kindle marks via OCR + computer use — JeremyNguyenPhD · 2026-10-13
- Professor Reveals Recipe Behind Cinematic Course Site: Blender + Codex + Planning + Iteration — CSProfKGD · 2026-10-13
- 1.7B laptops vs 21B IoT devices: the agent-concurrency gap nobody talks about — seanwbren · 2026-10-13