OpenLIT Enterprise: LLM Observability with RBAC, Audit Logs and Alerts

OpenLIT Enterprise is open source LLM observability on OpenTelemetry, plus RBAC, audit logs, per-feature API keys and alerts to Slack, PagerDuty and Datadog. Book a call with the founder: https://cal.com/aman.openlit/30min GitHub: https://github.com/openlit/openlit Enterprise docs: https://docs.openlit.io/latest/openlit/enterprise Site: https://openlit.io Someone changed a member's role on Tuesday at 5:48 PM. Who did it? The audit log knows: every privileged change is recorded, and nobody can edit or delete entries. OpenLIT Enterprise is the same Apache 2.0 OpenLIT, plus organisation-wide governance, compliance and alerting, unlocked per organisation with a signed license. What Enterprise adds: - Role-based access control (RBAC): custom role groups and direct user permissions - Per-feature API key access: restrict each key to only the features it needs - Audit logs: who did what, to which resource, and when - Alerts to Slack, email, webhook, Discord, PagerDuty, Opsgenie and Datadog the moment an API key, a Vault secret or someone's access changes - Datadog and New Relic data sources: read their traces, logs and metrics inside OpenLIT - GPUs page: fleet health, workloads, kernels, errors and cost The community edition stays free and Apache 2.0. Want Enterprise? Book a call with the founder: https://cal.com/aman.openlit/30min Chapters 0:00 Who changed this role at 5:48 PM? 0:07 Same OpenLIT, plus governance 0:14 Audit logs 0:24 Role-based access control 0:30 Restrict an API key per feature 0:36 Alerts to Slack, PagerDuty and more 0:44 Datadog and New Relic 0:49 GPU fleet health and cost 0:52 Community vs Enterprise 0:55 Book a call with the founder, star on GitHub 1:00 One question for you Every screen in this video is a real screenshot of the OpenLIT Enterprise build (v2.1.1), taken from the enterprise docs. Test and user emails are blurred. The 3D pillar transition is stylized. Music: Happy Beats & Business Moves Vol. 12 by Sascha Ende, ende.app, CC BY 4.0. Sound effects by Kenney, CC0. #OpenLIT #LLMObservability #OpenTelemetry #AuditLogs #RBAC Who changed what in your AI stack last week? Could you prove it? Tell us in the comments. #Shorts

All videosYouTube

Your agent called 5 tools and said Done. Which one failed? #Shorts

Your AI agent said "Done. The refund has been issued." It wasn't. The refund API was down, the tool swallowed the error and returned {}, and the agent told the customer it was done anyway. This is a real run: a LangGraph agent on a local Qwen 2.5 3B model (Ollama) with 5 tools. One line, openlit.init(), traced every LLM call, tool call and HTTP request into OpenLIT, so the empty tool result and the "Connection refused" error are right there in the trace. Star OpenLIT on GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Has your agent ever said "done" when it wasn't? Tell me what it skipped 👇 #Shorts #AIAgents #LangChain #OpenTelemetry #OpenLIT

All videosYouTube

What is OpenTelemetry? LLM Observability with OpenLIT #Shorts

Instrument once. Send your telemetry anywhere. That is OpenTelemetry, and OpenLIT is built natively on it. Docs: https://docs.openlit.io OpenTelemetry is the open, vendor-neutral CNCF standard for traces, metrics, and logs. Add one line, openlit.init(), and OpenLIT auto-instruments your LLMs, vector databases, agent frameworks, and GPUs, following the OpenTelemetry GenAI semantic conventions. Send the data to OpenLIT or to any OpenTelemetry backend, like Grafana or Jaeger. No vendor lock-in. Open source and self-hosted. GitHub: https://github.com/openlit/openlit Site: https://openlit.io Watch the full video: https://www.youtube.com/watch?v=foYnZK_ORDs Code and span cards in the video are illustrative. Screenshots come from docs.openlit.io. #OpenTelemetry #LLMObservability #OpenLIT #OpenSource #Shorts

All videosYouTube

AI Agent Said "Done" but Did Nothing? How to Debug It #Shorts

Traces docs: https://docs.openlit.io/latest/openlit/observability/telemetry/traces GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Your AI agent said "done." It did nothing. No error, no crash, and your logs say 200 OK. Agents fail silently: a wrong tool call, a loop, a lost context. To debug them, trace every step. Find the exact span where the run went wrong, then check tokens and cost. A loop shows up as a spike. OpenLIT traces LLM calls and agent/tool steps, and shows tokens and cost for each trace. It's open source, built on OpenTelemetry, and setup is one line: openlit.init(). What's the weirdest way your agent failed? Tell me in the comments 👇 #Shorts #AIAgents #OpenLIT #OpenTelemetry #LLMObservability #Debugging #AIEngineering #OpenSource

All videosYouTube

Traces in Tempo, memory in Mem0? Connect it all to OpenLIT #Shorts

Docs: https://docs.openlit.io/latest/openlit/connectors/overview GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Your traces, agent memory, and repos all live in different tools. OpenLIT Connectors plug into them from one place (Configuration → Connectors), scoped to your project and environment. • Data sources: keep ClickHouse, Grafana Tempo, Loki, Prometheus, or Jaeger, and route traces, logs, and metrics independently. • Memory: connect Claude, Mem0, or Zep to browse and search what your agents remember (capabilities vary by vendor). • Scanners: Trustabl scans GitHub repos for agent SDK, MCP, and policy findings. Which connector should we add next? Tell me in the comments. #Shorts #OpenLIT #OpenTelemetry #Observability #AIAgents #Mem0 #Grafana #MCP

All videosYouTube

What did Claude Code just cost you? Trace Cursor & Codex too #Shorts

Docs: https://docs.openlit.io/latest/openlit/coding-agents/overview GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Your coding agent just ran for 40 minutes. What did it actually do? OpenLIT traces Claude Code, Cursor, and Codex with one command (openlit coding install --vendor=all): no SDK, no code changes. Every session lands in the Coding Agents tab with tokens, cost, tool calls, code changed, acceptance %, commits, and pull requests. Self-hosted, built on OpenTelemetry. Which coding agent do you use? Tell me in the comments. #Shorts #ClaudeCode #Cursor #Codex #CodingAgents #OpenLIT #OpenTelemetry #AIAgents

All videosYouTube

Open-source LLM observability on OpenTelemetry | OpenLIT

GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Tired of proprietary LLM telemetry? OpenLIT is open-source AI engineering for LLM observability, evals, prompts, and cost — built on OpenTelemetry. Self-host free under Apache 2.0. #Shorts #OpenLIT #LLMObservability #OpenTelemetry #OpenSource #AIObservability #DevTools #LLM

All videosYouTube

LLM API keys + GPU metrics in OpenLIT

GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io API keys in env files, GPU metrics in another tool? OpenLIT Vault stores/rotates LLM keys; the GPU collector adds utilization, memory, temp, and power via OpenTelemetry. #Shorts #OpenLIT #Vault #GPUMonitoring #OpenTelemetry #LLMOps #OpenSource

All videosYouTube

Monitor AI agents end to end | OpenLIT

GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Your agent called five tools — which one failed? OpenLIT monitors AI agents with OpenTelemetry spans for every step, tool, and LLM call. #Shorts #OpenLIT #AIAgents #AgentObservability #OpenTelemetry #LLMOps #OpenSource

All videosYouTube

Compare LLMs side by side | OpenLIT OpenGround

GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Picking a model from a blog benchmark? OpenGround compares LLMs on your prompts — latency, cost, and quality — before you ship. #Shorts #OpenLIT #ModelComparison #LLM #AIEngineering #OpenSource #DevTools

All videosYouTube

Version prompts without shipping code | OpenLIT Prompt Hub

GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Prompts still hardcoded in the repo? OpenLIT Prompt Hub versions and deploys prompts without a code deploy — linked to traces and evals. #Shorts #OpenLIT #PromptEngineering #PromptManagement #LLMOps #OpenSource #AIEngineering

All videosYouTube

LLM-as-a-judge on live traces | OpenLIT

GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Shipped a prompt change and quality dropped? OpenLIT runs LLM evaluation on the same production traces — online scoring and CI. #Shorts #OpenLIT #LLMEvaluation #LLMasaJudge #AIQuality #OpenSource #DevTools

All videosYouTube

Trace LLMs with OpenTelemetry in 1 line | OpenLIT

GitHub: https://github.com/openlit/openlit Docs: https://docs.openlit.io Site: https://openlit.io Still guessing which LLM call burned your budget? OpenLIT adds OpenTelemetry LLM tracing with pip install openlit + openlit.init() — cost, latency, tokens, tools. #Shorts #OpenLIT #OpenTelemetry #LLMObservability #AIObservability #OpenSource #DevTools #LLM

All videosYouTube

Use arrow keys, mouse wheel, or swipe to move between Shorts. Press M to mute or unmute. Playback uses the official YouTube IFrame Player so views count on YouTube.