Standout question: "Can you catch issues in CI/CD before production?"
Yes! Observability isn't just for production. Set it up early to monitor AI agent behavior in CI/CD and catch issues before deployment.
Full recording: www.youtube.com/watch?v=WWEv...
Thank you to everyone who joined our AI agent debugging webinar with @inkeep!
What we covered:
- AI agent architecture & observability challenges
- Live tracing demo: agent execution, tool calls, token usage
- Real-time debugging with OpenTelemetry
- Continuous quality monitoring
They'll show how to trace agent execution, tool calls, latency spikes, and token usage using OpenTelemetry instrumentation.
If you're building AI agents or working with agent frameworks, this one's worth your time.
Register here: luma.com/f2t9hnia?tk=...
This Thursday (Jan 29, 11:30 AM PT), we're doing a live session with Inkeep on observing AI agents in production.
Engineers from our teams will walk through how Inkeep built observability into their agent framework using SigNoz.
We're a Gold Sponsor at Open Source North! 🎉
Our team is excited to connect with the open source community there.
05.28.2026 | St. Paul, MN
See you at the event!
8/ Panel re-renders no longer triggered by interaction
9/ Query builder improvements: limit parameter handling, metric type respect, formula case-insensitivity
10/ Endpoint stats, span ordering, and light mode contrast improvements
& more bugs squashed
signoz.io/changelog/
4/ Public dashboard APIs now available in OpenAPI specification
5/ Log pipelines no longer fail to save on first attempt
6/ Click-to-copy functionality restored for log body fields
7/ Alerts created from dashboards now open in new tab
The release also includes:
1/ Tooltips in quick filters now appear instantly and placement updated to prevent flickering
2/ Checkbox filter state now triggers query reload when toggling filters
3/ Free text search now works correctly with quick filters
v0.106.0
Keyboard Shortcuts Enhancement:
Navigate faster with an enhanced Command Palette (Cmd/Ctrl+K) supporting multi-key combinations like Shift+M+E for Metrics Explorer or Shift+T+F for Trace Funnels. Long-press Shift to display a visual overlay of popular shortcuts.
Click on a span, see p78 next to it. Expand to see p50, p90, p99 with durations. Can add filters to narrow the comparison to a specific deployment or endpoint.
Helps you figure out if a span is an outlier without leaving the trace view.
We kept seeing the same pattern with users debugging traces in SigNoz.
They'd find a span that took 2 seconds, then open another tab to query percentiles and figure out if that duration is actually slow or just normal for that operation.
So we added the percentile inline.
Thank you to everyone who's contributed code, filed issues, helped other developers, and chose to build with SigNoz.
2026, here we come ⭐
We crossed 25,000 GitHub stars!
Q4 2025 was huge:
1/ Shipped our 100th GitHub release
2/ Launched Olly (our mascot) at KubeCon North America
3/ Hit 500 customers
4/ Crossed 25,000 GitHub stars to close out the year
4/ Corrected filtering behavior in External APIs view
Full changelog: signoz.io/changelog/
The release also includes:
1/ Fixed column display issues in Logs and Traces explorers to prevent cross-contamination
2/ Improved CMD/CTRL+K command palette with better styling and keyboard navigation
3/ Fixed input overflow in metrics custom functions with dynamic resizing
SigNoz v0.105.0
SigNoz now auto-generates OpenAPI documentation for its APIs, making programmatic integration easier. The specification covers authentication, users, organizations, sessions, and preferences, is validated in CI.
What gets automatically instrumented:
- Sub-agent transfers
- Tool executions
- LLM calls (model, tokens, latency)
- Multi-step workflows
Developers building with Inkeep get observability out-of-the-box!
1. Trace API → query traces programmatically
2. ClickHouse → fast queries over high-cardinality data
3. OpenTelemetry-native → no proprietary instrumentation
4. Local + production parity → same debugging experience everywhere
AI agents are different beasts to monitor 🧵
@inkeep builds a framework for AI agents. Here's how they integrated @SigNozHQ to give their users automatic observability:
1/ Structure spans for login, token validation, permission checks
2/ Add attributes to filter by user or failure reason
3/ Propagate context across services
4/ Patterns for OAuth, JWT, sessions
When auth breaks, see exactly where and why.
Authentication failures are hard to debug in distributed systems.
So, we wrote a guide on modeling auth flow with OpenTelemetry: newsletter.signoz.io/p/eli5-auth-...
Also includes:
1/ Copy JSON nodes from log details
2/ Better error message display
3/ Cleaner traces explorer columns
Full changelog: signoz.io/changelog/
Changelog v0.104.0
Public Dashboard Sharing:
Share dashboards publicly via link without authentication. Set default time ranges and control viewer access for easy sharing with stakeholders.
Includes:
1/ IdP-initiated SAML authentication
2/ Fixed query editor cursor jumping
3/ Resolved Prometheus rule deadlock
4/ Live log filters persist on reconnect
5/ Regex filters work with $ in column names
6/ Retention settings apply correctly
7/ Air-gapped environment support
Changelog v0.103.1
Y-Axis Unit Formatting in Logs and Traces Explorer:
Custom Y-axis units (bytes, ms, %) now available in explorer views, making charts instantly readable without mental conversions.
One platform. Context stays intact as you move between metrics, traces, and logs.
www.youtube.com/watch?v=7MCi...
This is what full-stack observability should look like.
46% order success rate → click the metric → jump to traces → 502 error from external API → check logs → root cause found in 4 minutes!
Our CEO @pranay01 is speaking at #KubeCon today at 4:45 PM!
How do you trace something that's non-deterministic by design? Pranay's got 5 minutes to show you.
📍 Nov 12 | 4:45 PM EST | Building C, Level 3, Georgia Ballroom 2
