# ingero

> eBPF-based GPU causal observability agent

- **URL**: https://www.freshcrate.ai/projects/ingero
- **Author**: ingero-io
- **Category**: MCP Servers
- **Latest version**: `v0.19.0` (2026-05-24)
- **License**: Apache-2.0
- **Source**: https://github.com/ingero-io/ingero
- **Homepage**: https://ingero.io
- **Language**: Go
- **GitHub**: 54 stars, 4 forks
- **Registry**: github (`ingero-io/ingero`)
- **Tags**: `causal-tracing`, `cuda`, `cuda-graphs`, `ebpf`, `go`, `gpu`, `gpu-monitoring`, `gpu-observability`, `incident-response`, `model-context-protocol`

## Description

eBPF-based GPU causal observability agent

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v0.19.0` | 2026-05-24 | High | ## Changelog ### Documentation * c7bd321878cc110387a21db7d035ffcf2a2d7e21 docs: per-request summary entries in commands + detections + vLLM README ### Other * 6c5974d9a70b4a6c5e9a4da02599971261972bef cli + emitter: by-request flag-combo guards and emitter scope defaults * 7c27f8d130839438ce1335949e04b69c41304174 cli: Go-quote label values outside the safe charset in formatLabels * 8e998f80c28059cd985afcb98035ebdef010c55d cli: add explain --by-request flag with honesty caveats * 5955d9ed1fd68baa0 |
| `v0.17.0` | 2026-05-18 | High | ## Changelog ### Documentation * dbdbdbbd56c11032db919e1ec43ff209d403f385 docs: add v0.17.0 changelog entry * 0f3a35b7e5574593951d60a2f853a118e0e49163 docs: add vLLM integration example (#61) * e3bbbd7d319f30c09a434195ef46d622756a8d8d docs: backfill the 0.16.0 changelog entry * 9e88d3690d9007162eb10659c5eb9ae419790813 docs: cross-link the Fleet repo from agent docs * f600eab3ebb79514ee87cd9b8366b1d197ae0bdf docs: keep README to current behavior ### Other * 331758cdfe088845a565397237142c3e6352c97 |
| `v0.16.0` | 2026-05-11 | High | ### TL;DR  `ingero trace --inference` is now a first-class production daemon shape for GPU **inference** workloads (vLLM, SGLang, TGI, Triton), sitting alongside the training-flavoured `--fleet-workload-type` path that v0.10..v0.15 built out. One flag bundles sub-second causal windows, an event sampler attached to the on-disk store, DB file rollover with retention, a typed UDS outlier protocol, and a per-workload step-duration baseline + outlier classifier. The classifier is **phase-awa |
| `v0.15.0` | 2026-05-07 | High | ## Changelog ### Features * 4cdbe1aad0e32f70c72163000a44283f9ffd1dc3 feat(auth): add bearer-token compare + hardened TLS keypair loader * 928fa03b9f0d63e5a39991d0b7b276f6cb708569 feat(export): OTLP Histogram encoder + per-event memcpy duration histogram * 92c1095dc5a88dfffa1eb3d577228c70cb5e8017 feat(export): Prometheus aggregates for nccl.collective.* (F2) * bbacb0f82697e0e657827e530291786354341c75 feat(kprobe): experimental closed-driver kprobe + uprobe surface * 294bb0591745783967337f264a240a |
| `v0.11.0` | 2026-05-01 | High | ## Changelog ### Features * 7e41a69e1af5d0966b1fa2a27ed9f32f3aea5370 feat(alerter): ingero-alerter sidecar (Slack + PagerDuty backends) * 91c95a7590d7bf389a02e78a4f953bbe3c19029a feat(check): --support-bundle flag writes diagnostic tarball + adds arch diagram to README * 4ab13faffd65927d78f5f386afdb67a0ecbd547d feat(emitter): cost-of-problem gauges (ingero.node.info + ingero.node.world_size) * b7e4d3770c32cde4bba15e3aeaadc434c61b9dd2 feat(emitter): rank/world_size on resource attrs + per-event U |
| `v0.10.0` | 2026-04-21 | High | ## Changelog ### Documentation * f7998b120ffdc7a7ec9a5339c41a9974ae7e70be docs: drop unnecessary writeups; security fixes already in 01b181c ### Other * 0fdf9ef9257458db636807004d80d4467ca7c4a5 release: v0.10.0 |
| `v0.10.0-rc1` | 2026-04-20 | High | ## Changelog ### Features * ecaaad5767b858f2ca883ac85918710e41cc936c feat(health): - threshold consumption, degradation, classification * aa889dba874103e61ec926b6289477cdf7a1279f feat(health): agent-side Fleet integration * 4e0dd4b95deed4f1004a4739e2b3705a85c78fc6 feat(health): wire real signal collector, rotate DNS, tune warmup * 1235cf045a7bb4382ccfb5ca0086344578f35585 feat(helm): add fleet-push mode to ingero chart * d5f4ea7dfed6e3b614071993a2183f6c5096c842 feat: add Fleet interface contract |
| `v0.9.2` | 2026-04-16 | High | The in-kernel eBPF Python frame walker (`--py-walker=ebpf`) is the headline of this release. After extensive validation   on container and bare-metal Linux, through fork storms and lifecycle edge cases, it now delivers the advantages the   userspace walker cannot:    - works in containers (Docker, K8s) without debug symbols or host mounts   - works on distro-patched CPython (Ubuntu 24.04, etc.) via runtime offset harvesting   - works at `kernel.yama.ptrace_scope=3` with `--pid X` (PID-spe |
| `v0.9.1` | 2026-04-03 | Medium | > **Note:** The multi-node features in this release (fan-out queries, offline merge, Perfetto export) are interim solutions for cross-node GPU investigation. A dedicated cluster-level observability and diagnostics tool with native multi-node support is coming soon.  **One command. Every node. Full causal chain.**  Ingero can now investigate distributed GPU workloads across multiple nodes from a single CLI command, MCP tool call, or offline database merge. Diagnose which rank stalled, why, |
| `v0.9.0` | 2026-04-01 | Medium | Ingero can now trace the full CUDA Graph lifecycle — capture, instantiate, launch — via eBPF uprobes on libcudart.so.  Zero application modification, zero CUPTI dependency, production-safe overhead.   CUDA Graph Observability  - eBPF probes for cudaStreamBeginCapture, cudaStreamEndCapture, cudaGraphInstantiate, and cudaGraphLaunch — covers the stream capture path used by PyTorch torch.compile, vLLM, and TensorRT-LLM - Causal correlation connects graph events to system state: OOM during gr |

## Citation

- HTML: https://www.freshcrate.ai/projects/ingero
- Markdown: https://www.freshcrate.ai/projects/ingero.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/ingero/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._