Observability: OTel semantic conventions for agentic systems#33
Conversation
|
Skipping CI for Draft Pull Request. |
rubambiza
left a comment
There was a problem hiding this comment.
Thanks for kicking off the proposal @david-martin . I am a clean slate to the world of observability, so excuse any obvious questions in my review 😅
|
Capturing the outcome of conversation on slack & meet.
|
✅ Deploy Preview for kube-agentic-networking ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
Signed-off-by: David Martin <davmarti@redhat.com>
Signed-off-by: Evaline Ju <69598118+evaline-ju@users.noreply.github.com>
…n attributes from otel Signed-off-by: David Martin <davmarti@redhat.com>
Signed-off-by: Evaline Ju <69598118+evaline-ju@users.noreply.github.com>
Signed-off-by: Evaline Ju <69598118+evaline-ju@users.noreply.github.com>
Signed-off-by: Evaline Ju <69598118+evaline-ju@users.noreply.github.com>
📝 Consolidate span attributes
…for the most part Signed-off-by: David Martin <davmarti@redhat.com>
Simplify proposal to make it clear we're not defining new attributes …
Signed-off-by: Evaline Ju <69598118+evaline-ju@users.noreply.github.com>
📝 Add alternate conventions section
|
Overall this proposal looks good to me, thanks @david-martin! /approve /hold Will defer to @gkhom for final approval. |
Signed-off-by: David Martin <davmarti@redhat.com>
Signed-off-by: David Martin <davmarti@redhat.com>
7a3c85e to
cc31050
Compare
| ### Guardrailing | ||
|
|
||
| This shows a trace example of a guardrail blocking a request at a gateway. |
There was a problem hiding this comment.
👋 from OpenTelemetry GenAI SIG
I don't have a ton of context on this proposal, but I wanted to share a proposal for Guardrails open-telemetry/semantic-conventions#3233
| │ | ||
| │ POST /mcp | ||
| │ Headers: | ||
| │ traceparent: 00-c3d4e5f6a7b8c9d0e1f2a3b4c5d6e7f8-9c1d2e3f4a5b6c7d-01 |
There was a problem hiding this comment.
FYI, propagation needs to happen in the JSON-RPC layer for MCP because of it's stateful nature.
rubambiza
left a comment
There was a problem hiding this comment.
Mostly clarifying questions, not sure if they may require any changes. They certainly do not need to block merging.
| │ } | ||
| │ } | ||
| │ } | ||
| ▼ |
There was a problem hiding this comment.
Is this the point where the traffic hits a k8s cluster?
There was a problem hiding this comment.
I would say the request is inside the MCP aware router/gateway, or in the MCP Server at this point.
Either of those could be in a k8s cluster, yes.
After this point, it's up to the MCP server where that traffic goes (if anywhere), before the response is returned.
Signed-off-by: Evaline Ju <69598118+evaline-ju@users.noreply.github.com>
Signed-off-by: Evaline Ju <69598118+evaline-ju@users.noreply.github.com>
📝 Update traceparent and JSON-RPC bodies in examples
|
@david-martin anything else thats blocking you from getting this in? |
|
@LiorLieberman I think we're good to go here. Thanks to everyone for the help and feedback on getting this to a mergable state. |
|
/approve |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: david-martin, LiorLieberman The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
/unhold |
What type of PR is this?
/kind documentation
What this PR does / why we need it:
Defines tracing schemas for agentic systems by following OpenTelemetry semantic conventions, with proposed extensions for permission enforcement outcomes.