feat(harness): add Agent Inspector and Mantle (OpenAI-compatible) examples#1644
Open
di-brasil wants to merge 2 commits into
Open
feat(harness): add Agent Inspector and Mantle (OpenAI-compatible) examples#1644di-brasil wants to merge 2 commits into
di-brasil wants to merge 2 commits into
Conversation
di-brasil
force-pushed
the
feat/harness-observability
branch
2 times, most recently
from
dae7a07 to
21be6be
Compare
di-brasil
changed the title
Adds an AgentCore CLI harness sample under 01-features/01-harness/01-advanced-examples/06-getting-started-with-agent-inspector/. It deploys a harness from scratch, invokes it to generate telemetry, confirms the OpenTelemetry spans landed in the aws/spans CloudWatch log group, then launches the Agent Inspector (agentcore dev) so you can chat with the agent and watch its sessions, traces, and spans live. Deploy comes first because the Inspector has nothing to show until the harness is deployed and invoked. Includes a recording-safe demo.sh (account ID and local username masked), zero-arg cleanup.sh, README, system prompt, and a GIF shot-list. Tested end-to-end on @aws/agentcore@1.0.0-preview.12 in us-east-1.
New advanced example 08-mantle-endpoint/ showing how to run a Bedrock harness through the OpenAI-compatible bedrock-mantle endpoint via --api-format responses, using OpenAI's open-weight gpt-oss-120b (no API key — uses the execution role). Self-contained: README, demo.sh, cleanup.sh, system-prompt.md, .gitignore, images. Verified end-to-end on @aws/agentcore@1.0.0-preview.13 in us-east-1 (deploy + invoke + full GenAI spans in aws/spans). Documents the gotcha that the Mantle model id drops the -1:0 suffix (openai.gpt-oss-120b, not openai.gpt-oss-120b-1:0).
di-brasil
force-pushed
the
feat/harness-observability
branch
from
21be6be to
8ff6fd5
Compare
di-brasil
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds an AgentCore CLI harness sample under
01-features/01-harness/01-advanced-examples/06-getting-started-with-agent-inspector/that shows how to deploy a harness from scratch and then open the Agent Inspector (agentcore dev) to watch its sessions, traces, and spans live.The order matters, and the sample makes it explicit: deploy first, then inspect. The Agent Inspector reads the harness's telemetry from CloudWatch, so there is nothing to inspect until the harness is deployed and has handled an invocation.
demo.shdeploys, invokes the agent, confirms the OpenTelemetry spans landed in theaws/spanslog group, and only then launches the Inspector — so the data is already there when the UI opens.Every harness invocation is auto-instrumented by the service (no ADOT, no
OTEL_*variables, no code). The only account-level setup is enabling CloudWatch Transaction Search once.Tested end-to-end on
@aws/agentcore@1.0.0-preview.12inus-east-1(deploy + live invoke + spans confirmed inaws/spans+ Agent Inspector).What the sample includes
<ACCOUNT>and the local username/home path as<USER>throughout, so a recording is safe to share. Names are auto-generated per run, so it can be run repeatedly without collisions../demo.sh --self-testruns offline checks (no AWS calls).Notes
OTEL_*variables — instrumentation is service-side..gitignorekeeps the local recording (*.mov), workspace, and generatedaws-targets.jsonout of git.Testing
agentcore deploy+ threeagentcore invoketurns +agentcore devran end-to-end.aws/spanslog group, including thechatmodel-call span withgen_ai.usage.*token counts and thesession.idattribute, and the same data rendering in the Agent Inspector's Traces pane.