Skip to content

[TRTLLM-12842][feat] Maximal LLMAPI capture in usage telemetry#14398

Merged
venkywonka merged 2 commits into
NVIDIA:mainfrom
venkywonka:venky/telemetry-config-redesign
Jun 13, 2026
Merged

[TRTLLM-12842][feat] Maximal LLMAPI capture in usage telemetry#14398
venkywonka merged 2 commits into
NVIDIA:mainfrom
venkywonka:venky/telemetry-config-redesign

Conversation

@venkywonka

@venkywonka venkywonka commented May 21, 2026

Copy link
Copy Markdown
Collaborator

Motivation

Usage telemetry had coarse feature flags, not effective LLM API runtime-shape config. Needed: know how deployed configs are shaped without collecting private or user-controlled data.

Hard rule: no PII, no prompts, no outputs, no model/tokenizer paths, no env vars, no raw YAML, no arbitrary fields, no arbitrary strings.

Design

Full list and related TRTLLM documentation preview: https://venkywonka.github.io/TensorRT-LLM/developer-guide/telemetry.html

  • Add field-local Field(..., telemetry=...) opt-in metadata for approved LLM args fields.
  • Walk only declared Pydantic fields from validated effective args. Unmarked fields stay invisible.
  • Emit deterministic llmApiConfigJson for sanitized config values.
  • Emit llmApiConfigMetaJson with capture version, policy version, args class, digests, counts, source, and unsafe-exclusion status.
  • Fail closed in sanitizer: reject paths, raw strings without allowlists, dict/object payloads, Any, callables, unsafe unions, and heterogeneous tuples.
  • Keep featuresJson for legacy summary consumers. Add TODO to dedupe with llmApiConfigJson later.
  • Generate telemetry docs during Sphinx build from llm_args.py annotations so exhaustive captured-key docs do not drift manually.
  • Preserve existing usage opt-out behavior and precompiled editable validation path for Python/schema-only changes.

Testing

  • python3 -m json.tool tensorrt_llm/usage/schemas/trtllm_usage_event_schema.json
  • python3 -m py_compile on telemetry usage modules and docs generator
  • ruff check / ruff format --check on touched Python files
  • git diff --check
  • Usage unit coverage: 199 passed, 5 skipped
  • Forced real-GPU usage capture smoke: 2 passed
  • Generated-docs unit coverage: 2 passed
  • Full Sphinx HTML build completed; telemetry page renders README link, schema link, privacy invariant, and exhaustive LLM API config field tables.
  • Real GPU precomputed-wheel validation path: 54 passed
  • Real-CUDA smoke payloads passed for single-GPU, TP=2 compile, and scheduler/KV-cache privacy shapes.
  • TRT-LLM MCP runtime contract check: status=ready

Summary by CodeRabbit

  • Documentation

    • Enhanced README telemetry section with explicit enumeration of excluded data categories (outputs, model paths, PII, raw configurations, persistent identifiers).
    • Expanded developer guide with detailed telemetry capture reference and configuration field allowlists.
  • New Features

    • Introduced sanitized, opt-in LLM API configuration telemetry with deterministic capture and validation; schema version bumped to 0.2.
    • Added disaggregated launcher telemetry environment propagation for deployment tracking.
  • Tests

    • Comprehensive test coverage for configuration telemetry capture, sanitization, and end-to-end validation.

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot kill

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49757 [ kill ] triggered by Bot. Commit: 2ea9ea6 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49757 [ kill ] completed with state SUCCESS. Commit: 2ea9ea6
Successfully killed previous jobs for commit 2ea9ea6

Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49758 [ run ] triggered by Bot. Commit: 2ea9ea6 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49758 [ run ] completed with state SUCCESS. Commit: 2ea9ea6
/LLM/main/L0_MergeRequest_PR pipeline #39358 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot kill

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49763 [ kill ] triggered by Bot. Commit: f8773b6 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49763 [ kill ] completed with state SUCCESS. Commit: f8773b6
Successfully killed previous jobs for commit f8773b6

Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49764 [ run ] triggered by Bot. Commit: f8773b6 Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot kill

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49771 [ kill ] triggered by Bot. Commit: 1159335 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49772 [ run ] triggered by Bot. Commit: 1159335 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49771 [ kill ] completed with state ABORTED. Commit: 1159335

Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49764 [ run ] completed with state ABORTED. Commit: f8773b6

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot kill

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49778 [ run ] triggered by Bot. Commit: 5966c67 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49779 [ kill ] triggered by Bot. Commit: 5966c67 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49778 [ run ] completed with state ABORTED. Commit: 5966c67

Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49772 [ run ] completed with state ABORTED. Commit: 1159335

Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49779 [ kill ] completed with state SUCCESS. Commit: 5966c67
Successfully killed previous jobs for commit 5966c67

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot kill

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49781 [ run ] triggered by Bot. Commit: 5e06861 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49782 [ kill ] triggered by Bot. Commit: 5e06861 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #49781 [ run ] completed with state ABORTED. Commit: 5e06861

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #53714 [ run ] triggered by Bot. Commit: b9b771d Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #53714 [ run ] completed with state FAILURE. Commit: b9b771d
/LLM/main/L0_MergeRequest_PR pipeline #42843 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #53908 [ run ] triggered by Bot. Commit: b9b771d Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #53908 [ run ] completed with state SUCCESS. Commit: b9b771d
/LLM/main/L0_MergeRequest_PR pipeline #43006 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

Capture schema-stable LLM API config telemetry from the type-driven manifest.

Add the generated manifest golden, docs updates, and runtime/config capture test coverage.

Signed-off-by: venkywonka <23023424+venkywonka@users.noreply.github.com>
Signed-off-by: venkywonka <23023424+venkywonka@users.noreply.github.com>
@venkywonka
venkywonka force-pushed the venky/telemetry-config-redesign branch from b9b771d to 010df91 Compare June 13, 2026 04:37
@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54010 [ run ] triggered by Bot. Commit: 010df91 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54010 [ run ] completed with state SUCCESS. Commit: 010df91
/LLM/main/L0_MergeRequest_PR pipeline #43094 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54031 [ run ] triggered by Bot. Commit: 010df91 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54031 [ run ] completed with state FAILURE. Commit: 010df91
/LLM/main/L0_MergeRequest_PR pipeline #43115 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54034 [ run ] triggered by Bot. Commit: 010df91 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54034 [ run ] completed with state FAILURE. Commit: 010df91
/LLM/main/L0_MergeRequest_PR pipeline #43118 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54039 [ run ] triggered by Bot. Commit: 010df91 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54039 [ run ] completed with state FAILURE. Commit: 010df91
/LLM/main/L0_MergeRequest_PR pipeline #43123 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54040 [ run ] triggered by Bot. Commit: 010df91 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54040 [ run ] completed with state FAILURE. Commit: 010df91
/LLM/main/L0_MergeRequest_PR pipeline #43124 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

@venkywonka

Copy link
Copy Markdown
Collaborator Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54044 [ run ] triggered by Bot. Commit: 010df91 Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #54044 [ run ] completed with state SUCCESS. Commit: 010df91
/LLM/main/L0_MergeRequest_PR pipeline #43128 completed with status: 'SUCCESS'

CI Report

Link to invocation

@venkywonka
venkywonka merged commit 4e1776a into NVIDIA:main Jun 13, 2026
7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

7 participants