Skip to content

feature: Enabling http-headers to support authenticated triton connections#1002

Open
satias10 wants to merge 2 commits intotriton-inference-server:mainfrom
satias10:feature/headers
Open

feature: Enabling http-headers to support authenticated triton connections#1002
satias10 wants to merge 2 commits intotriton-inference-server:mainfrom
satias10:feature/headers

Conversation

@satias10
Copy link
Copy Markdown

What does this PR do?
This PR adds a new configuration option, --triton-http-headers ,which allows users to inject custom HTTP headers into Model Analyzer's internal TritonClient.

Why is this needed?
Previously, users could pass headers to perf_analyzer, but the internal Python client—responsible for loading/unloading models and checking server readiness—did not support custom headers.

Key Changes:

  • CLI & Config: --triton-http-headers which accepts a JSON string (CLI) or dictionary (YAML).
  • Client: Updated TritonHTTPClient to pass these headers during load_model, unload_model, is_server_ready, and get_model_repository_index.

Verification:
[x] Unit Tests: All tests passed (python3 -m unittest discover -s tests -t .).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

1 participant