Repository navigation
AgentX v1.0.7
Install AgentX without cloning or building the repository:
pip install agentx-harness==1.0.7
agentx profile --url "$SERVER_URL" --model "$SERVED_MODEL_NAME" \
--tokenizer "$TOKENIZER" --concurrency "$CONC" \
--public-dataset "$DATASET" --output-artifact-dir "$OUTPUT_DIR"- Add the
agentxcommand, which selects the AgentX scenario automatically for flag-based profiles. Explicit scenario overrides and YAML configurations retain their existing behavior. - Publish
agentx-harnessto PyPI automatically on GitHub releases, with the package version matching the release tag. - Keep upstream AIPerf source metadata and the existing
aiperfcommand intact; packaging changes are applied only in the release build. - Include the datasets dependency required by the public AgentX trace corpora.
Wheel and source installations passed streaming benchmark integration checks on Python 3.11–3.13: https://github.com/SemiAnalysisAI/agentx-harness/actions/runs/38314954918
See #51. InferenceX integration: SemiAnalysisAI/InferenceX#3940.