docs: restructure README and add harness-eval references - #77
Conversation
Moved gates architecture, submission formats, extensibility, and persistence sections into dedicated files under Docs/. README reduced from 691 lines to 139 with links to the detailed documentation.
Updated all user-facing references in docs, comments, docstrings, YAML comments, and metadata. The Python package name (abevalflow) and OpenShift namespace (ab-eval-flow) are unchanged as those are deployed infrastructure names.
|
Added a second commit that renames all remaining "ABEvalFlow" references to "Agentic Eval Flow" across the codebase (47 files, 101 replacements). Updated: Docs/, Python docstrings and comments, YAML comments, metadata files, test assertions, plan.md. Kept as-is: |
PR ReviewOverall: Looks good, with a couple of minor issues to flag. What's Good
Issues
VerdictThe PR is solid -- the restructuring makes the docs much more navigable and the harness-eval references are placed appropriately. I'd suggest fixing the space-in-path issues in |
- Fixed space-in-path in plan.md and aeh_downloadverifiererror.md (quoted paths, use actual directory name agentic_eval_flow) - Fixed repo structure tree in implementation_plan.md to show agentic_eval_flow/ instead of Agentic Eval Flow/ - Replaced em dashes with -- across all Docs/ files for consistency with README style
|
Fixed all 4 review items:
|
Summary
The README was 691 lines long, mixing quick-start content with full reference documentation. This PR slims it down to 139 lines and moves the detailed sections into dedicated files under
Docs/.What moved
Docs/gates-architecture.mdDocs/submission-formats.mdDocs/extensibility.mdDocs/persistence.mdCompass Facts Integration was already in
Docs/compass_facts_integration.md.What's new
Added harness-eval references:
Docs/gates-architecture.md: documented both security scanners (Cisco + harness-eval) and both quality sources (LLM review + harness-eval)What stayed in README