Skip to content

v0.31.0 - Namespace Architecture Alignment & BDD Quality

Choose a tag to compare

View all tags
@manykarim manykarim released this 23 Mar 08:53
· 66 commits to main since this release
v0.31.0
95b36e5

rf-mcp v0.31.0

Previous Release: v0.30.1 (2026-02-26)

New Features

Namespace & Execution Architecture Alignment (ADR-020)

  • Fixed zombie variable scope bug — scope depth was 4 instead of correct 3 due to redundant namespace.start_test() / end_test() calls
  • Removed _try_execute_from_library — direct Python method calls bypassed RF's ${var} resolution. All keywords now go through runner.run() or BuiltIn.run_keyword()
  • Initial MCP test is now tracked in current_run_test/current_res_test so start_test_in_context can auto-end it properly

BDD & Data-Driven Test Suite Generation (ADR-019)

  • BDD scenario quality improvement — word-level embedded name building, instrumentation filtering, semantic phase detection, content-aware keyword naming
  • Data-driven template support — suite_template, column_headers, named_data_rows, add_data_row action with auto-promotion mode
  • bdd_group / bdd_intent params on execute_step for behavioral keyword grouping

Persistent Semantic Memory (ADR-014)

  • sqlite-vec + model2vec based memory with 3-tier fallback
  • 5 MCP tools: recall_step, recall_fix, recall_locator, store_knowledge, get_memory_status
  • Proactive response augmentation — memory hints injected into tool responses automatically

Token Reduction (ADR-015/016/017/018)

  • Artifact externalization + fetch_artifact MCP tool
  • Slim profiles with AutoProfileSelection for 8K model support
  • Delta state via mode/since_version params on get_session_state
  • 71-88% token savings across all tools; 15-step workflow: 8,763 -> 2,508 tokens

Output Noise Reduction

  • session_variables removed from responses (80% reduction per step)
  • changed_variables tracking — only non-builtin vars that changed
  • Hint deduplication, grouped warnings, compact HTTP responses, auto-delta mode

Improvements

  • Guarded double variable assignment — RF's VariableAssigner handles RunKeyword.assign; manual assignment now only runs for BuiltIn fallback
  • ExecutionStatus re-raise guard prevents BuiltIn retry on assertion failures
  • Copilot CLI E2E test integration in CI with multi-model support and rate limit detection

Bug Fixes

  • Fixed zombie variable scope in RF execution context creating 4 scopes instead of 3
  • Fixed direct method bypass skipping ${var} resolution in keyword arguments
  • Fixed double variable assignment when RunKeyword.assign already handles it
  • Fixed requests import crash (PR #55)
  • Fixed Windows SQLite file-lock in memory integration test teardown
  • Fixed BDD stripping, enum mismatch, and missing E2E fields in CI
  • 5 data-driven fixes: template rendering, BDD wrapper args, empty values (${EMPTY}), builtin var filtering

Installation

pip install rf-mcp==0.31.0

# With optional features
pip install rf-mcp[web]==0.31.0      # Browser + SeleniumLibrary
pip install rf-mcp[api]==0.31.0      # RequestsLibrary
pip install rf-mcp[memory]==0.31.0   # Persistent memory
pip install rf-mcp[all]==0.31.0      # Everything
claude mcp add rf-mcp -- uvx rf-mcp@0.31.0
Assets 6
Loading