Spaces:
Sleeping
Sleeping
Commit History
-
fix: Use proper Callable type annotation instead of callable builtin
Nihar Shah8afc900
-
Fix: Add arbitrary_types_allowed to Task model and update test expectations for new score range (0.05-0.95)
Nihar Shah1462b73 -
Add explicit graders module for OpenEnv validation
Nihar Shah- Created app/graders.py with explicit grader functions for each task - Exported GRADERS registry with 4 graders (requirement: ≥3) - Updated app/__init__.py to export graders - Updated openenv.yaml to reflect score bounds (0, 1) - All graders return scores strictly between 0 and 160604aa -
Fix Phase 2: ensure task scores strictly between 0 and 1
Nihar Shah- Modified TaskGrader.grade_task() score range from [0.0, 1.0] to [0.05, 0.95] - Updated fallback scores in inference.py from 0.0 to 0.05 - All 4 tasks now produce valid scores in exclusive range (0, 1) - Verified with comprehensive test suite1943b8f -
🎉 Success: HF Spaces deployment verified - ReDoc and all endpoints working! ✅
Nihar Shahad21580
-
Fix: Support both Pydantic models and dicts in step() and reward calculation
Nihar Shah- Updated env.py step() to handle both dict and Action Pydantic models - Updated reward.py calculate() to handle both dict and Pydantic models - All 120 tests pass - API endpoints validated and working correctly - Supports flexible action format: dict or Action model8577b4b -
Clean up: Remove all .md files except README.md - keep project lean
Nihar Shahfd35c86 -
Add Priority 1 improvements: Procedural task generation, business-aware rewards, constrained environment
Nihar Shah4150d9b
-
Add ultra-advanced features: semantic memory, analytics, curriculum learning, neural ranking (38 new tests, 1300+ lines)
Nihar Shahbb291ef
-
Initial commit: OpenEnv Business CRM Query Environment with memory system and multi-agent architecture
Nihar Shah8f7a08e