Spaces:
Sleeping
Sleeping
Commit History
-
one more
affu1436079dab -
one more time
affu14313c6248 -
cmmit kodthuna
affu143f18f9f9 -
Copied source project
Deepika3d87f50
-
Add diagnostic logging to task_registry for validator discovery debugging
85033ed -
Register all 7 tasks in task_registry with wrapped graders
1b4713c -
Revert: restore (0.001, 0.999) clamping per validator spec
efb92b3 -
Fix: grader reward clamping - use [0.0, 1.0] range per validator spec
1aaa0e8 -
Revert: environment.step() is now grading-only (LLM solving happens in inference.py)
61cc465 -
Fix: use HF_TOKEN or API_KEY for autonomous solver credentials
d255d10 -
Integrate LLM agent into environment.step() - autonomously generate fixes when validator sends empty fixed_config
02c2ba4 -
Enhanced validator payload diagnostic logging - capture action schema and fixed_config details
101d812 -
Debug: log what validator is actually submitting to step endpoint
eac318a -
Add deployment marker to verify Space is running latest code
bfaedde -
Debug: log final config string sent to grader before evaluation
2794216 -
Debug: log raw model output to inspect formatting issues
4ed1d8a -
Optimize LLM prompts and model selection for better config-debug task solving
68dc097 -
Dual-interface compatibility: graders return float for validator, environment converts to tuple internally
79145d3 -
Debug: add grader output logging to inspect raw reward values
d661cb5 -
Critical: fix grader wrapper to return tuple (reward, error_msg, bugs_fixed) with clamped reward
7eba106 -
Fix: implement grader wrapper API with float returns and strict (0.001, 0.999) bounds
1818234 -
Fix: limit task registry to 3 validated graded tasks (JSON, YAML, Dockerfile)
6a53124 -
Fix: add reward clamping to all graders for strict (0,1) score range
d15f73e -
Fix: resolve syntax error in k8s_grader string formatting
0eb6f46 -
Fix: adjust all grader scores to strictly between 0 and 1 (not including 0 or 1)
5c4143f -
Fix: add guaranteed proxy validation call immediately after client creation
d88004f -
Fix: use strict os.environ[] for API_KEY and API_BASE_URL to match validator requirements
0c3bdac -
Fix: use gpt-4o-mini as default model for better proxy compatibility
4b57735 -
Replace repo with updated config-debug files
cda147c