Align inference with hackathon validator

Publish structured task grader metadata

Harden OpenEnv grader entrypoints

Restore submission-38 task validation surface with proxy-safe inference

Normalize exported score/reward fields to strict (0,1)

Eliminate zero rewards and raise score floors for validator safety

Normalize task-id aliases across reset and grading paths

Harden manifest task schema and canonical grader paths

Return plain float from all manifest-facing graders

Ensure fallback inference output always includes 3 valid tasks

Handle inference exceptions and avoid strict-mode crashes

Enforce strict inference: no fallback without proxy vars

Use validator proxy API_BASE_URL+API_KEY in strict inference mode

Implement non-constant deterministic graders and align required inference env vars

Make graders dict+numeric compatible for deep validator callers

Harden deep-validator grader discovery compatibility

Simplify graders to return plain float (OpenEnv spec)

Make grader outputs dict+numeric compatible

Add deterministic validator-safe grader mode

Clamp step reward values to strict (0,1) for validator compatibility

Match OpenEnv runtime health and MCP contracts

Return plain float from task grader entry points

Package task grader modules for deep validator imports

Maximize deep-validator grader compatibility

Add root endpoint for platform probe compatibility

Ignore Python packaging metadata artifacts

Harden grader compatibility for deep validators

chore: trigger space rebuild

Stabilize task grader discovery paths for Phase 2 validation

Fix validator discovery for Phase 2: point openenv.yaml back to tasks.X.grader:grade

chore: trigger Hugging Face Space rebuild

Use package grader entry points in openenv manifest

Fix OpenEnv grader discovery by including tasks package in Docker image

fix: rename task ids to use underscores instead of hyphens

fix: rename 'final_score' to 'score' in inference output payload

fix: restructure graders strictly to tasks.task_name.grader:grade

fix: grader functions must return float, not dict

fix: add openenv-core>=0.2.0 dependency required by openenv validate checker

feat: implement full OpenEnv spec compliance

fix(critical): add task graders to openenv.yaml for Phase 2 validation

fix(critical): rewrite inference.py to use in-process environment

fix(inference): add rule-based fallback when API keys are absent

fix: replace bad output JSONs with valid strictly-bounded scores

fix(validator): use strict (0<score<1) bounds to match hackathon Phase 2 check

chore: migrate image files to Git LFS

chore: add LFS tracking for image files (png, jpg, jpeg, gif, webp)

Merge branch 'main' of https://huggingface.co/spaces/Prince9868/Guardian-Agent

fix: eliminate all remaining 0.0/1.0 boundary score values

fix: clamp evidence_score and all boundary score values to strict (0,1) range

fix(eval): completely purge all 0.0 and 1.0 explicit bounds