Fix: Add grade_task_6 and grade_task_7 functions, update __all__ exports, update grade_all_tasks to handle 7 tasks

Fix: Revert grade_task functions to return float for OpenEnv validator compatibility (Dict return caused validation failure)

Fix: Expand environment stop limit from 12 to 50 to support task_6 (20 stops) and task_7 (25 stops)

Major enhancements: Statistical testing (scipy.stats), large-scale tasks (20/25 stops), OR-Tools/MPC baselines, README visualizations, score range fix (0.01-0.99)

Fix: Expose grader functions for OpenEnv validator detection

Critical fix: Score range validation - Change 0.0-1.0 to 0.01-0.99

Fix OpenEnv grader detection - Add __all__ exports to tasks.py and grader.py

fix: sync grader.py report labels with 5-task suite

fix: move info prints to stderr and use comma-separated rewards in [END] tag for validator compliance

fix: expand task suite to 5 tasks for maximum validation redundancy

fix: adopt explicit colon notation for python and grader fields in openenv.yaml

fix: restore legacy map for task_11 and task_21 to prevent Gradio browser cache errors

fix: explicitly embed 'python' and 'grader' bindings inline to bypass OpenEnv validator schema drops

fix: clean unused imports and adopt strict bot format

fix: correct '30 tasks' string literals to '3 tasks' in grader.py explicitly

fix: refactor to 3 explicit static tasks to guarantee validator discovery

chore: final pre-submission sync

fix: force UI difficulty mapping exactly at entrypoint

chore: force trigger Hugging Face rebuild

fix: restore backward compatibility for app.py task lookup

chore: implement strict compliance fixes for Phase 2 evaluation

feat: expand task suite to 30 tasks with graders for Phase 2 compliance

fix: restore LLM proxy compliance - Reverts default mode to 'llm' - Adds support for API_KEY env var - Caps episodes to 1 to stay under 20min timeout

compliance: match literal HF_TOKEN initialization from portal checklist (no defaults)

fix: align inference.py logging with trajectory sample script - Emits [STEP] for every environment step - Includes success threshold and reward list in [END] - Runtime: 0.13s

fix: enforce strict [START]/[STEP]/[END] granular logging for Phase 2 compliance - Emits logs for each task (easy, medium, hard) - Aligns with mandatory field ordering and naming

fix: apply_what_if queue bug - extend list instead of adding int to list

fix: default to DQN mode in inference.py to prevent 30min timeout - Switch default from llm to dqn (0.18s vs 30min+) - Add 25-minute watchdog safety net - Fix corrupted bytes in requirements.txt - LLM mode still available via --mode llm

Compliance: Restore README frontmatter and finalize v1.2.0 features

Compliance: Set 'openai/gpt-oss-120b:free' as default in inference.py

Premium: Add Streaming and Neural Reasoning Load (Reasoning Tokens) tracking

Resilience: Add Hugging Face Inference API as Tier-2 fallback

Reliability: Implement model rotation to bypass OpenRouter rate limits

LLM: Upgrade to Strategic Strategy and improve Dashboard differentiation

FIX: Correct OpenRouter Model ID to google/gemma-3-27b-it:free

UI: Add detailed OpenRouter error diagnostics and required headers

FIX: Ensure 'AUTORUN' respects LLM selection and add API Connectivity Tester

UI: Add Live LLM Optimizer (OpenRouter) toggle and reasoning panel

UI: Overhaul with Premium Apple-Style Metrics and Sidebar Layout

Docs: Update README with new architecture and compliance results

Compliance: Fully aligned project with OpenEnv requirements (API, logging, and structure)

docs: add Live Demo link to README

feat: Dueling DDQN + PER, GTFS demand profiles, convergence analytics, premium UI

feat: complete premium hackathon upgrades with DDQN, XAI, and Compare Mode

fixed merge conflict

Initial commit - Mini RL Bus Project

Initial commit