chore: removing logs

fix: remove exception

feat: rewards upgrade

feat: updating prompt and reward consdition

chore: updating the [END] log

feat: LLM agent model change

fix: clamp all rewards and scores to [0.10, 0.90]

fix: clamp all score paths to (0.01, 0.99), fix reward field name, add per-task score line

fix: reduced the number of scenario in inference

fix: logs in inference

chore: doc string update and remove unused import

fic: score condition

fix: score format and range

fix: score range

fix: enforce reward bounds (0.01–0.99) and 2 decimal precision across grader, env, and inference

fix: reward scores are updated to be between 0 and 1

chore: code cleanup

chore: logs format update

chore: updating logs

fix: rewards field shows single composite score instead of step CSV

chore: 2 sceanrios per difficulty - inference

chore: only one scenario is run per difficulty

chore: restrict stdout to START/STEP/END for eval compliance

chore: clean up all the unnecessary comments

chore: updating logs

fix: comply with openenv stdout spec, preserve inspection data in history, sharpen medium-tier label rules

fix: error handling for episode run loop

fix: updating the logs to align with the evaluation standards

fix: error handling for HF_TOKEN env

fix: tighten label rules for underfitting, overfitting, and vanishing gradients

feat: add judge fallback

refactor: moving llm judge inside server dir

fix: harden label rules to prevent missing_regularization misfires

feat(inference): add label decision rules and stop rules to system prompt

feat: add local agent support

feat: upgrading the system and user prompt, upgrading the _make_env() function

feat: adding [END] log for each episode and error handling for websocket

chore: docstring update

fix: adding reasoning in episode run loop and refactor commments

feat: implementing judge LLM which contributes to 15% of scoring

feat: 3 modes of difficulty and updating the logs

feat: init inference

refactor: WhyDidItFailAction and WhyDidItFailObservation classes

Initial commit