Sync github 89af584 + 4b5b65c: demo video link + blog image-ref formatting fix

Sync github 100e431 + aa873a5: slim README, add blog.md

custom: bigger panel title, 'CaptainRL — Live cockpit' (mirrors github af58f9b)

custom: in-process episode driver + CaptainRL rename (mirrors github c17c1ba)

Sync teammate's commit 22246ba: integrate Cartesia TTS for live commentary

ui: opponent + captain selectors with explicit presets (mirrors github 54e156d)

sync: pull latest from main (model_server.py, captain LLM toggle in ui.py, 0.6B configs, SUBMISSION + RUNTIME_DURABILITY docs)

sync: today's source updates (XML-only prompt, reward unclip, neg-reward on loss, pinned versions, configs reorg)

cleanup: drop illustrations/ (large, optional run logs)

cleanup: drop unsloth_compiled_cache (not needed in Space)

docs: sync tool count and descriptions across manifest, models, and PRD

Fix field placements: 9 fielders per preset, enforce powerplay cap, correct boundary default

illus: all 3 opponent modes verified across 5-over inference + train-smoke

feat: CricsheetOpponentPolicy multi-format support (T5/T20/ODI)

feat: CricsheetOpponentPolicy — real match data as opponent decisions

docs: update README, slides, illustrations with format mapper, tool budget, player rosters, 5/20-over baselines

feat: llm_live as default opponent (graceful fallback to heuristic if no API key)

fix: inference max_overs not reaching env + live per-delivery log

feat: merge PR#2 (player rosters), tool budget fines, roster-aware selection

feat: format-aware batter/bowler rules + Cricsheet ODI transitions

fix: Markov calibration, opponent plan leak removed, pre-create test folders

fix: complete UI rewrite — per-tool groups, correct field dispatch, auto-play

fix: incremental log writes, openenv.yaml descriptions, openai dep

feat: all-12-tool UI, auto-play, timestamped run folders, model fix

Set captain and opponent models to google/gemma-4-e4b-it via YAML config

Ensure non-zero rewards for all valid tool calls

chore: remove training notebook artifact

feat: add field-aware hybrid ball simulation and timing logs

fix: tighten cricket rules and reward visibility

fix: match OpenEnv Gradio builder signature

fix: enforce bowler changes at over boundaries

test: add 5-over training rollout smoke

fix: preserve match outcome reward objective

feat: mount Gradio UI in OpenEnv app

docs: clarify HF Gemma runs and OpenEnv traces

Merge pull request #1 from pratinavseth/dream11-reward

Added Dream-11 reward

feat: add YAML config for server and runners

data: add cached Gemma-2B opponent plans

feat: add HF router LLM opponent defaults

fix: wire reset options for max_overs

docs: surface benchmark guides in readme

docs: add experiment workflow guide

docs: explain adaptive benchmark architecture

feat: align benchmark data and training roadmap

feat: add adaptive cricket benchmark

data: added dream 11 points rubric

chore: add HF Spaces YAML frontmatter to README

chore: add OpenEnv submission files (Dockerfile, validate, uv.lock)

docs: comprehensive README update + fix ui.py dead code