feat(commentary): introduce universal rafters-critic for scenario commentary and cadence control

feat(media): introduce MediaRouter and stubs for image and speech generation

feat: Update model references to MiniCPM5 and adjust related documentation and tests for transformers 5.x compatibility

Refactor local provider to support on-device loading with accelerate

feat: Add use_cache parameter to support KV cache handling in models and update tests

Refactor local model configurations and update tests

feat: Update model routing and local provider to support multi-sponsor tiers and auto-class resolution

feat: Ensure full materialization of model weights to prevent meta tensor issues in ZeroGPU

fix: Ensure proper unpacking of BatchEncoding in model generation to prevent AttributeError

feat: Replace llama.cpp backend with in-process transformers backend for local GPU inference

Implement local in-process inference backend for transformers models

feat: Implement llama.cpp backend for local inference with GGUF models

Merge remote-tracking branch 'origin/main' into worktree-scenarios

feat: make every scenario arena-grade with a competition contract

feat: Implement well-known typed fields for verdicts in output models

feat: implement per-event model attribution for enhanced event tracking

feat: update agent personas and scheduling in YAML configurations; enhance event handling and memory deduplication in core logic

feat: update HF model catalogue to prioritize chat-capable model and adjust tests for new routing logic

feat: require live infrastructure for event store and inference, removing offline mode

feat(observability): instrument models, conductor, memory, ledger (Units 2-6)

refactor: tidy HF backend per /simplify review

feat: add Hugging Face inference backend next to Modal

feat: implement streaming agent responses and enhance conversation flow with repeat guards

feat(core): add resilient loop and shared blackboard

feat: Enhance model routing and deployment flexibility

Merge branch 'feat/live-open-table' into feat/live-integration

feat: add Open Table live-chat scenario (curious/skeptic/host)

feat: add the-steeped spy-game scenario (word-pair bluff)

feat: add Fishbowl UI presenter layer and say-vs-think engine support

feat: unify model catalogue and self-hosted routing

feat: Instructor-validated structured output on the live path

feat: LiteLLM gateway routing profiles to Modal-served small models

feat: update setup instructions and dependencies for uv integration

Refactor code structure for improved readability and maintainability

feat: phase 1 — memory layer, governor, second scenario, 70 passing tests

feat: scaffold hackathon agent theater