Add Sotopia Arena (auth, Werewolf, ELO, EC2 deployment)#346
Open
nikitachaudharicodes wants to merge 87 commits intosotopia-lab:mainfrom
Open
Add Sotopia Arena (auth, Werewolf, ELO, EC2 deployment)#346nikitachaudharicodes wants to merge 87 commits intosotopia-lab:mainfrom
nikitachaudharicodes wants to merge 87 commits intosotopia-lab:mainfrom
Conversation
with minor bugs, will fix in future iterations
contain minor bugs, will fix in future iterations
Fixes several bugs preventing custom models (via custom/model@url format) from working:
- Fix parameter name in generate.py: api_base → base_url (line 257)
- Fix hardcoded "gpt-4" evaluator models in server.py (lines 309, 401)
Now uses model_dict.get("evaluator", model_dict["env"])
- Add markdown code block stripping in PydanticOutputParser
Many local LLMs wrap JSON in ```json...```, parser now handles this
- Fix format_bad_output to support custom models
Passes base_url/api_key through error recovery path
Conditionally uses response_format (custom servers may not support it)
…ility issues in the game
Refactor SocialDeductionGame for real-time history and cleaner prompts
- ParallelSotopiaEnv: Added `include_turn_marker` flag to control environment turn messages.
- SocialDeductionGame:
- Disabled environment turn markers to avoid duplication.
- Implemented real-time history appending via `recv_message` override and `agent_message_buffer`.
- Populated `action_instruction` in `Observation` for dynamic prompt instructions.
- Observation: Added `action_instruction` field.
- generate.py: Added `fill_template` helper for partial string formatting.
- LLMAgent: Updated `aact` to use `fill_template` to inject `action_instructions` into `custom_template`.
- Werewolves: Updated config description to populate `{agent_names}` dynamically.
next step, change script_like to false, and fix the rest errors that may cause
…_Sell_custom_models.py
previous commit reverted too much..
Found and fix the evaluation and generation error on the negotiation arena examples. - **Termination Fix**: Updated `ParallelSotopiaEnv` to pass the `env` instance to evaluators. Modified `RuleBasedTerminatedEvaluator` to correctly count active agents using `env.agents` instead of relying solely on message history, which caused early termination in the first turn. - **LiteLLM Support**: Updated `generate.py` to handle OpenAI schema limitations. Added `_fix_schema` to convert `prefixItems` (tuples) to `items` (arrays) and set `strict=False` to support dynamic dictionary outputs (Evaluator maps) while preventing `BadRequestError`.
…r arena games - generate.py: add action_instruction param to agenerate_action, inject as action_instruction_text in prompts - output_parsers.py: normalize non-canonical action_type (e.g. vote, kill, lynch) to 'action' with verb in argument Enables Werewolf phase-specific prompts and robust parsing; extensible for other arena games.
- Register/login with bcrypt + JWT, /auth/me and PUT /auth/me - /auth/me/games and /auth/me/stats for game history and stats - Legacy /auth/identity for backward compatibility with existing frontend
- GET /oauth/providers, /oauth/login/{provider}, /oauth/callback/{provider}
- Create/link UserProfile and OAuthAccount, issue JWT for frontend
- POST /oauth/link, GET /oauth/accounts, DELETE /oauth/unlink
- elo.py: ELO calculation, AI base ratings, update_user_elo_after_game,
process_game_elo, rank tiers (Bronze–Master), get_rank_tier/get_rank_progress
- leaderboard.py: GET /leaderboard, /ranks, /user/{pk}, /top/{count}
- GET /profile/me, /profile/me/elo-history, /profile/me/matches (optional game_type)
- PUT /profile/me; GET /profile/{user_pk}, /profile/{user_pk}/matches
- _build_profile_stats shared by get_my_profile and get_user_profile; role_stats, win streaks, rank_tier
…game routing - Include auth_router, oauth_router, leaderboard_router, profile_router - SimulationState: _human_players, _pending_actions, _pack_chats, _game_types; human action queue - WebSocket: parse game_type, auth_token, human_agent; route to werewolf or Sotopia sim - /games/queue, /games/leaderboard, /games/history, /memory stubs; /games/werewolf/config, POST /games/werewolf/sessions/create
- WebSocketHumanAgent: human-controlled agent consuming actions from queue; normalize vote/kill/lynch to action type - get_env_agents: explicit environment_messages fallback - WebSocketSotopiaSimulator: game_type, human_agent_index, action_queue_getter params
- WerewolfGameEndEvaluator, WerewolfActionHandler, WerewolfEnv - WebSocketWerewolfRunner: config load, human+LLM agents, run loop, pack chat - Persist GameResult, process ELO; run_werewolf_simulation entrypoint - Fix witch_have_poison typo in runner
- config: API_BASE_URL, API_ROUTES for backend and game endpoints
- auth-api: JWT storage, register/login/me/logout, OAuth login URL (/oauth/login/{provider})
- auth-api: getProfile, getEloHistory, getMatchHistory; getLeaderboard with limit/offset, backend-aligned types
- auth-api: handleOAuthTokenFromUrl, authRequest
- SimulationWebSocket: connect, startSimulation, sendClientMessage, on/off, disconnect - WSMessageType, WSMessage, SimulationStartPayload - generateSessionToken, getOrCreateSessionToken
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #
📑 Description
Adds the Sotopia Arena: a production-style web platform where humans play social games (starting with Werewolf) against LLM agents, to support research on human–AI social interaction (deception, coordination, trust in adversarial settings).
Scope of this PR:
scripts/ec2/(deploy, redeploy, Nginx, cache clear, etc.), Nginx reverse-proxy configThe branch is large by design (single feature set). Commits are organized so reviewers can follow the story (auth → backend → frontend → EC2). Happy to split into smaller follow-up PRs if the team prefers.
✅ Checks
type/description(e.g.feature/add-llm-agents)ℹ Additional Information