Spaces:
Running
Running
Commit History
-
feat: implement episode finalization and adjust session support flag
ba9b704
-
fix: properly handle task_id in reset endpoint
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>7c50ec6 -
fix: add root endpoint and API documentation
- Add GET / endpoint with welcome message and API info - Enable /docs (Swagger UI) and /redoc endpoints - Provide usage instructions in root response Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>4938245 -
refactor: unify data format and add stateful HTTP server
Data Format Changes: - Replace corpus.json/tasks.json with per-seed JSON files - Rename fields: gold_answer→truth, gold_chunk_ids→supporting_items, difficulty→level - Add SupportingItem and DistractorItem models with verification fields - Add sample/ directory with mock data matching generator output format Training Module Updates: - Change difficulty (easy/medium/hard) to level (0/1/2) throughout - Update CurriculumScheduler to use integer levels - Update metrics to use success_rate_by_level dict - Update evaluation to group by level instead of difficulty Server Changes: - Replace stateless openenv server with custom stateful FastAPI server - Environment state now persists across HTTP requests - Add /tasks endpoint to list available tasks Production Data Loading: - Add load_tasks_from_directory() for per-seed file loading - Add load_verified_tasks() and load_tasks_by_level() helpers - Add get_directory_statistics() for production data analysis All 76 tests pass. Co-Authored-By: C…79b7d56 -
remove: delete server/requirements.txt (using uv)
ad7575d -
remove: delete server/tasks/ (moved to data/)
914a8af -
remove: delete old server/retrieval.py and rewards.py
da927c9 -
fix: update app imports for new package structure
178f9f9 -
refactor: update server to use searcharena package
421f883
-
refactor: update type hints to use built-in generic types
e597175 -
Refactor inference and logging mechanisms; remove log_utils.py
- Simplified logging functions in inference.py, replacing the Logger class with standalone functions for logging start, step, and end events. - Removed log_utils.py as its functionality was integrated directly into inference.py. - Updated Config class to include a benchmark field and adjusted environment variable handling. - Enhanced error handling in run_episode and get_action functions. - Streamlined the SearchEnvironment class by removing unused result classes and simplifying return structures. - Improved document and task loading with caching mechanisms in tasks module. - Updated tests to reflect changes in logging and environment behavior, removing unnecessary tests for stepping after episode completion.dfe9450
-
refactor: update environment variables and reward validation range
5cb1be7
-
updated readme and project.toml
a3dd1cb -
Add comprehensive tests for Search RL Environment
- Introduced shared pytest fixtures in `conftest.py` for reusable test setups. - Created edge case tests in `test_edge_cases.py` to validate search, read, prune, and answer actions. - Implemented environment functionality tests in `test_environment.py` to ensure correct reset and step behavior. - Developed retrieval tests in `test_retrieval.py` for BM25 index and document corpus functionality. - Added reward calculation tests in `test_rewards.py` to verify reward mechanisms and beta scheduling. - Included task loading and difficulty tests in `test_tasks.py` to ensure task integrity and diversity.b765bdb
-
refactor: remove Serper backend support and related configurations
cdba309
-
added support for the serper api
51d4cae
-
enhance reward calculation with F-beta weight and trajectory floor adjustments
1df4090 -
added rewards mechanisms
96b86e0 -
wiring up the search environment loop
0176e87 -
added BM25 retrieval layer
8550b16 -
initial commit
5d108aa