synth-ai 0.2.13.dev2__py3-none-any.whl → 0.2.16__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.


This version of synth-ai might be problematic. Click here for more details.

Files changed (293) hide show
  1. examples/README.md +1 -0
  2. examples/multi_step/SFT_README.md +147 -0
  3. examples/multi_step/configs/README_verilog_rl.md +77 -0
  4. examples/multi_step/configs/VERILOG_REWARDS.md +90 -0
  5. examples/multi_step/configs/VERILOG_RL_CHECKLIST.md +183 -0
  6. examples/multi_step/configs/crafter_eval_synth_qwen4b.toml +35 -0
  7. examples/multi_step/configs/crafter_eval_text_only_groq_qwen32b.toml +36 -0
  8. examples/multi_step/configs/crafter_rl_stepwise_hosted_judge.toml +12 -11
  9. examples/multi_step/configs/crafter_sft_qwen30b_lora.toml +62 -0
  10. examples/multi_step/configs/crafter_synth_backend.md +40 -0
  11. examples/multi_step/configs/verilog_eval_groq_qwen32b.toml +31 -0
  12. examples/multi_step/configs/verilog_eval_synth_qwen8b.toml +33 -0
  13. examples/multi_step/configs/verilog_rl_lora.toml +190 -0
  14. examples/multi_step/convert_traces_to_sft.py +84 -0
  15. examples/multi_step/judges/crafter_backend_judge.py +220 -0
  16. examples/multi_step/judges/verilog_backend_judge.py +234 -0
  17. examples/multi_step/readme.md +48 -0
  18. examples/multi_step/run_sft_qwen30b.sh +45 -0
  19. examples/multi_step/verilog_rl_lora.md +218 -0
  20. examples/qwen_coder/configs/coder_lora_30b.toml +3 -2
  21. examples/qwen_coder/configs/coder_lora_4b.toml +2 -1
  22. examples/qwen_coder/configs/coder_lora_small.toml +2 -1
  23. examples/qwen_vl/BUGS_AND_FIXES.md +232 -0
  24. examples/qwen_vl/IMAGE_VALIDATION_COMPLETE.md +271 -0
  25. examples/qwen_vl/IMAGE_VALIDATION_SUMMARY.md +260 -0
  26. examples/qwen_vl/INFERENCE_SFT_TESTS.md +412 -0
  27. examples/qwen_vl/NEXT_STEPS_2B.md +325 -0
  28. examples/qwen_vl/QUICKSTART.md +327 -0
  29. examples/qwen_vl/QUICKSTART_RL_VISION.md +110 -0
  30. examples/qwen_vl/README.md +154 -0
  31. examples/qwen_vl/RL_VISION_COMPLETE.md +475 -0
  32. examples/qwen_vl/RL_VISION_TESTING.md +333 -0
  33. examples/qwen_vl/SDK_VISION_INTEGRATION.md +328 -0
  34. examples/qwen_vl/SETUP_COMPLETE.md +275 -0
  35. examples/qwen_vl/VISION_TESTS_COMPLETE.md +490 -0
  36. examples/qwen_vl/VLM_PIPELINE_COMPLETE.md +242 -0
  37. examples/qwen_vl/__init__.py +2 -0
  38. examples/qwen_vl/collect_data_via_cli.md +423 -0
  39. examples/qwen_vl/collect_vision_traces.py +368 -0
  40. examples/qwen_vl/configs/crafter_rl_vision_qwen3vl4b.toml +127 -0
  41. examples/qwen_vl/configs/crafter_vlm_sft_example.toml +60 -0
  42. examples/qwen_vl/configs/eval_gpt4o_mini_vision.toml +43 -0
  43. examples/qwen_vl/configs/eval_gpt4o_vision_proper.toml +29 -0
  44. examples/qwen_vl/configs/eval_gpt5nano_vision.toml +45 -0
  45. examples/qwen_vl/configs/eval_qwen2vl_vision.toml +44 -0
  46. examples/qwen_vl/configs/filter_qwen2vl_sft.toml +50 -0
  47. examples/qwen_vl/configs/filter_vision_sft.toml +53 -0
  48. examples/qwen_vl/configs/filter_vision_test.toml +8 -0
  49. examples/qwen_vl/configs/sft_qwen3_vl_2b_test.toml +54 -0
  50. examples/qwen_vl/crafter_gpt5nano_agent.py +308 -0
  51. examples/qwen_vl/crafter_qwen_vl_agent.py +300 -0
  52. examples/qwen_vl/run_vision_comparison.sh +62 -0
  53. examples/qwen_vl/run_vision_sft_pipeline.sh +175 -0
  54. examples/qwen_vl/test_image_validation.py +201 -0
  55. examples/qwen_vl/test_sft_vision_data.py +110 -0
  56. examples/rl/README.md +1 -1
  57. examples/rl/configs/eval_base_qwen.toml +17 -0
  58. examples/rl/configs/eval_rl_qwen.toml +13 -0
  59. examples/rl/configs/rl_from_base_qwen.toml +37 -0
  60. examples/rl/configs/rl_from_base_qwen17.toml +76 -0
  61. examples/rl/configs/rl_from_ft_qwen.toml +37 -0
  62. examples/rl/run_eval.py +436 -0
  63. examples/rl/run_rl_and_save.py +111 -0
  64. examples/rl/task_app/README.md +22 -0
  65. examples/rl/task_app/math_single_step.py +990 -0
  66. examples/rl/task_app/math_task_app.py +111 -0
  67. examples/sft/README.md +5 -5
  68. examples/sft/configs/crafter_fft_qwen0p6b.toml +4 -2
  69. examples/sft/configs/crafter_lora_qwen0p6b.toml +4 -3
  70. examples/sft/evaluate.py +4 -4
  71. examples/sft/export_dataset.py +7 -4
  72. examples/sft/generate_traces.py +2 -0
  73. examples/swe/task_app/README.md +1 -1
  74. examples/swe/task_app/grpo_swe_mini.py +1 -1
  75. examples/swe/task_app/grpo_swe_mini_task_app.py +0 -12
  76. examples/swe/task_app/hosted/envs/mini_swe/environment.py +13 -13
  77. examples/swe/task_app/hosted/policy_routes.py +0 -2
  78. examples/swe/task_app/hosted/rollout.py +2 -8
  79. examples/task_apps/IMAGE_ONLY_EVAL_QUICKSTART.md +258 -0
  80. examples/task_apps/crafter/CREATE_SFT_DATASET.md +273 -0
  81. examples/task_apps/crafter/EVAL_IMAGE_ONLY_RESULTS.md +152 -0
  82. examples/task_apps/crafter/FILTER_COMMAND_STATUS.md +174 -0
  83. examples/task_apps/crafter/FILTER_COMMAND_SUCCESS.md +268 -0
  84. examples/task_apps/crafter/QUERY_EXAMPLES.md +203 -0
  85. examples/task_apps/crafter/README_IMAGE_ONLY_EVAL.md +316 -0
  86. examples/task_apps/crafter/eval_image_only_gpt4o.toml +28 -0
  87. examples/task_apps/crafter/eval_text_only_groq_llama.toml +36 -0
  88. examples/task_apps/crafter/filter_sft_dataset.toml +16 -0
  89. examples/task_apps/crafter/task_app/__init__.py +3 -0
  90. examples/task_apps/crafter/task_app/grpo_crafter.py +309 -14
  91. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/environment.py +10 -0
  92. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/policy.py +75 -4
  93. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/react_agent.py +17 -2
  94. examples/task_apps/crafter/task_app/synth_envs_hosted/inference/openai_client.py +55 -3
  95. examples/task_apps/crafter/task_app/synth_envs_hosted/policy_routes.py +114 -32
  96. examples/task_apps/crafter/task_app/synth_envs_hosted/rollout.py +127 -27
  97. examples/task_apps/crafter/task_app/synth_envs_hosted/utils.py +156 -0
  98. examples/task_apps/enron/__init__.py +1 -0
  99. examples/task_apps/enron/filter_sft.toml +5 -0
  100. examples/task_apps/enron/tests/__init__.py +2 -0
  101. examples/task_apps/enron/tests/integration/__init__.py +2 -0
  102. examples/task_apps/enron/tests/integration/test_enron_eval.py +2 -0
  103. examples/task_apps/enron/tests/unit/__init__.py +2 -0
  104. examples/task_apps/pokemon_red/EVAL_IMAGE_ONLY_COMPLETE.md +283 -0
  105. examples/task_apps/pokemon_red/EVAL_IMAGE_ONLY_STATUS.md +155 -0
  106. examples/task_apps/pokemon_red/README_IMAGE_ONLY_EVAL.md +415 -0
  107. examples/task_apps/pokemon_red/eval_image_only_gpt4o.toml +29 -0
  108. examples/task_apps/pokemon_red/pallet_town_rl_config.toml +2 -0
  109. examples/task_apps/pokemon_red/task_app.py +199 -6
  110. examples/task_apps/pokemon_red/test_pallet_town_rewards.py +2 -0
  111. examples/task_apps/sokoban/filter_sft.toml +5 -0
  112. examples/task_apps/sokoban/tests/__init__.py +2 -0
  113. examples/task_apps/sokoban/tests/integration/__init__.py +2 -0
  114. examples/task_apps/sokoban/tests/unit/__init__.py +2 -0
  115. examples/task_apps/verilog/eval_groq_qwen32b.toml +8 -4
  116. examples/task_apps/verilog/filter_sft.toml +5 -0
  117. examples/task_apps/verilog/task_app/grpo_verilog.py +258 -23
  118. examples/task_apps/verilog/tests/__init__.py +2 -0
  119. examples/task_apps/verilog/tests/integration/__init__.py +2 -0
  120. examples/task_apps/verilog/tests/integration/test_verilog_eval.py +2 -0
  121. examples/task_apps/verilog/tests/unit/__init__.py +2 -0
  122. examples/vlm/README.md +3 -3
  123. examples/vlm/configs/crafter_vlm_gpt4o.toml +2 -0
  124. examples/vlm/crafter_openai_vlm_agent.py +3 -5
  125. examples/vlm/filter_image_rows.py +1 -1
  126. examples/vlm/run_crafter_vlm_benchmark.py +2 -2
  127. examples/warming_up_to_rl/_utils.py +92 -0
  128. examples/warming_up_to_rl/analyze_trace_db.py +1 -1
  129. examples/warming_up_to_rl/configs/crafter_fft.toml +2 -0
  130. examples/warming_up_to_rl/configs/crafter_fft_4b.toml +2 -0
  131. examples/warming_up_to_rl/configs/eval_fft_qwen4b.toml +2 -0
  132. examples/warming_up_to_rl/configs/eval_groq_qwen32b.toml +2 -0
  133. examples/warming_up_to_rl/configs/eval_modal_qwen4b.toml +2 -1
  134. examples/warming_up_to_rl/configs/rl_from_base_qwen4b.toml +2 -1
  135. examples/warming_up_to_rl/configs/rl_from_ft.toml +2 -0
  136. examples/warming_up_to_rl/export_trace_sft.py +174 -60
  137. examples/warming_up_to_rl/groq_test.py +2 -0
  138. examples/warming_up_to_rl/readme.md +63 -132
  139. examples/warming_up_to_rl/run_fft_and_save.py +1 -1
  140. examples/warming_up_to_rl/run_local_rollout.py +2 -0
  141. examples/warming_up_to_rl/run_local_rollout_modal.py +2 -0
  142. examples/warming_up_to_rl/run_local_rollout_parallel.py +2 -0
  143. examples/warming_up_to_rl/run_local_rollout_traced.py +2 -0
  144. examples/warming_up_to_rl/run_rl_and_save.py +1 -1
  145. examples/warming_up_to_rl/run_rollout_remote.py +2 -0
  146. examples/warming_up_to_rl/task_app/README.md +42 -0
  147. examples/warming_up_to_rl/task_app/grpo_crafter.py +696 -0
  148. examples/warming_up_to_rl/task_app/grpo_crafter_task_app.py +135 -0
  149. examples/warming_up_to_rl/task_app/synth_envs_hosted/README.md +173 -0
  150. examples/warming_up_to_rl/task_app/synth_envs_hosted/__init__.py +5 -0
  151. examples/warming_up_to_rl/task_app/synth_envs_hosted/branching.py +143 -0
  152. examples/warming_up_to_rl/task_app/synth_envs_hosted/environment_routes.py +1226 -0
  153. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/__init__.py +1 -0
  154. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/__init__.py +6 -0
  155. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/app.py +1 -0
  156. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/environment.py +522 -0
  157. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/policy.py +478 -0
  158. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/react_agent.py +108 -0
  159. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/shared.py +305 -0
  160. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/tools.py +47 -0
  161. examples/warming_up_to_rl/task_app/synth_envs_hosted/hosted_app.py +204 -0
  162. examples/warming_up_to_rl/task_app/synth_envs_hosted/inference/__init__.py +5 -0
  163. examples/warming_up_to_rl/task_app/synth_envs_hosted/inference/openai_client.py +618 -0
  164. examples/warming_up_to_rl/task_app/synth_envs_hosted/main.py +100 -0
  165. examples/warming_up_to_rl/task_app/synth_envs_hosted/policy_routes.py +1081 -0
  166. examples/warming_up_to_rl/task_app/synth_envs_hosted/registry.py +195 -0
  167. examples/warming_up_to_rl/task_app/synth_envs_hosted/rollout.py +1861 -0
  168. examples/warming_up_to_rl/task_app/synth_envs_hosted/storage/__init__.py +5 -0
  169. examples/warming_up_to_rl/task_app/synth_envs_hosted/storage/volume.py +211 -0
  170. examples/warming_up_to_rl/task_app/synth_envs_hosted/test_agents.py +161 -0
  171. examples/warming_up_to_rl/task_app/synth_envs_hosted/test_service.py +137 -0
  172. examples/warming_up_to_rl/task_app/synth_envs_hosted/utils.py +62 -0
  173. synth_ai/__init__.py +44 -30
  174. synth_ai/_utils/__init__.py +47 -0
  175. synth_ai/_utils/base_url.py +10 -0
  176. synth_ai/_utils/http.py +10 -0
  177. synth_ai/_utils/prompts.py +10 -0
  178. synth_ai/_utils/task_app_state.py +12 -0
  179. synth_ai/_utils/user_config.py +10 -0
  180. synth_ai/api/models/supported.py +145 -7
  181. synth_ai/api/train/__init__.py +13 -1
  182. synth_ai/api/train/cli.py +30 -7
  183. synth_ai/api/train/config_finder.py +18 -11
  184. synth_ai/api/train/env_resolver.py +13 -10
  185. synth_ai/cli/__init__.py +66 -49
  186. synth_ai/cli/_modal_wrapper.py +9 -6
  187. synth_ai/cli/_typer_patch.py +0 -2
  188. synth_ai/cli/_validate_task_app.py +22 -4
  189. synth_ai/cli/legacy_root_backup.py +3 -1
  190. synth_ai/cli/lib/__init__.py +10 -0
  191. synth_ai/cli/lib/task_app_discovery.py +7 -0
  192. synth_ai/cli/lib/task_app_env.py +518 -0
  193. synth_ai/cli/recent.py +1 -0
  194. synth_ai/cli/setup.py +266 -0
  195. synth_ai/cli/task_app_deploy.py +16 -0
  196. synth_ai/cli/task_app_list.py +25 -0
  197. synth_ai/cli/task_app_modal_serve.py +16 -0
  198. synth_ai/cli/task_app_serve.py +18 -0
  199. synth_ai/cli/task_apps.py +392 -141
  200. synth_ai/cli/train.py +18 -0
  201. synth_ai/cli/tui.py +62 -0
  202. synth_ai/demos/__init__.py +10 -0
  203. synth_ai/demos/core/__init__.py +28 -1
  204. synth_ai/demos/crafter/__init__.py +1 -0
  205. synth_ai/demos/crafter/crafter_fft_4b.toml +55 -0
  206. synth_ai/demos/crafter/grpo_crafter_task_app.py +185 -0
  207. synth_ai/demos/crafter/rl_from_base_qwen4b.toml +74 -0
  208. synth_ai/demos/demo_registry.py +176 -0
  209. synth_ai/demos/demo_task_apps/crafter/grpo_crafter_task_app.py +1 -1
  210. synth_ai/demos/math/__init__.py +1 -0
  211. synth_ai/demos/math/_common.py +16 -0
  212. synth_ai/demos/math/app.py +38 -0
  213. synth_ai/demos/math/config.toml +76 -0
  214. synth_ai/demos/math/deploy_modal.py +54 -0
  215. synth_ai/demos/math/modal_task_app.py +702 -0
  216. synth_ai/demos/math/task_app_entry.py +51 -0
  217. synth_ai/environments/environment/core.py +7 -1
  218. synth_ai/environments/examples/bandit/engine.py +0 -1
  219. synth_ai/environments/examples/bandit/environment.py +0 -1
  220. synth_ai/environments/examples/crafter_classic/environment.py +1 -1
  221. synth_ai/environments/examples/verilog/engine.py +76 -10
  222. synth_ai/environments/examples/wordle/environment.py +0 -1
  223. synth_ai/evals/base.py +16 -5
  224. synth_ai/evals/client.py +1 -1
  225. synth_ai/inference/client.py +1 -1
  226. synth_ai/learning/client.py +1 -1
  227. synth_ai/learning/health.py +1 -1
  228. synth_ai/learning/jobs.py +1 -1
  229. synth_ai/learning/rl/client.py +1 -1
  230. synth_ai/learning/rl/env_keys.py +1 -1
  231. synth_ai/learning/rl/secrets.py +1 -1
  232. synth_ai/learning/sft/client.py +1 -1
  233. synth_ai/learning/sft/data.py +407 -4
  234. synth_ai/learning/validators.py +4 -1
  235. synth_ai/task/__init__.py +11 -1
  236. synth_ai/task/apps/__init__.py +5 -2
  237. synth_ai/task/config.py +259 -0
  238. synth_ai/task/contracts.py +15 -2
  239. synth_ai/task/rubrics/__init__.py +4 -2
  240. synth_ai/task/rubrics/loaders.py +27 -4
  241. synth_ai/task/rubrics/scoring.py +3 -0
  242. synth_ai/task/rubrics.py +219 -0
  243. synth_ai/task/trace_correlation_helpers.py +328 -0
  244. synth_ai/task/tracing_utils.py +14 -3
  245. synth_ai/task/validators.py +145 -2
  246. synth_ai/tracing_v3/config.py +15 -13
  247. synth_ai/tracing_v3/constants.py +21 -0
  248. synth_ai/tracing_v3/db_config.py +3 -1
  249. synth_ai/tracing_v3/decorators.py +10 -7
  250. synth_ai/tracing_v3/session_tracer.py +10 -0
  251. synth_ai/tracing_v3/turso/daemon.py +2 -2
  252. synth_ai/tracing_v3/turso/native_manager.py +108 -77
  253. synth_ai/tracing_v3/utils.py +1 -1
  254. synth_ai/tui/__init__.py +5 -0
  255. synth_ai/tui/__main__.py +13 -0
  256. synth_ai/tui/cli/__init__.py +1 -0
  257. synth_ai/tui/cli/query_experiments.py +164 -0
  258. synth_ai/tui/cli/query_experiments_v3.py +164 -0
  259. synth_ai/tui/dashboard.py +911 -0
  260. synth_ai/utils/__init__.py +101 -0
  261. synth_ai/utils/base_url.py +94 -0
  262. synth_ai/utils/cli.py +131 -0
  263. synth_ai/utils/env.py +287 -0
  264. synth_ai/utils/http.py +169 -0
  265. synth_ai/utils/modal.py +308 -0
  266. synth_ai/utils/process.py +212 -0
  267. synth_ai/utils/prompts.py +39 -0
  268. synth_ai/utils/sqld.py +122 -0
  269. synth_ai/utils/task_app_discovery.py +882 -0
  270. synth_ai/utils/task_app_env.py +186 -0
  271. synth_ai/utils/task_app_state.py +318 -0
  272. synth_ai/utils/user_config.py +137 -0
  273. synth_ai/v0/config/__init__.py +1 -5
  274. synth_ai/v0/config/base_url.py +1 -7
  275. synth_ai/v0/tracing/config.py +1 -1
  276. synth_ai/v0/tracing/decorators.py +1 -1
  277. synth_ai/v0/tracing/upload.py +1 -1
  278. synth_ai/v0/tracing_v1/config.py +1 -1
  279. synth_ai/v0/tracing_v1/decorators.py +1 -1
  280. synth_ai/v0/tracing_v1/upload.py +1 -1
  281. {synth_ai-0.2.13.dev2.dist-info → synth_ai-0.2.16.dist-info}/METADATA +85 -31
  282. {synth_ai-0.2.13.dev2.dist-info → synth_ai-0.2.16.dist-info}/RECORD +286 -135
  283. synth_ai/cli/man.py +0 -106
  284. synth_ai/compound/cais.py +0 -0
  285. synth_ai/core/experiment.py +0 -13
  286. synth_ai/core/system.py +0 -15
  287. synth_ai/demo_registry.py +0 -295
  288. synth_ai/handshake.py +0 -109
  289. synth_ai/http.py +0 -26
  290. {synth_ai-0.2.13.dev2.dist-info → synth_ai-0.2.16.dist-info}/WHEEL +0 -0
  291. {synth_ai-0.2.13.dev2.dist-info → synth_ai-0.2.16.dist-info}/entry_points.txt +0 -0
  292. {synth_ai-0.2.13.dev2.dist-info → synth_ai-0.2.16.dist-info}/licenses/LICENSE +0 -0
  293. {synth_ai-0.2.13.dev2.dist-info → synth_ai-0.2.16.dist-info}/top_level.txt +0 -0
@@ -0,0 +1,195 @@
1
+ from __future__ import annotations
2
+
3
+ import uuid
4
+ from dataclasses import dataclass, field
5
+ from datetime import datetime
6
+ from typing import Any
7
+
8
+
9
+ @dataclass
10
+ class EnvHandle:
11
+ """In-memory handle for an active environment instance."""
12
+
13
+ env_id: str
14
+ env: Any # StatefulEnvironment or wrapper
15
+ last_observation: dict[str, Any] | None
16
+ last_info: dict[str, Any] | None
17
+ step_idx: int
18
+ seed: int | None
19
+ rl_run_id: str
20
+ created_at: datetime = field(default_factory=datetime.utcnow)
21
+
22
+
23
+ @dataclass
24
+ class PolicyHandle:
25
+ """In-memory handle for an active policy instance."""
26
+
27
+ policy_id: str
28
+ policy: Any # Policy instance
29
+ bound_env_id: str | None
30
+ rl_run_id: str
31
+ created_at: datetime = field(default_factory=datetime.utcnow)
32
+
33
+
34
+ @dataclass
35
+ class RunHandle:
36
+ """Track run status for abort support."""
37
+
38
+ run_id: str
39
+ status: str # "running" | "aborted" | "completed"
40
+ started_at: datetime
41
+ finished_at: datetime | None = None
42
+
43
+
44
+ @dataclass
45
+ class SnapshotMeta:
46
+ """Metadata for a stored snapshot."""
47
+
48
+ snapshot_id: str
49
+ kind: str # "env" | "policy"
50
+ rl_run_id: str
51
+ parent_snapshot_id: str | None
52
+ size: int
53
+ created_at: datetime
54
+ path: str
55
+
56
+
57
+ class Registry:
58
+ """In-memory registries for the service."""
59
+
60
+ def __init__(self) -> None:
61
+ self.envs: dict[str, EnvHandle] = {}
62
+ self.policies: dict[str, PolicyHandle] = {}
63
+ self.runs: dict[str, RunHandle] = {}
64
+ self.snapshots: dict[str, SnapshotMeta] = {}
65
+
66
+ def generate_id(self) -> str:
67
+ """Generate a UUID for unique identification."""
68
+ return str(uuid.uuid4())
69
+
70
+ def register_env(
71
+ self,
72
+ env: Any,
73
+ seed: int | None,
74
+ rl_run_id: str,
75
+ last_observation: dict[str, Any] | None = None,
76
+ last_info: dict[str, Any] | None = None,
77
+ ) -> str:
78
+ """Register a new environment instance."""
79
+ env_id = self.generate_id()
80
+ handle = EnvHandle(
81
+ env_id=env_id,
82
+ env=env,
83
+ last_observation=last_observation,
84
+ last_info=last_info,
85
+ step_idx=0,
86
+ seed=seed,
87
+ rl_run_id=rl_run_id,
88
+ )
89
+ self.envs[env_id] = handle
90
+ return env_id
91
+
92
+ def register_policy(
93
+ self,
94
+ policy: Any,
95
+ rl_run_id: str,
96
+ bound_env_id: str | None = None,
97
+ ) -> str:
98
+ """Register a new policy instance."""
99
+ policy_id = self.generate_id()
100
+ handle = PolicyHandle(
101
+ policy_id=policy_id,
102
+ policy=policy,
103
+ bound_env_id=bound_env_id,
104
+ rl_run_id=rl_run_id,
105
+ )
106
+ self.policies[policy_id] = handle
107
+ return policy_id
108
+
109
+ def register_run(self, run_id: str | None = None) -> str:
110
+ """Register a new run."""
111
+ if run_id is None:
112
+ run_id = self.generate_id()
113
+ handle = RunHandle(
114
+ run_id=run_id,
115
+ status="running",
116
+ started_at=datetime.utcnow(),
117
+ )
118
+ self.runs[run_id] = handle
119
+ return run_id
120
+
121
+ def abort_run(self, run_id: str) -> bool:
122
+ """Mark a run as aborted."""
123
+ if run_id in self.runs:
124
+ self.runs[run_id].status = "aborted"
125
+ self.runs[run_id].finished_at = datetime.utcnow()
126
+ return True
127
+ return False
128
+
129
+ def complete_run(self, run_id: str) -> bool:
130
+ """Mark a run as completed."""
131
+ if run_id in self.runs:
132
+ self.runs[run_id].status = "completed"
133
+ self.runs[run_id].finished_at = datetime.utcnow()
134
+ return True
135
+ return False
136
+
137
+ def is_run_aborted(self, run_id: str) -> bool:
138
+ """Check if a run has been aborted."""
139
+ return run_id in self.runs and self.runs[run_id].status == "aborted"
140
+
141
+ def register_snapshot(
142
+ self,
143
+ kind: str,
144
+ rl_run_id: str,
145
+ size: int,
146
+ path: str,
147
+ parent_snapshot_id: str | None = None,
148
+ ) -> str:
149
+ """Register a new snapshot."""
150
+ snapshot_id = self.generate_id()
151
+ meta = SnapshotMeta(
152
+ snapshot_id=snapshot_id,
153
+ kind=kind,
154
+ rl_run_id=rl_run_id,
155
+ parent_snapshot_id=parent_snapshot_id,
156
+ size=size,
157
+ created_at=datetime.utcnow(),
158
+ path=path,
159
+ )
160
+ self.snapshots[snapshot_id] = meta
161
+ return snapshot_id
162
+
163
+ def get_env(self, env_id: str) -> EnvHandle | None:
164
+ """Get an environment handle by ID."""
165
+ return self.envs.get(env_id)
166
+
167
+ def get_policy(self, policy_id: str) -> PolicyHandle | None:
168
+ """Get a policy handle by ID."""
169
+ return self.policies.get(policy_id)
170
+
171
+ def get_run(self, run_id: str) -> RunHandle | None:
172
+ """Get a run handle by ID."""
173
+ return self.runs.get(run_id)
174
+
175
+ def get_snapshot(self, snapshot_id: str) -> SnapshotMeta | None:
176
+ """Get snapshot metadata by ID."""
177
+ return self.snapshots.get(snapshot_id)
178
+
179
+ def remove_env(self, env_id: str) -> bool:
180
+ """Remove an environment from the registry."""
181
+ if env_id in self.envs:
182
+ del self.envs[env_id]
183
+ return True
184
+ return False
185
+
186
+ def remove_policy(self, policy_id: str) -> bool:
187
+ """Remove a policy from the registry."""
188
+ if policy_id in self.policies:
189
+ del self.policies[policy_id]
190
+ return True
191
+ return False
192
+
193
+
194
+ # Global registry instance
195
+ registry = Registry()