synth-ai 0.2.8.dev4__py3-none-any.whl → 0.2.23.dev3__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.
Files changed (889) hide show
  1. examples/README.md +1 -0
  2. examples/__init__.py +16 -0
  3. examples/analyze_semantic_words.sh +17 -0
  4. examples/baseline/banking77_baseline.py +243 -0
  5. examples/baseline/banking77_pipeline_baseline.py +294 -0
  6. examples/baseline/crafter_baseline.py +407 -0
  7. examples/baseline/pokemon_red_baseline.py +326 -0
  8. examples/baseline/simple_baseline.py +56 -0
  9. examples/baseline/warming_up_to_rl_baseline.py +239 -0
  10. examples/blog_posts/gepa/README.md +355 -0
  11. examples/blog_posts/gepa/configs/banking77_gepa_local.toml +95 -0
  12. examples/blog_posts/gepa/configs/banking77_gepa_test.toml +80 -0
  13. examples/blog_posts/gepa/configs/banking77_mipro_local.toml +50 -0
  14. examples/blog_posts/gepa/configs/banking77_pipeline_gepa_local.toml +101 -0
  15. examples/blog_posts/gepa/configs/banking77_pipeline_gepa_test.toml +96 -0
  16. examples/blog_posts/gepa/configs/hotpotqa_gepa_local.toml +57 -0
  17. examples/blog_posts/gepa/configs/hotpotqa_gepa_qwen.toml +35 -0
  18. examples/blog_posts/gepa/configs/hotpotqa_mipro_local.toml +51 -0
  19. examples/blog_posts/gepa/configs/hover_gepa_local.toml +57 -0
  20. examples/blog_posts/gepa/configs/hover_gepa_qwen.toml +35 -0
  21. examples/blog_posts/gepa/configs/hover_mipro_local.toml +51 -0
  22. examples/blog_posts/gepa/configs/ifbench_gepa_local.toml +57 -0
  23. examples/blog_posts/gepa/configs/ifbench_gepa_qwen.toml +35 -0
  24. examples/blog_posts/gepa/configs/ifbench_mipro_local.toml +51 -0
  25. examples/blog_posts/gepa/configs/pupa_gepa_local.toml +58 -0
  26. examples/blog_posts/gepa/configs/pupa_mipro_local.toml +52 -0
  27. examples/blog_posts/gepa/deploy_banking77_task_app.sh +54 -0
  28. examples/blog_posts/gepa/gepa_baseline.py +204 -0
  29. examples/blog_posts/gepa/query_prompts_example.py +97 -0
  30. examples/blog_posts/gepa/run_gepa_banking77.sh +112 -0
  31. examples/blog_posts/gepa/run_gepa_banking77_pipeline.sh +163 -0
  32. examples/blog_posts/gepa/task_apps.py +105 -0
  33. examples/blog_posts/gepa/test_gepa_local.sh +67 -0
  34. examples/blog_posts/gepa/verify_banking77_setup.sh +123 -0
  35. examples/blog_posts/mipro/README.md +415 -0
  36. examples/blog_posts/mipro/configs/banking77_mipro_local.toml +91 -0
  37. examples/blog_posts/mipro/configs/banking77_mipro_test.toml +87 -0
  38. examples/blog_posts/mipro/configs/banking77_pipeline_mipro_gemini_flash_lite_local.toml +98 -0
  39. examples/blog_posts/mipro/configs/banking77_pipeline_mipro_gpt41mini_local.toml +96 -0
  40. examples/blog_posts/mipro/configs/banking77_pipeline_mipro_local.toml +94 -0
  41. examples/blog_posts/mipro/configs/banking77_pipeline_mipro_test.toml +170 -0
  42. examples/blog_posts/mipro/deploy_banking77_pipeline_task_app.sh +59 -0
  43. examples/blog_posts/mipro/deploy_banking77_task_app.sh +41 -0
  44. examples/blog_posts/mipro/multi_step.md +79 -0
  45. examples/blog_posts/mipro/run_mipro_banking77.sh +191 -0
  46. examples/blog_posts/mipro/run_mipro_banking77_pipeline.sh +171 -0
  47. examples/blog_posts/mipro/run_mipro_banking77_pipeline_gemini_flash_lite.sh +177 -0
  48. examples/blog_posts/mipro/run_mipro_banking77_pipeline_gpt41mini.sh +173 -0
  49. examples/blog_posts/mipro/verify_banking77_setup.sh +117 -0
  50. examples/blog_posts/pokemon_vl/README.md +98 -0
  51. examples/blog_posts/pokemon_vl/configs/eval_gpt5nano.toml +26 -0
  52. examples/blog_posts/pokemon_vl/configs/eval_qwen3_vl.toml +27 -0
  53. examples/blog_posts/pokemon_vl/configs/eval_rl_final.toml +24 -0
  54. examples/blog_posts/pokemon_vl/configs/filter_high_reward.toml +10 -0
  55. examples/blog_posts/pokemon_vl/configs/train_rl_from_sft.toml +43 -0
  56. examples/blog_posts/pokemon_vl/configs/train_sft_qwen4b_vl.toml +40 -0
  57. examples/blog_posts/pokemon_vl/extract_images.py +239 -0
  58. examples/blog_posts/pokemon_vl/pokemon_vl_baseline.py +326 -0
  59. examples/blog_posts/pokemon_vl/run_eval_extract_images.py +209 -0
  60. examples/blog_posts/pokemon_vl/run_qwen_eval_extract_images.py +212 -0
  61. examples/blog_posts/pokemon_vl/text_box_analysis.md +106 -0
  62. examples/blog_posts/warming_up_to_rl/ARCHITECTURE.md +195 -0
  63. examples/blog_posts/warming_up_to_rl/FINAL_TEST_RESULTS.md +127 -0
  64. examples/blog_posts/warming_up_to_rl/INFERENCE_SUCCESS.md +132 -0
  65. examples/blog_posts/warming_up_to_rl/README.md +158 -0
  66. examples/blog_posts/warming_up_to_rl/SMOKE_TESTING.md +164 -0
  67. examples/blog_posts/warming_up_to_rl/SMOKE_TEST_COMPLETE.md +253 -0
  68. examples/blog_posts/warming_up_to_rl/configs/eval_baseline_qwen32b_10x20.toml +25 -0
  69. examples/blog_posts/warming_up_to_rl/configs/eval_ft_qwen4b.toml +25 -0
  70. examples/blog_posts/warming_up_to_rl/configs/eval_ft_qwen4b_10x20.toml +26 -0
  71. examples/blog_posts/warming_up_to_rl/configs/eval_groq_qwen32b.toml +25 -0
  72. examples/blog_posts/warming_up_to_rl/configs/eval_openai_gpt_oss_120b.toml +29 -0
  73. examples/blog_posts/warming_up_to_rl/configs/filter_high_reward_dataset.toml +10 -0
  74. examples/blog_posts/warming_up_to_rl/configs/smoke_test.toml +75 -0
  75. examples/blog_posts/warming_up_to_rl/configs/train_rl_from_sft.toml +91 -0
  76. examples/blog_posts/warming_up_to_rl/configs/train_sft_qwen4b.toml +40 -0
  77. examples/blog_posts/warming_up_to_rl/warming_up_to_rl_baseline.py +187 -0
  78. examples/crafter_debug_render.py +186 -0
  79. examples/dev/qwen3_32b_qlora_4xh100.toml +45 -0
  80. examples/gepa/banking77_pipeline_gepa.toml +96 -0
  81. examples/gepa/multi_stage_gepa_example.toml +84 -0
  82. examples/gepa/run_gepa_banking77_pipeline.sh +157 -0
  83. examples/multi_step/SFT_README.md +147 -0
  84. examples/multi_step/configs/README_verilog_rl.md +77 -0
  85. examples/multi_step/configs/VERILOG_REWARDS.md +103 -0
  86. examples/multi_step/configs/VERILOG_RL_CHECKLIST.md +196 -0
  87. examples/multi_step/configs/crafter_eval_synth_qwen4b.toml +35 -0
  88. examples/multi_step/configs/crafter_eval_text_only_groq_qwen32b.toml +36 -0
  89. examples/multi_step/configs/crafter_rl_outcome.toml +75 -0
  90. examples/multi_step/configs/crafter_rl_stepwise_hosted_judge.toml +145 -0
  91. examples/multi_step/configs/crafter_rl_stepwise_shaped.toml +84 -0
  92. examples/multi_step/configs/crafter_rl_stepwise_simple.toml +79 -0
  93. examples/multi_step/configs/crafter_rl_stepwise_simple_NEW_FORMAT.toml +105 -0
  94. examples/multi_step/configs/crafter_sft_qwen30b_lora.toml +62 -0
  95. examples/multi_step/configs/crafter_synth_backend.md +40 -0
  96. examples/multi_step/configs/verilog_eval_groq_qwen32b.toml +31 -0
  97. examples/multi_step/configs/verilog_eval_synth_qwen8b.toml +33 -0
  98. examples/multi_step/configs/verilog_rl_lora.toml +147 -0
  99. examples/multi_step/convert_traces_to_sft.py +84 -0
  100. examples/multi_step/crafter_rl_lora.md +70 -0
  101. examples/multi_step/judges/crafter_backend_judge.py +220 -0
  102. examples/multi_step/judges/verilog_backend_judge.py +234 -0
  103. examples/multi_step/readme.md +48 -0
  104. examples/multi_step/run_sft_qwen30b.sh +45 -0
  105. examples/multi_step/sse_metrics_streaming_notes.md +357 -0
  106. examples/multi_step/task_app_config_notes.md +494 -0
  107. examples/multi_step/verilog_rl_lora.md +218 -0
  108. examples/qwen_coder/README.md +102 -0
  109. examples/qwen_coder/_shared.py +113 -0
  110. examples/qwen_coder/configs/coder_lora_30b.toml +60 -0
  111. examples/qwen_coder/configs/coder_lora_4b.toml +61 -0
  112. examples/qwen_coder/configs/coder_lora_small.toml +57 -0
  113. examples/qwen_coder/generate_dataset.py +98 -0
  114. examples/qwen_coder/infer_ft_smoke.py +65 -0
  115. examples/qwen_coder/infer_prod_proxy.py +73 -0
  116. examples/qwen_coder/infer_via_synth.py +87 -0
  117. examples/qwen_coder/scripts/infer_coder.sh +19 -0
  118. examples/qwen_coder/scripts/train_coder_30b.sh +22 -0
  119. examples/qwen_coder/sft_full_17b.py +103 -0
  120. examples/qwen_coder/sft_lora_30b.py +110 -0
  121. examples/qwen_coder/subset_jsonl.py +39 -0
  122. examples/qwen_coder/todos.md +38 -0
  123. examples/qwen_coder/validate_jsonl.py +60 -0
  124. examples/qwen_vl/BUGS_AND_FIXES.md +232 -0
  125. examples/qwen_vl/IMAGE_VALIDATION_COMPLETE.md +271 -0
  126. examples/qwen_vl/IMAGE_VALIDATION_SUMMARY.md +260 -0
  127. examples/qwen_vl/INFERENCE_SFT_TESTS.md +412 -0
  128. examples/qwen_vl/NEXT_STEPS_2B.md +325 -0
  129. examples/qwen_vl/QUICKSTART.md +327 -0
  130. examples/qwen_vl/QUICKSTART_RL_VISION.md +110 -0
  131. examples/qwen_vl/README.md +152 -0
  132. examples/qwen_vl/RL_VISION_COMPLETE.md +475 -0
  133. examples/qwen_vl/RL_VISION_TESTING.md +333 -0
  134. examples/qwen_vl/SDK_VISION_INTEGRATION.md +328 -0
  135. examples/qwen_vl/SETUP_COMPLETE.md +274 -0
  136. examples/qwen_vl/VISION_TESTS_COMPLETE.md +489 -0
  137. examples/qwen_vl/VLM_PIPELINE_COMPLETE.md +242 -0
  138. examples/qwen_vl/__init__.py +2 -0
  139. examples/qwen_vl/collect_data_via_cli.md +415 -0
  140. examples/qwen_vl/collect_vision_traces.py +368 -0
  141. examples/qwen_vl/configs/crafter_rl_vision_qwen3vl4b.toml +110 -0
  142. examples/qwen_vl/configs/crafter_vlm_sft_example.toml +59 -0
  143. examples/qwen_vl/configs/eval_gpt4o_mini_vision.toml +26 -0
  144. examples/qwen_vl/configs/eval_gpt4o_vision_proper.toml +29 -0
  145. examples/qwen_vl/configs/eval_gpt5nano_vision.toml +26 -0
  146. examples/qwen_vl/configs/eval_qwen3vl_vision.toml +26 -0
  147. examples/qwen_vl/configs/filter_qwen3vl_sft.toml +49 -0
  148. examples/qwen_vl/configs/filter_vision_sft.toml +52 -0
  149. examples/qwen_vl/configs/filter_vision_test.toml +8 -0
  150. examples/qwen_vl/configs/sft_qwen3_vl_2b_test.toml +54 -0
  151. examples/qwen_vl/crafter_gpt5nano_agent.py +308 -0
  152. examples/qwen_vl/crafter_qwen_vl_agent.py +300 -0
  153. examples/qwen_vl/run_vision_comparison.sh +61 -0
  154. examples/qwen_vl/run_vision_sft_pipeline.sh +175 -0
  155. examples/qwen_vl/test_image_validation.py +201 -0
  156. examples/qwen_vl/test_sft_vision_data.py +110 -0
  157. examples/rl/README.md +169 -0
  158. examples/rl/configs/eval_base_qwen.toml +17 -0
  159. examples/rl/configs/eval_rl_qwen.toml +13 -0
  160. examples/rl/configs/rl_from_base_qwen.toml +62 -0
  161. examples/rl/configs/rl_from_base_qwen17.toml +80 -0
  162. examples/rl/configs/rl_from_ft_qwen.toml +37 -0
  163. examples/rl/download_dataset.py +80 -0
  164. examples/rl/run_eval.py +436 -0
  165. examples/rl/run_rl_and_save.py +111 -0
  166. examples/rl/task_app/README.md +21 -0
  167. examples/rl/task_app/math_single_step.py +990 -0
  168. examples/rl/task_app/math_task_app.py +111 -0
  169. examples/run_crafter_demo.sh +10 -0
  170. examples/sdk_prompt_learning_example.py +55 -0
  171. examples/sft/README.md +139 -0
  172. examples/sft/configs/crafter_fft_qwen0p6b.toml +49 -0
  173. examples/sft/configs/crafter_lora_qwen0p6b.toml +49 -0
  174. examples/sft/evaluate.py +117 -0
  175. examples/sft/export_dataset.py +120 -0
  176. examples/sft/generate_traces.py +164 -0
  177. examples/swe/__init__.py +12 -0
  178. examples/swe/task_app/README.md +135 -0
  179. examples/swe/task_app/__init__.py +2 -0
  180. examples/swe/task_app/grpo_swe_mini.py +604 -0
  181. examples/swe/task_app/grpo_swe_mini_task_app.py +124 -0
  182. examples/swe/task_app/hosted/README.md +173 -0
  183. examples/swe/task_app/hosted/__init__.py +5 -0
  184. examples/swe/task_app/hosted/branching.py +143 -0
  185. examples/swe/task_app/hosted/environment_routes.py +1289 -0
  186. examples/swe/task_app/hosted/envs/__init__.py +1 -0
  187. examples/swe/task_app/hosted/envs/crafter/__init__.py +6 -0
  188. examples/swe/task_app/hosted/envs/crafter/app.py +1 -0
  189. examples/swe/task_app/hosted/envs/crafter/environment.py +522 -0
  190. examples/swe/task_app/hosted/envs/crafter/policy.py +478 -0
  191. examples/swe/task_app/hosted/envs/crafter/react_agent.py +108 -0
  192. examples/swe/task_app/hosted/envs/crafter/shared.py +305 -0
  193. examples/swe/task_app/hosted/envs/crafter/tools.py +47 -0
  194. examples/swe/task_app/hosted/envs/mini_swe/__init__.py +8 -0
  195. examples/swe/task_app/hosted/envs/mini_swe/environment.py +1191 -0
  196. examples/swe/task_app/hosted/envs/mini_swe/policy.py +355 -0
  197. examples/swe/task_app/hosted/envs/mini_swe/shared.py +83 -0
  198. examples/swe/task_app/hosted/envs/mini_swe/tools.py +96 -0
  199. examples/swe/task_app/hosted/hosted_app.py +204 -0
  200. examples/swe/task_app/hosted/inference/__init__.py +5 -0
  201. examples/swe/task_app/hosted/inference/openai_client.py +584 -0
  202. examples/swe/task_app/hosted/main.py +100 -0
  203. examples/swe/task_app/hosted/policy_routes.py +1094 -0
  204. examples/swe/task_app/hosted/registry.py +195 -0
  205. examples/swe/task_app/hosted/rollout.py +1905 -0
  206. examples/swe/task_app/hosted/storage/__init__.py +5 -0
  207. examples/swe/task_app/hosted/storage/volume.py +211 -0
  208. examples/swe/task_app/hosted/test_agents.py +161 -0
  209. examples/swe/task_app/hosted/test_service.py +136 -0
  210. examples/swe/task_app/hosted/utils.py +62 -0
  211. examples/swe/task_app/morph_backend.py +178 -0
  212. examples/task_apps/IMAGE_ONLY_EVAL_QUICKSTART.md +258 -0
  213. examples/task_apps/TESTING.md +275 -0
  214. examples/task_apps/banking77/__init__.py +6 -0
  215. examples/task_apps/banking77/banking77_task_app.py +912 -0
  216. examples/task_apps/banking77/deploy_wrapper.py +46 -0
  217. examples/task_apps/banking77_pipeline/__init__.py +6 -0
  218. examples/task_apps/banking77_pipeline/banking77_pipeline_task_app.py +489 -0
  219. examples/task_apps/banking77_pipeline/deploy_wrapper.py +50 -0
  220. examples/task_apps/crafter/CREATE_SFT_DATASET.md +286 -0
  221. examples/task_apps/crafter/EVAL_IMAGE_ONLY_RESULTS.md +152 -0
  222. examples/task_apps/crafter/FILTER_COMMAND_STATUS.md +187 -0
  223. examples/task_apps/crafter/FILTER_COMMAND_SUCCESS.md +281 -0
  224. examples/task_apps/crafter/QUERY_EXAMPLES.md +203 -0
  225. examples/task_apps/crafter/README_IMAGE_ONLY_EVAL.md +316 -0
  226. examples/task_apps/crafter/eval_image_only_gpt4o.toml +28 -0
  227. examples/task_apps/crafter/eval_text_only_groq_llama.toml +36 -0
  228. examples/task_apps/crafter/filter_sft_dataset.toml +16 -0
  229. examples/task_apps/crafter/task_app/README.md +42 -0
  230. examples/task_apps/crafter/task_app/__init__.py +5 -0
  231. examples/task_apps/crafter/task_app/grpo_crafter.py +1055 -0
  232. examples/task_apps/crafter/task_app/grpo_crafter_task_app.py +146 -0
  233. examples/task_apps/crafter/task_app/synth_envs_hosted/README.md +173 -0
  234. examples/task_apps/crafter/task_app/synth_envs_hosted/__init__.py +5 -0
  235. examples/task_apps/crafter/task_app/synth_envs_hosted/branching.py +143 -0
  236. examples/task_apps/crafter/task_app/synth_envs_hosted/environment_routes.py +1226 -0
  237. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/__init__.py +1 -0
  238. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/__init__.py +6 -0
  239. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/app.py +1 -0
  240. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/environment.py +532 -0
  241. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/policy.py +583 -0
  242. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/react_agent.py +122 -0
  243. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/shared.py +305 -0
  244. examples/task_apps/crafter/task_app/synth_envs_hosted/envs/crafter/tools.py +47 -0
  245. examples/task_apps/crafter/task_app/synth_envs_hosted/hosted_app.py +253 -0
  246. examples/task_apps/crafter/task_app/synth_envs_hosted/inference/__init__.py +5 -0
  247. examples/task_apps/crafter/task_app/synth_envs_hosted/inference/openai_client.py +999 -0
  248. examples/task_apps/crafter/task_app/synth_envs_hosted/main.py +100 -0
  249. examples/task_apps/crafter/task_app/synth_envs_hosted/policy_routes.py +1252 -0
  250. examples/task_apps/crafter/task_app/synth_envs_hosted/registry.py +195 -0
  251. examples/task_apps/crafter/task_app/synth_envs_hosted/rollout.py +2233 -0
  252. examples/task_apps/crafter/task_app/synth_envs_hosted/storage/__init__.py +5 -0
  253. examples/task_apps/crafter/task_app/synth_envs_hosted/storage/volume.py +211 -0
  254. examples/task_apps/crafter/task_app/synth_envs_hosted/test_agents.py +161 -0
  255. examples/task_apps/crafter/task_app/synth_envs_hosted/test_service.py +136 -0
  256. examples/task_apps/crafter/task_app/synth_envs_hosted/utils.py +411 -0
  257. examples/task_apps/dev/pokemon_emerald/__init__.py +2 -0
  258. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/README.md +811 -0
  259. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/agent/__init__.py +120 -0
  260. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/agent/action.py +160 -0
  261. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/agent/memory.py +155 -0
  262. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/agent/perception.py +69 -0
  263. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/agent/planning.py +96 -0
  264. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/agent/simple.py +1502 -0
  265. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/agent/system_prompt.py +4 -0
  266. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/grab_map.py +68 -0
  267. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/manual.py +216 -0
  268. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/pokemon_env/__init__.py +35 -0
  269. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/pokemon_env/emerald_utils.py +631 -0
  270. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/pokemon_env/emulator.py +1544 -0
  271. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/pokemon_env/enums.py +1428 -0
  272. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/pokemon_env/memory_reader.py +4848 -0
  273. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/pokemon_env/types.py +41 -0
  274. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/pokemon_env/utils.py +298 -0
  275. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/pyproject.toml +95 -0
  276. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/run.py +204 -0
  277. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/server/app.py +2152 -0
  278. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/server/client.py +429 -0
  279. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/server/frame_server.py +155 -0
  280. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/README.md +78 -0
  281. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/run_tests.py +122 -0
  282. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_agent_direct.py +76 -0
  283. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_agent_prompts.py +413 -0
  284. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_battle_state_formatting.py +204 -0
  285. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_dialogue_detection.py +133 -0
  286. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_dialogue_detection_comprehensive.py +229 -0
  287. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_direct_agent_emulator.py +300 -0
  288. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_fps_adjustment_pytest.py +205 -0
  289. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_house_to_outside_direct.py +200 -0
  290. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_house_to_outside_transition.py +284 -0
  291. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_map_ground_truth_comparison.py +468 -0
  292. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_memory_map.py +575 -0
  293. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_server_map_validation.py +311 -0
  294. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests/test_torchic_state.py +259 -0
  295. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/anticheat.py +372 -0
  296. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/checkpoint.py +296 -0
  297. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/error_handler.py +275 -0
  298. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/get_local_ip.py +22 -0
  299. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/helpers.py +44 -0
  300. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/llm_logger.py +514 -0
  301. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/map_formatter.py +415 -0
  302. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/map_stitcher.py +1763 -0
  303. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/map_stitcher_singleton.py +33 -0
  304. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/map_trimmer.py +106 -0
  305. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/map_visualizer.py +334 -0
  306. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/ocr_dialogue.py +1020 -0
  307. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/recording.py +188 -0
  308. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/state_formatter.py +1481 -0
  309. examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils/vlm.py +862 -0
  310. examples/task_apps/dev/pokemon_emerald/modal_app.py +114 -0
  311. examples/task_apps/dev/pokemon_emerald/task_app/README.md +81 -0
  312. examples/task_apps/dev/pokemon_emerald/task_app/__init__.py +6 -0
  313. examples/task_apps/dev/pokemon_emerald/task_app/pokemon_emerald.py +685 -0
  314. examples/task_apps/enron/__init__.py +2 -0
  315. examples/task_apps/enron/eval_groq_qwen32.toml +16 -0
  316. examples/task_apps/enron/filter_sft.toml +5 -0
  317. examples/task_apps/enron/task_app/README.md +14 -0
  318. examples/task_apps/enron/task_app/__init__.py +1 -0
  319. examples/task_apps/enron/task_app/grpo_enron.py +906 -0
  320. examples/task_apps/enron/task_app/grpo_enron_task_app.py +146 -0
  321. examples/task_apps/enron/tests/__init__.py +4 -0
  322. examples/task_apps/enron/tests/conftest.py +115 -0
  323. examples/task_apps/enron/tests/integration/__init__.py +4 -0
  324. examples/task_apps/enron/tests/integration/test_enron_eval.py +179 -0
  325. examples/task_apps/enron/tests/integration/test_enron_rollout.py +135 -0
  326. examples/task_apps/enron/tests/unit/__init__.py +4 -0
  327. examples/task_apps/enron/tests/unit/test_enron_environment.py +126 -0
  328. examples/task_apps/gepa_benchmarks/__init__.py +7 -0
  329. examples/task_apps/gepa_benchmarks/common.py +260 -0
  330. examples/task_apps/gepa_benchmarks/hotpotqa_task_app.py +507 -0
  331. examples/task_apps/gepa_benchmarks/hover_task_app.py +436 -0
  332. examples/task_apps/gepa_benchmarks/ifbench_task_app.py +563 -0
  333. examples/task_apps/gepa_benchmarks/pupa_task_app.py +460 -0
  334. examples/task_apps/math/README.md +21 -0
  335. examples/task_apps/math/math_single_step.py +1000 -0
  336. examples/task_apps/math/math_task_app.py +115 -0
  337. examples/task_apps/pokemon_battle/__init__.py +2 -0
  338. examples/task_apps/pokemon_battle/modal_app.py +104 -0
  339. examples/task_apps/pokemon_battle/task_app/README.md +68 -0
  340. examples/task_apps/pokemon_battle/task_app/__init__.py +6 -0
  341. examples/task_apps/pokemon_battle/task_app/pokemon_showdown.py +932 -0
  342. examples/task_apps/pokemon_red/EVAL_IMAGE_ONLY_COMPLETE.md +283 -0
  343. examples/task_apps/pokemon_red/EVAL_IMAGE_ONLY_STATUS.md +155 -0
  344. examples/task_apps/pokemon_red/README.md +356 -0
  345. examples/task_apps/pokemon_red/README_IMAGE_ONLY_EVAL.md +428 -0
  346. examples/task_apps/pokemon_red/__init__.py +3 -0
  347. examples/task_apps/pokemon_red/eval_image_only_gpt4o.toml +30 -0
  348. examples/task_apps/pokemon_red/eval_pokemon_red_policy.py +224 -0
  349. examples/task_apps/pokemon_red/pallet_town_rl_config.toml +75 -0
  350. examples/task_apps/pokemon_red/task_app.py +1048 -0
  351. examples/task_apps/pokemon_red/test_pallet_town_rewards.py +193 -0
  352. examples/task_apps/sokoban/README.md +306 -0
  353. examples/task_apps/sokoban/__init__.py +3 -0
  354. examples/task_apps/sokoban/eval_groq_qwen32.toml +16 -0
  355. examples/task_apps/sokoban/eval_openai_gpt5.toml +16 -0
  356. examples/task_apps/sokoban/filter_sft.toml +5 -0
  357. examples/task_apps/sokoban/task_app.py +1058 -0
  358. examples/task_apps/sokoban/tests/__init__.py +4 -0
  359. examples/task_apps/sokoban/tests/conftest.py +113 -0
  360. examples/task_apps/sokoban/tests/integration/__init__.py +4 -0
  361. examples/task_apps/sokoban/tests/integration/test_sokoban_eval.py +57 -0
  362. examples/task_apps/sokoban/tests/integration/test_sokoban_rollout.py +198 -0
  363. examples/task_apps/sokoban/tests/unit/__init__.py +4 -0
  364. examples/task_apps/sokoban/tests/unit/test_sokoban_environment.py +114 -0
  365. examples/task_apps/verilog/__init__.py +1 -0
  366. examples/task_apps/verilog/eval_groq_qwen32b.toml +22 -0
  367. examples/task_apps/verilog/filter_sft.toml +5 -0
  368. examples/task_apps/verilog/task_app/README.md +12 -0
  369. examples/task_apps/verilog/task_app/__init__.py +1 -0
  370. examples/task_apps/verilog/task_app/grpo_verilog.py +1166 -0
  371. examples/task_apps/verilog/task_app/grpo_verilog_task_app.py +145 -0
  372. examples/task_apps/verilog/tests/__init__.py +4 -0
  373. examples/task_apps/verilog/tests/conftest.py +115 -0
  374. examples/task_apps/verilog/tests/integration/__init__.py +4 -0
  375. examples/task_apps/verilog/tests/integration/test_verilog_eval.py +181 -0
  376. examples/task_apps/verilog/tests/integration/test_verilog_rollout.py +55 -0
  377. examples/task_apps/verilog/tests/unit/__init__.py +4 -0
  378. examples/task_apps/verilog/tests/unit/test_verilog_scoring.py +118 -0
  379. examples/tunnel_gepa_banking77/README.md +106 -0
  380. examples/tunnel_gepa_banking77/banking77_gepa_tunnel.toml +95 -0
  381. examples/tunnel_gepa_banking77/keep_tunnel_running.py +60 -0
  382. examples/tunnel_gepa_banking77/run_gepa_with_tunnel.sh +226 -0
  383. examples/vlm/PROPOSAL.md +53 -0
  384. examples/vlm/README.md +68 -0
  385. examples/vlm/configs/crafter_vlm_gpt4o.toml +49 -0
  386. examples/vlm/crafter_image_only_agent.py +207 -0
  387. examples/vlm/crafter_openai_vlm_agent.py +275 -0
  388. examples/vlm/filter_image_rows.py +63 -0
  389. examples/vlm/run_crafter_vlm_benchmark.py +316 -0
  390. examples/warming_up_to_rl/_utils.py +92 -0
  391. examples/warming_up_to_rl/analyze_trace_db.py +422 -0
  392. examples/warming_up_to_rl/configs/crafter_fft.toml +53 -0
  393. examples/warming_up_to_rl/configs/crafter_fft_4b.toml +54 -0
  394. examples/warming_up_to_rl/configs/eval_fft_qwen4b.toml +22 -0
  395. examples/warming_up_to_rl/configs/eval_groq_qwen32b.toml +15 -0
  396. examples/warming_up_to_rl/configs/eval_modal_qwen4b.toml +24 -0
  397. examples/warming_up_to_rl/configs/eval_stepwise_complex.toml +35 -0
  398. examples/warming_up_to_rl/configs/eval_stepwise_consistent.toml +26 -0
  399. examples/warming_up_to_rl/configs/eval_stepwise_per_achievement.toml +36 -0
  400. examples/warming_up_to_rl/configs/eval_stepwise_simple.toml +32 -0
  401. examples/warming_up_to_rl/configs/rl_from_base_qwen4b.toml +85 -0
  402. examples/warming_up_to_rl/configs/rl_from_ft.toml +58 -0
  403. examples/warming_up_to_rl/export_trace_sft.py +837 -0
  404. examples/warming_up_to_rl/groq_test.py +97 -0
  405. examples/warming_up_to_rl/manage_secrets.py +131 -0
  406. examples/warming_up_to_rl/old/event_rewards.md +234 -0
  407. examples/warming_up_to_rl/old/notes.md +73 -0
  408. examples/warming_up_to_rl/readme.md +110 -0
  409. examples/warming_up_to_rl/run_eval.py +736 -0
  410. examples/warming_up_to_rl/run_fft_and_save.py +380 -0
  411. examples/warming_up_to_rl/run_local_rollout.py +239 -0
  412. examples/warming_up_to_rl/run_local_rollout_modal.py +248 -0
  413. examples/warming_up_to_rl/run_local_rollout_parallel.py +405 -0
  414. examples/warming_up_to_rl/run_local_rollout_traced.py +477 -0
  415. examples/warming_up_to_rl/run_rl_and_save.py +124 -0
  416. examples/warming_up_to_rl/run_rollout_remote.py +156 -0
  417. examples/warming_up_to_rl/task_app/README.md +42 -0
  418. examples/warming_up_to_rl/task_app/grpo_crafter.py +876 -0
  419. examples/warming_up_to_rl/task_app/grpo_crafter_task_app.py +135 -0
  420. examples/warming_up_to_rl/task_app/synth_envs_hosted/README.md +173 -0
  421. examples/warming_up_to_rl/task_app/synth_envs_hosted/__init__.py +5 -0
  422. examples/warming_up_to_rl/task_app/synth_envs_hosted/branching.py +143 -0
  423. examples/warming_up_to_rl/task_app/synth_envs_hosted/environment_routes.py +1226 -0
  424. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/__init__.py +1 -0
  425. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/__init__.py +6 -0
  426. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/app.py +1 -0
  427. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/environment.py +522 -0
  428. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/policy.py +454 -0
  429. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/react_agent.py +108 -0
  430. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/shared.py +305 -0
  431. examples/warming_up_to_rl/task_app/synth_envs_hosted/envs/crafter/tools.py +47 -0
  432. examples/warming_up_to_rl/task_app/synth_envs_hosted/hosted_app.py +253 -0
  433. examples/warming_up_to_rl/task_app/synth_envs_hosted/inference/__init__.py +5 -0
  434. examples/warming_up_to_rl/task_app/synth_envs_hosted/inference/openai_client.py +729 -0
  435. examples/warming_up_to_rl/task_app/synth_envs_hosted/main.py +100 -0
  436. examples/warming_up_to_rl/task_app/synth_envs_hosted/policy_routes.py +1114 -0
  437. examples/warming_up_to_rl/task_app/synth_envs_hosted/registry.py +195 -0
  438. examples/warming_up_to_rl/task_app/synth_envs_hosted/rollout.py +1891 -0
  439. examples/warming_up_to_rl/task_app/synth_envs_hosted/storage/__init__.py +5 -0
  440. examples/warming_up_to_rl/task_app/synth_envs_hosted/storage/volume.py +211 -0
  441. examples/warming_up_to_rl/task_app/synth_envs_hosted/test_agents.py +161 -0
  442. examples/warming_up_to_rl/task_app/synth_envs_hosted/test_service.py +137 -0
  443. examples/warming_up_to_rl/task_app/synth_envs_hosted/utils.py +129 -0
  444. examples/workflows/math_rl/configs/eval_base_qwen.toml +15 -0
  445. examples/workflows/math_rl/configs/eval_rl_qwen.toml +11 -0
  446. examples/workflows/math_rl/configs/rl_from_base_qwen.toml +62 -0
  447. examples/workflows/math_rl/configs/rl_from_base_qwen17.toml +80 -0
  448. examples/workflows/math_rl/configs/rl_from_ft_qwen.toml +35 -0
  449. examples/workflows/math_rl/download_dataset.py +80 -0
  450. examples/workflows/math_rl/run_eval.py +436 -0
  451. examples/workflows/math_rl/run_rl_and_save.py +111 -0
  452. synth_ai/__init__.py +47 -23
  453. synth_ai/_utils/__init__.py +47 -0
  454. synth_ai/_utils/base_url.py +10 -0
  455. synth_ai/_utils/http.py +10 -0
  456. synth_ai/_utils/prompts.py +10 -0
  457. synth_ai/_utils/task_app_state.py +12 -0
  458. synth_ai/_utils/user_config.py +10 -0
  459. synth_ai/api/models/supported.py +514 -0
  460. synth_ai/api/train/__init__.py +63 -0
  461. synth_ai/api/train/builders.py +473 -0
  462. synth_ai/api/train/cli.py +1185 -0
  463. synth_ai/api/train/config_finder.py +246 -0
  464. synth_ai/api/train/configs/__init__.py +65 -0
  465. synth_ai/api/train/configs/prompt_learning.py +496 -0
  466. synth_ai/api/train/configs/rl.py +188 -0
  467. synth_ai/api/train/configs/sft.py +99 -0
  468. synth_ai/api/train/configs/shared.py +81 -0
  469. synth_ai/api/train/env_resolver.py +352 -0
  470. synth_ai/api/train/pollers.py +91 -0
  471. synth_ai/api/train/prompt_learning.py +425 -0
  472. synth_ai/api/train/sft.py +390 -0
  473. synth_ai/api/train/supported_algos.py +147 -0
  474. synth_ai/api/train/task_app.py +195 -0
  475. synth_ai/api/train/utils.py +244 -0
  476. synth_ai/api/train/validators.py +1117 -0
  477. synth_ai/api/tunnel.py +49 -0
  478. synth_ai/auth/credentials.py +94 -0
  479. synth_ai/baseline/__init__.py +25 -0
  480. synth_ai/baseline/config.py +209 -0
  481. synth_ai/baseline/discovery.py +214 -0
  482. synth_ai/baseline/execution.py +146 -0
  483. synth_ai/cfgs.py +227 -0
  484. synth_ai/cli/__init__.py +90 -45
  485. synth_ai/cli/_modal_wrapper.py +31 -0
  486. synth_ai/cli/_storage.py +20 -0
  487. synth_ai/cli/_typer_patch.py +47 -0
  488. synth_ai/cli/_validate_task_app.py +29 -0
  489. synth_ai/cli/balance.py +16 -4
  490. synth_ai/cli/calc.py +36 -21
  491. synth_ai/cli/claude.py +70 -0
  492. synth_ai/cli/codex.py +267 -0
  493. synth_ai/cli/commands/__init__.py +18 -0
  494. synth_ai/cli/commands/baseline/__init__.py +12 -0
  495. synth_ai/cli/commands/baseline/core.py +637 -0
  496. synth_ai/cli/commands/baseline/list.py +93 -0
  497. synth_ai/cli/commands/demo/__init__.py +6 -0
  498. synth_ai/cli/commands/demo/core.py +163 -0
  499. synth_ai/cli/commands/eval/__init__.py +19 -0
  500. synth_ai/cli/commands/eval/core.py +1112 -0
  501. synth_ai/cli/commands/eval/errors.py +81 -0
  502. synth_ai/cli/commands/eval/validation.py +133 -0
  503. synth_ai/cli/commands/filter/__init__.py +12 -0
  504. synth_ai/cli/commands/filter/core.py +424 -0
  505. synth_ai/cli/commands/filter/errors.py +55 -0
  506. synth_ai/cli/commands/filter/validation.py +77 -0
  507. synth_ai/cli/commands/help/__init__.py +185 -0
  508. synth_ai/cli/commands/help/core.py +72 -0
  509. synth_ai/cli/commands/smoke/__init__.py +7 -0
  510. synth_ai/cli/commands/smoke/core.py +1437 -0
  511. synth_ai/cli/commands/status/__init__.py +66 -0
  512. synth_ai/cli/commands/status/client.py +192 -0
  513. synth_ai/cli/commands/status/config.py +92 -0
  514. synth_ai/cli/commands/status/errors.py +20 -0
  515. synth_ai/cli/commands/status/formatters.py +164 -0
  516. synth_ai/cli/commands/status/subcommands/__init__.py +9 -0
  517. synth_ai/cli/commands/status/subcommands/files.py +79 -0
  518. synth_ai/cli/commands/status/subcommands/jobs.py +334 -0
  519. synth_ai/cli/commands/status/subcommands/models.py +79 -0
  520. synth_ai/cli/commands/status/subcommands/pricing.py +22 -0
  521. synth_ai/cli/commands/status/subcommands/runs.py +81 -0
  522. synth_ai/cli/commands/status/subcommands/session.py +183 -0
  523. synth_ai/cli/commands/status/subcommands/summary.py +47 -0
  524. synth_ai/cli/commands/status/subcommands/usage.py +203 -0
  525. synth_ai/cli/commands/status/utils.py +114 -0
  526. synth_ai/cli/commands/train/__init__.py +53 -0
  527. synth_ai/cli/commands/train/core.py +21 -0
  528. synth_ai/cli/commands/train/errors.py +117 -0
  529. synth_ai/cli/commands/train/judge_schemas.py +200 -0
  530. synth_ai/cli/commands/train/judge_validation.py +305 -0
  531. synth_ai/cli/commands/train/validation.py +386 -0
  532. synth_ai/cli/demo.py +32 -140
  533. synth_ai/cli/deploy.py +233 -0
  534. synth_ai/cli/eval/__init__.py +36 -0
  535. synth_ai/cli/eval/core.py +5 -0
  536. synth_ai/cli/eval/errors.py +31 -0
  537. synth_ai/cli/eval/validation.py +5 -0
  538. synth_ai/cli/filter/__init__.py +28 -0
  539. synth_ai/cli/filter/core.py +5 -0
  540. synth_ai/cli/filter/errors.py +23 -0
  541. synth_ai/cli/filter/validation.py +5 -0
  542. synth_ai/cli/legacy_root_backup.py +28 -22
  543. synth_ai/cli/lib/__init__.py +10 -0
  544. synth_ai/cli/lib/task_app_discovery.py +7 -0
  545. synth_ai/cli/lib/task_app_env.py +518 -0
  546. synth_ai/cli/mcp.py +34 -0
  547. synth_ai/cli/modal_serve/__init__.py +12 -0
  548. synth_ai/cli/modal_serve/core.py +14 -0
  549. synth_ai/cli/modal_serve/errors.py +8 -0
  550. synth_ai/cli/modal_serve/validation.py +11 -0
  551. synth_ai/cli/opencode.py +256 -0
  552. synth_ai/cli/recent.py +13 -7
  553. synth_ai/cli/rl_demo.py +166 -114
  554. synth_ai/cli/root.py +143 -112
  555. synth_ai/cli/serve/__init__.py +12 -0
  556. synth_ai/cli/serve/core.py +14 -0
  557. synth_ai/cli/serve/errors.py +8 -0
  558. synth_ai/cli/serve/validation.py +11 -0
  559. synth_ai/cli/setup.py +49 -0
  560. synth_ai/cli/status.py +7 -125
  561. synth_ai/cli/task_app_deploy.py +7 -0
  562. synth_ai/cli/task_app_list.py +25 -0
  563. synth_ai/cli/task_app_modal_serve.py +11 -0
  564. synth_ai/cli/task_app_serve.py +11 -0
  565. synth_ai/cli/task_apps.py +3134 -0
  566. synth_ai/cli/traces.py +9 -5
  567. synth_ai/cli/train/__init__.py +12 -0
  568. synth_ai/cli/train/core.py +21 -0
  569. synth_ai/cli/train/errors.py +8 -0
  570. synth_ai/cli/train/validation.py +24 -0
  571. synth_ai/cli/train.py +5 -0
  572. synth_ai/cli/turso.py +73 -0
  573. synth_ai/cli/watch.py +13 -18
  574. synth_ai/demos/__init__.py +10 -0
  575. synth_ai/demos/core/__init__.py +28 -1
  576. synth_ai/demos/core/cli.py +745 -416
  577. synth_ai/demos/crafter/__init__.py +1 -0
  578. synth_ai/demos/crafter/crafter_fft_4b.toml +55 -0
  579. synth_ai/demos/crafter/grpo_crafter_task_app.py +185 -0
  580. synth_ai/demos/crafter/rl_from_base_qwen4b.toml +74 -0
  581. synth_ai/demos/demo_registry.py +176 -0
  582. synth_ai/demos/demo_task_apps/__init__.py +7 -1
  583. synth_ai/demos/demo_task_apps/core.py +75 -37
  584. synth_ai/demos/demo_task_apps/crafter/__init__.py +1 -0
  585. synth_ai/demos/demo_task_apps/crafter/configs/crafter_fft_4b.toml +53 -0
  586. synth_ai/demos/demo_task_apps/crafter/configs/rl_from_base_qwen4b.toml +73 -0
  587. synth_ai/demos/demo_task_apps/crafter/grpo_crafter_task_app.py +184 -0
  588. synth_ai/demos/demo_task_apps/math/_common.py +1 -2
  589. synth_ai/demos/demo_task_apps/math/app.py +2 -1
  590. synth_ai/demos/demo_task_apps/math/config.toml +55 -110
  591. synth_ai/demos/demo_task_apps/math/deploy_modal.py +3 -6
  592. synth_ai/demos/demo_task_apps/math/modal_task_app.py +491 -166
  593. synth_ai/demos/demo_task_apps/math/task_app_entry.py +37 -0
  594. synth_ai/demos/math/__init__.py +1 -0
  595. synth_ai/demos/math/_common.py +16 -0
  596. synth_ai/demos/math/app.py +38 -0
  597. synth_ai/demos/math/config.toml +76 -0
  598. synth_ai/demos/math/deploy_modal.py +54 -0
  599. synth_ai/demos/math/modal_task_app.py +703 -0
  600. synth_ai/demos/math/task_app_entry.py +51 -0
  601. synth_ai/environments/environment/core.py +7 -1
  602. synth_ai/environments/examples/bandit/engine.py +12 -5
  603. synth_ai/environments/examples/bandit/environment.py +0 -1
  604. synth_ai/environments/examples/bandit/taskset.py +4 -4
  605. synth_ai/environments/examples/crafter_classic/engine_deterministic_patch.py +7 -4
  606. synth_ai/environments/examples/crafter_classic/engine_serialization_patch_v3.py +9 -5
  607. synth_ai/environments/examples/crafter_classic/environment.py +93 -2
  608. synth_ai/environments/examples/crafter_classic/world_config_patch_simple.py +4 -3
  609. synth_ai/environments/examples/enron/engine.py +7 -2
  610. synth_ai/environments/examples/enron/environment.py +68 -0
  611. synth_ai/environments/examples/red/engine.py +60 -12
  612. synth_ai/environments/examples/red/engine_helpers/memory_map.py +7 -0
  613. synth_ai/environments/examples/red/engine_helpers/reward_components.py +151 -179
  614. synth_ai/environments/examples/red/engine_helpers/reward_library/pallet_town_progression.py +477 -0
  615. synth_ai/environments/examples/red/engine_helpers/state_extraction.py +32 -0
  616. synth_ai/environments/examples/red/environment.py +86 -0
  617. synth_ai/environments/examples/red/trace_hooks_v3.py +168 -0
  618. synth_ai/environments/examples/sokoban/taskset.py +116 -0
  619. synth_ai/environments/examples/verilog/engine.py +104 -12
  620. synth_ai/environments/examples/wordle/environment.py +0 -1
  621. synth_ai/environments/reproducibility/tree.py +5 -6
  622. synth_ai/environments/service/app.py +11 -12
  623. synth_ai/environments/service/core_routes.py +10 -9
  624. synth_ai/environments/stateful/engine.py +1 -1
  625. synth_ai/environments/tasks/core.py +1 -0
  626. synth_ai/environments/tasks/filters.py +5 -6
  627. synth_ai/environments/tasks/utils.py +4 -5
  628. synth_ai/evals/__init__.py +15 -0
  629. synth_ai/evals/base.py +14 -5
  630. synth_ai/evals/client.py +82 -0
  631. synth_ai/evals/types.py +42 -0
  632. synth_ai/http.py +8 -22
  633. synth_ai/http_client.py +45 -12
  634. synth_ai/inference/__init__.py +0 -2
  635. synth_ai/inference/client.py +21 -7
  636. synth_ai/jobs/client.py +129 -80
  637. synth_ai/judge_schemas.py +127 -0
  638. synth_ai/learning/__init__.py +51 -6
  639. synth_ai/learning/algorithms.py +14 -0
  640. synth_ai/learning/client.py +122 -30
  641. synth_ai/learning/config.py +2 -40
  642. synth_ai/learning/constants.py +0 -2
  643. synth_ai/learning/ft_client.py +4 -56
  644. synth_ai/learning/health.py +14 -8
  645. synth_ai/learning/jobs.py +43 -47
  646. synth_ai/learning/prompt_learning_client.py +276 -0
  647. synth_ai/learning/prompt_learning_types.py +185 -0
  648. synth_ai/{rl → learning/rl}/__init__.py +14 -5
  649. synth_ai/learning/rl/client.py +269 -0
  650. synth_ai/learning/rl/config.py +31 -0
  651. synth_ai/{rl → learning/rl}/contracts.py +5 -10
  652. synth_ai/{rl → learning/rl}/env_keys.py +45 -16
  653. synth_ai/learning/rl/secrets.py +13 -0
  654. synth_ai/learning/rl_client.py +2 -253
  655. synth_ai/learning/sft/__init__.py +29 -0
  656. synth_ai/learning/sft/client.py +68 -0
  657. synth_ai/learning/sft/config.py +270 -0
  658. synth_ai/learning/sft/data.py +698 -0
  659. synth_ai/learning/sse.py +25 -26
  660. synth_ai/learning/validators.py +29 -25
  661. synth_ai/mcp/__init__.py +5 -0
  662. synth_ai/mcp/__main__.py +8 -0
  663. synth_ai/mcp/main.py +254 -0
  664. synth_ai/mcp/setup.py +100 -0
  665. synth_ai/modal.py +257 -0
  666. synth_ai/pricing/__init__.py +3 -0
  667. synth_ai/pricing/model_pricing.py +64 -0
  668. synth_ai/session/__init__.py +75 -0
  669. synth_ai/session/client.py +383 -0
  670. synth_ai/session/constants.py +63 -0
  671. synth_ai/session/exceptions.py +105 -0
  672. synth_ai/session/manager.py +139 -0
  673. synth_ai/session/models.py +89 -0
  674. synth_ai/session/query.py +110 -0
  675. synth_ai/spec/__init__.py +46 -0
  676. synth_ai/spec/dataclasses.py +149 -0
  677. synth_ai/spec/loader.py +144 -0
  678. synth_ai/spec/serializer.py +199 -0
  679. synth_ai/spec/validation.py +250 -0
  680. synth_ai/streaming/__init__.py +29 -0
  681. synth_ai/streaming/config.py +94 -0
  682. synth_ai/streaming/handlers.py +589 -0
  683. synth_ai/streaming/streamer.py +320 -0
  684. synth_ai/streaming/types.py +95 -0
  685. synth_ai/task/__init__.py +116 -3
  686. synth_ai/task/apps/__init__.py +132 -0
  687. synth_ai/task/auth.py +165 -0
  688. synth_ai/task/client.py +167 -0
  689. synth_ai/task/config.py +261 -0
  690. synth_ai/task/contracts.py +173 -57
  691. synth_ai/task/datasets.py +108 -0
  692. synth_ai/task/errors.py +50 -0
  693. synth_ai/task/health.py +17 -11
  694. synth_ai/task/inference_api.py +101 -0
  695. synth_ai/task/json.py +111 -0
  696. synth_ai/task/proxy.py +251 -0
  697. synth_ai/task/rubrics/__init__.py +55 -0
  698. synth_ai/task/rubrics/loaders.py +156 -0
  699. synth_ai/task/rubrics/models.py +57 -0
  700. synth_ai/task/rubrics/scoring.py +116 -0
  701. synth_ai/task/rubrics/strict.py +149 -0
  702. synth_ai/task/rubrics.py +219 -0
  703. synth_ai/task/server.py +432 -0
  704. synth_ai/task/trace_correlation_helpers.py +328 -0
  705. synth_ai/task/tracing_utils.py +95 -0
  706. synth_ai/task/validators.py +449 -6
  707. synth_ai/task/vendors.py +59 -0
  708. synth_ai/tracing_v3/__init__.py +4 -0
  709. synth_ai/tracing_v3/abstractions.py +21 -4
  710. synth_ai/tracing_v3/config.py +167 -22
  711. synth_ai/tracing_v3/constants.py +21 -0
  712. synth_ai/tracing_v3/db_config.py +42 -29
  713. synth_ai/tracing_v3/decorators.py +80 -45
  714. synth_ai/tracing_v3/examples/basic_usage.py +15 -9
  715. synth_ai/tracing_v3/hooks.py +6 -4
  716. synth_ai/tracing_v3/llm_call_record_helpers.py +161 -61
  717. synth_ai/tracing_v3/migration_helper.py +1 -2
  718. synth_ai/tracing_v3/replica_sync.py +12 -7
  719. synth_ai/tracing_v3/serialization.py +130 -0
  720. synth_ai/tracing_v3/session_tracer.py +86 -21
  721. synth_ai/tracing_v3/storage/base.py +98 -12
  722. synth_ai/tracing_v3/storage/config.py +63 -16
  723. synth_ai/tracing_v3/storage/factory.py +11 -9
  724. synth_ai/tracing_v3/storage/utils.py +15 -11
  725. synth_ai/tracing_v3/trace_utils.py +317 -0
  726. synth_ai/tracing_v3/turso/__init__.py +8 -21
  727. synth_ai/tracing_v3/turso/daemon.py +123 -15
  728. synth_ai/tracing_v3/turso/models.py +5 -2
  729. synth_ai/tracing_v3/turso/native_manager.py +1293 -0
  730. synth_ai/tracing_v3/utils.py +5 -4
  731. synth_ai/tunnel.py +143 -0
  732. synth_ai/tunnel_deploy.py +278 -0
  733. synth_ai/types.py +8 -0
  734. synth_ai/urls.py +11 -0
  735. synth_ai/utils/__init__.py +166 -0
  736. synth_ai/utils/agents.py +74 -0
  737. synth_ai/utils/apps.py +152 -0
  738. synth_ai/utils/base_url.py +94 -0
  739. synth_ai/utils/bin.py +39 -0
  740. synth_ai/utils/claude.py +36 -0
  741. synth_ai/utils/cli.py +284 -0
  742. synth_ai/utils/config.py +81 -0
  743. synth_ai/utils/env.py +346 -0
  744. synth_ai/utils/errors.py +85 -0
  745. synth_ai/utils/http.py +172 -0
  746. synth_ai/utils/json.py +72 -0
  747. synth_ai/utils/log_filter.py +99 -0
  748. synth_ai/utils/logging.py +198 -0
  749. synth_ai/utils/modal.py +299 -0
  750. synth_ai/utils/paths.py +95 -0
  751. synth_ai/utils/process.py +233 -0
  752. synth_ai/utils/prompts.py +39 -0
  753. synth_ai/utils/sqld.py +122 -0
  754. synth_ai/utils/ssl.py +25 -0
  755. synth_ai/utils/task_app_discovery.py +882 -0
  756. synth_ai/utils/task_app_env.py +186 -0
  757. synth_ai/utils/task_app_state.py +318 -0
  758. synth_ai/utils/tunnel/__init__.py +12 -0
  759. synth_ai/utils/tunnel/config.py +55 -0
  760. synth_ai/utils/user_config.py +137 -0
  761. synth_ai/uvicorn.py +77 -0
  762. synth_ai-0.2.23.dev3.dist-info/METADATA +357 -0
  763. synth_ai-0.2.23.dev3.dist-info/RECORD +983 -0
  764. {synth_ai-0.2.8.dev4.dist-info → synth_ai-0.2.23.dev3.dist-info}/entry_points.txt +0 -1
  765. {synth_ai-0.2.8.dev4.dist-info → synth_ai-0.2.23.dev3.dist-info}/top_level.txt +1 -0
  766. synth_ai/cli/man.py +0 -106
  767. synth_ai/core/experiment.py +0 -15
  768. synth_ai/core/system.py +0 -15
  769. synth_ai/environments/examples/sokoban/units/astar_common.py +0 -95
  770. synth_ai/experimental/synth_oss.py +0 -446
  771. synth_ai/handshake.py +0 -63
  772. synth_ai/install_sqld.sh +0 -40
  773. synth_ai/learning/offline/dpo.py +0 -0
  774. synth_ai/learning/offline/providers.py +0 -7
  775. synth_ai/learning/offline/sft.py +0 -0
  776. synth_ai/learning/offline/shared.py +0 -0
  777. synth_ai/learning/online/grpo.py +0 -0
  778. synth_ai/learning/online/irft.py +0 -0
  779. synth_ai/learning/prompts/banking77_injection_eval.py +0 -168
  780. synth_ai/learning/prompts/gepa.py +0 -0
  781. synth_ai/learning/prompts/hello_world_in_context_injection_ex.py +0 -213
  782. synth_ai/learning/prompts/mipro.py +0 -289
  783. synth_ai/learning/prompts/random_search.py +0 -246
  784. synth_ai/learning/prompts/run_mipro_banking77.py +0 -172
  785. synth_ai/learning/prompts/run_random_search_banking77.py +0 -324
  786. synth_ai/lm/__init__.py +0 -51
  787. synth_ai/lm/caching/constants.py +0 -6
  788. synth_ai/lm/caching/dbs.py +0 -0
  789. synth_ai/lm/caching/ephemeral.py +0 -102
  790. synth_ai/lm/caching/handler.py +0 -137
  791. synth_ai/lm/caching/initialize.py +0 -11
  792. synth_ai/lm/caching/persistent.py +0 -114
  793. synth_ai/lm/config.py +0 -110
  794. synth_ai/lm/constants.py +0 -32
  795. synth_ai/lm/core/__init__.py +0 -8
  796. synth_ai/lm/core/all.py +0 -73
  797. synth_ai/lm/core/exceptions.py +0 -7
  798. synth_ai/lm/core/main.py +0 -319
  799. synth_ai/lm/core/main_v3.py +0 -594
  800. synth_ai/lm/core/synth_models.py +0 -48
  801. synth_ai/lm/core/vendor_clients.py +0 -188
  802. synth_ai/lm/cost/monitor.py +0 -1
  803. synth_ai/lm/cost/statefulness.py +0 -1
  804. synth_ai/lm/injection.py +0 -80
  805. synth_ai/lm/overrides.py +0 -206
  806. synth_ai/lm/provider_support/__init__.py +0 -8
  807. synth_ai/lm/provider_support/anthropic.py +0 -972
  808. synth_ai/lm/provider_support/openai.py +0 -1139
  809. synth_ai/lm/provider_support/suppress_logging.py +0 -31
  810. synth_ai/lm/structured_outputs/handler.py +0 -440
  811. synth_ai/lm/structured_outputs/inject.py +0 -297
  812. synth_ai/lm/structured_outputs/rehabilitate.py +0 -185
  813. synth_ai/lm/tools/__init__.py +0 -3
  814. synth_ai/lm/tools/base.py +0 -172
  815. synth_ai/lm/unified_interface.py +0 -202
  816. synth_ai/lm/vendors/base.py +0 -81
  817. synth_ai/lm/vendors/core/anthropic_api.py +0 -387
  818. synth_ai/lm/vendors/core/gemini_api.py +0 -292
  819. synth_ai/lm/vendors/core/mistral_api.py +0 -322
  820. synth_ai/lm/vendors/core/openai_api.py +0 -225
  821. synth_ai/lm/vendors/core/synth_dev_api.py +0 -0
  822. synth_ai/lm/vendors/local/ollama.py +0 -0
  823. synth_ai/lm/vendors/openai_standard.py +0 -780
  824. synth_ai/lm/vendors/openai_standard_responses.py +0 -256
  825. synth_ai/lm/vendors/retries.py +0 -22
  826. synth_ai/lm/vendors/supported/custom_endpoint.py +0 -417
  827. synth_ai/lm/vendors/supported/deepseek.py +0 -69
  828. synth_ai/lm/vendors/supported/grok.py +0 -75
  829. synth_ai/lm/vendors/supported/groq.py +0 -16
  830. synth_ai/lm/vendors/supported/ollama.py +0 -15
  831. synth_ai/lm/vendors/supported/openrouter.py +0 -74
  832. synth_ai/lm/vendors/supported/together.py +0 -11
  833. synth_ai/lm/vendors/synth_client.py +0 -808
  834. synth_ai/lm/warmup.py +0 -186
  835. synth_ai/rl/secrets.py +0 -19
  836. synth_ai/scripts/verify_rewards.py +0 -100
  837. synth_ai/tracing/__init__.py +0 -30
  838. synth_ai/tracing_v1/__init__.py +0 -33
  839. synth_ai/tracing_v3/turso/manager.py +0 -760
  840. synth_ai/v0/tracing/abstractions.py +0 -224
  841. synth_ai/v0/tracing/base_client.py +0 -91
  842. synth_ai/v0/tracing/client_manager.py +0 -131
  843. synth_ai/v0/tracing/config.py +0 -142
  844. synth_ai/v0/tracing/context.py +0 -146
  845. synth_ai/v0/tracing/decorators.py +0 -682
  846. synth_ai/v0/tracing/events/__init__.py +0 -0
  847. synth_ai/v0/tracing/events/manage.py +0 -147
  848. synth_ai/v0/tracing/events/scope.py +0 -86
  849. synth_ai/v0/tracing/events/store.py +0 -228
  850. synth_ai/v0/tracing/immediate_client.py +0 -151
  851. synth_ai/v0/tracing/local.py +0 -18
  852. synth_ai/v0/tracing/log_client_base.py +0 -73
  853. synth_ai/v0/tracing/retry_queue.py +0 -186
  854. synth_ai/v0/tracing/trackers.py +0 -515
  855. synth_ai/v0/tracing/upload.py +0 -512
  856. synth_ai/v0/tracing/utils.py +0 -9
  857. synth_ai/v0/tracing_v1/__init__.py +0 -16
  858. synth_ai/v0/tracing_v1/abstractions.py +0 -224
  859. synth_ai/v0/tracing_v1/base_client.py +0 -91
  860. synth_ai/v0/tracing_v1/client_manager.py +0 -131
  861. synth_ai/v0/tracing_v1/config.py +0 -142
  862. synth_ai/v0/tracing_v1/context.py +0 -146
  863. synth_ai/v0/tracing_v1/decorators.py +0 -703
  864. synth_ai/v0/tracing_v1/events/__init__.py +0 -0
  865. synth_ai/v0/tracing_v1/events/manage.py +0 -147
  866. synth_ai/v0/tracing_v1/events/scope.py +0 -86
  867. synth_ai/v0/tracing_v1/events/store.py +0 -228
  868. synth_ai/v0/tracing_v1/immediate_client.py +0 -151
  869. synth_ai/v0/tracing_v1/local.py +0 -18
  870. synth_ai/v0/tracing_v1/log_client_base.py +0 -73
  871. synth_ai/v0/tracing_v1/retry_queue.py +0 -186
  872. synth_ai/v0/tracing_v1/trackers.py +0 -515
  873. synth_ai/v0/tracing_v1/upload.py +0 -527
  874. synth_ai/v0/tracing_v1/utils.py +0 -9
  875. synth_ai/zyk/__init__.py +0 -30
  876. synth_ai-0.2.8.dev4.dist-info/METADATA +0 -129
  877. synth_ai-0.2.8.dev4.dist-info/RECORD +0 -420
  878. {synth_ai/lm/caching → examples/task_apps}/__init__.py +0 -0
  879. {synth_ai/lm/cost → examples/task_apps/crafter}/__init__.py +0 -0
  880. {synth_ai/lm/structured_outputs → examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/server}/__init__.py +0 -0
  881. {synth_ai/lm/vendors → examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/tests}/__init__.py +0 -0
  882. {synth_ai/lm/vendors/core → examples/task_apps/dev/pokemon_emerald/external/pokeagent-speedrun/utils}/__init__.py +0 -0
  883. {synth_ai/lm/vendors/local → examples/task_apps/math}/__init__.py +0 -0
  884. {synth_ai/lm/vendors/supported → examples/workflows}/__init__.py +0 -0
  885. {synth_ai/v0/tracing → examples/workflows/math_rl}/__init__.py +0 -0
  886. /synth_ai/{compound/cais.py → cli/__main__.py} +0 -0
  887. /synth_ai/{learning/filtering.py → py.typed} +0 -0
  888. {synth_ai-0.2.8.dev4.dist-info → synth_ai-0.2.23.dev3.dist-info}/WHEEL +0 -0
  889. {synth_ai-0.2.8.dev4.dist-info → synth_ai-0.2.23.dev3.dist-info}/licenses/LICENSE +0 -0
@@ -3,15 +3,14 @@ from __future__ import annotations
3
3
  import json
4
4
  import os
5
5
  import subprocess
6
- import sys
7
- import time
8
- from dataclasses import dataclass
9
- from typing import Any, Dict, Optional, Tuple
10
-
11
6
  import urllib.request
7
+ from dataclasses import dataclass
8
+ from typing import Any
12
9
 
13
10
  from synth_ai.config.base_url import PROD_BASE_URL_DEFAULT
14
11
 
12
+ DEFAULT_TASK_APP_SECRET_NAME = "hendrycks-math-task-app-secret"
13
+
15
14
 
16
15
  @dataclass
17
16
  class DemoEnv:
@@ -20,7 +19,7 @@ class DemoEnv:
20
19
  env_api_key: str = ""
21
20
  task_app_base_url: str = ""
22
21
  task_app_name: str = ""
23
- task_app_secret_name: str = ""
22
+ task_app_secret_name: str = DEFAULT_TASK_APP_SECRET_NAME
24
23
 
25
24
 
26
25
  def _mask(value: str, keep: int = 4) -> str:
@@ -33,7 +32,7 @@ def _state_path() -> str:
33
32
  return os.path.expanduser("~/.synth-ai/demo.json")
34
33
 
35
34
 
36
- def _read_state() -> Dict[str, Any]:
35
+ def _read_state() -> dict[str, Any]:
37
36
  try:
38
37
  path = _state_path()
39
38
  if os.path.isfile(path):
@@ -45,7 +44,7 @@ def _read_state() -> Dict[str, Any]:
45
44
  return {}
46
45
 
47
46
 
48
- def _write_state(data: Dict[str, Any]) -> None:
47
+ def _write_state(data: dict[str, Any]) -> None:
49
48
  try:
50
49
  path = _state_path()
51
50
  os.makedirs(os.path.dirname(path), exist_ok=True)
@@ -55,8 +54,8 @@ def _write_state(data: Dict[str, Any]) -> None:
55
54
  pass
56
55
 
57
56
 
58
- def load_dotenv_file(path: str) -> Dict[str, str]:
59
- out: Dict[str, str] = {}
57
+ def load_dotenv_file(path: str) -> dict[str, str]:
58
+ out: dict[str, str] = {}
60
59
  try:
61
60
  with open(path) as fh:
62
61
  for raw in fh:
@@ -70,7 +69,7 @@ def load_dotenv_file(path: str) -> Dict[str, str]:
70
69
  return out
71
70
 
72
71
 
73
- def _persist_dotenv_values(path: str, values: Dict[str, str]) -> None:
72
+ def _persist_dotenv_values(path: str, values: dict[str, str]) -> None:
74
73
  """Ensure ``values`` are present in ``path`` (.env style)."""
75
74
 
76
75
  try:
@@ -80,7 +79,7 @@ def _persist_dotenv_values(path: str, values: Dict[str, str]) -> None:
80
79
  existing_lines = fh.read().splitlines()
81
80
  else:
82
81
  os.makedirs(os.path.dirname(path) or ".", exist_ok=True)
83
- mapping: Dict[str, str] = {}
82
+ mapping: dict[str, str] = {}
84
83
  order: list[str] = []
85
84
  for line in existing_lines:
86
85
  if not line or line.startswith("#") or "=" not in line:
@@ -108,7 +107,7 @@ def _persist_dotenv_values(path: str, values: Dict[str, str]) -> None:
108
107
  pass
109
108
 
110
109
 
111
- def persist_dotenv_values(values: Dict[str, str], *, cwd: str | None = None) -> str:
110
+ def persist_dotenv_values(values: dict[str, str], *, cwd: str | None = None) -> str:
112
111
  path = os.path.join(cwd or os.getcwd(), ".env")
113
112
  _persist_dotenv_values(path, values)
114
113
  return path
@@ -120,14 +119,41 @@ def persist_env_api_key(key: str) -> None:
120
119
  _write_state(data)
121
120
 
122
121
 
123
- def modal_auth_status() -> Tuple[bool, str]:
122
+ def persist_demo_dir(demo_dir: str) -> None:
123
+ """Store the demo directory path for subsequent commands."""
124
+ data = _read_state()
125
+ data["DEMO_DIR"] = demo_dir
126
+ _write_state(data)
127
+
128
+
129
+ def load_demo_dir() -> str | None:
130
+ """Load the stored demo directory path, if any."""
131
+ data = _read_state()
132
+ return data.get("DEMO_DIR")
133
+
134
+
135
+ def persist_env_file_path(env_path: str) -> None:
136
+ """Store the .env file path for subsequent commands."""
137
+ data = _read_state()
138
+ data["ENV_FILE_PATH"] = env_path
139
+ _write_state(data)
140
+
141
+
142
+ def load_env_file_path() -> str | None:
143
+ """Load the stored .env file path, if any."""
144
+ data = _read_state()
145
+ return data.get("ENV_FILE_PATH")
146
+
147
+
148
+ def modal_auth_status() -> tuple[bool, str]:
124
149
  """Return (ok, message) describing Modal CLI credential status."""
125
150
 
126
151
  env_token_id = (os.environ.get("MODAL_TOKEN_ID") or "").strip()
127
152
  env_token_secret = (os.environ.get("MODAL_TOKEN_SECRET") or "").strip()
128
153
 
129
154
  try:
130
- from modal.config import config as modal_config, user_config_path
155
+ from modal.config import config as modal_config
156
+ from modal.config import user_config_path
131
157
  except Exception as exc: # pragma: no cover - modal optional in some envs
132
158
  return False, f"Modal client unavailable ({exc})"
133
159
 
@@ -182,7 +208,7 @@ def load_env() -> DemoEnv:
182
208
  """
183
209
  env = DemoEnv()
184
210
 
185
- os_env: Dict[str, str] = dict(os.environ)
211
+ os_env: dict[str, str] = dict(os.environ)
186
212
 
187
213
  # CWD .env
188
214
  cwd_env_path = os.path.join(os.getcwd(), ".env")
@@ -191,7 +217,9 @@ def load_env() -> DemoEnv:
191
217
  # Repo/package .envs (fallbacks)
192
218
  repo_root = os.path.abspath(os.path.join(os.path.dirname(__file__), "../../.."))
193
219
  repo_env = load_dotenv_file(os.path.join(repo_root, ".env"))
194
- pkg_env = load_dotenv_file(os.path.join(repo_root, "synth_ai", "demos", "demo_task_apps", "math", ".env"))
220
+ pkg_env = load_dotenv_file(
221
+ os.path.join(repo_root, "synth_ai", "demos", "demo_task_apps", "math", ".env")
222
+ )
195
223
  examples_env = load_dotenv_file(os.path.join(repo_root, "examples", "rl", ".env"))
196
224
 
197
225
  state = _read_state()
@@ -215,15 +243,12 @@ def load_env() -> DemoEnv:
215
243
  or pkg_env.get("DEV_BACKEND_URL")
216
244
  or ""
217
245
  ).strip()
218
- use_dev = False
219
246
  if backend_override:
220
247
  dev_url = backend_override
221
- use_dev = True
222
248
  elif dev_env:
223
249
  lower = dev_env.lower()
224
250
  if "localhost" in lower or "127.0.0.1" in lower or lower.endswith(":8000"):
225
251
  dev_url = dev_env
226
- use_dev = True
227
252
  else:
228
253
  dev_url = prod_default
229
254
  else:
@@ -240,7 +265,11 @@ def load_env() -> DemoEnv:
240
265
  or str(state.get("SYNTH_API_KEY") or "")
241
266
  )
242
267
  if not synth_api_key:
243
- mode = "prod" if default_root in dev_url else ("local" if ("localhost" in dev_url or "127.0.0.1" in dev_url) else "dev")
268
+ mode = (
269
+ "prod"
270
+ if default_root in dev_url
271
+ else ("local" if ("localhost" in dev_url or "127.0.0.1" in dev_url) else "dev")
272
+ )
244
273
  if mode == "prod":
245
274
  synth_api_key = (
246
275
  os_env.get("PROD_SYNTH_API_KEY")
@@ -285,7 +314,7 @@ def load_env() -> DemoEnv:
285
314
  )
286
315
 
287
316
  task_app_name = str(state.get("TASK_APP_NAME") or "")
288
- task_app_secret_name = str(state.get("TASK_APP_SECRET_NAME") or "")
317
+ task_app_secret_name = str(state.get("TASK_APP_SECRET_NAME") or DEFAULT_TASK_APP_SECRET_NAME)
289
318
 
290
319
  env.dev_backend_url = dev_url.rstrip("/")
291
320
  env.synth_api_key = synth_api_key
@@ -309,7 +338,9 @@ def load_env() -> DemoEnv:
309
338
  return env
310
339
 
311
340
 
312
- def assert_http_ok(url: str, method: str = "GET", allow_redirects: bool = True, timeout: float = 10.0) -> bool:
341
+ def assert_http_ok(
342
+ url: str, method: str = "GET", allow_redirects: bool = True, timeout: float = 10.0
343
+ ) -> bool:
313
344
  try:
314
345
  import ssl
315
346
 
@@ -367,16 +398,17 @@ def persist_task_url(url: str, *, name: str | None = None) -> None:
367
398
  if data.get("TASK_APP_NAME") != name:
368
399
  data["TASK_APP_NAME"] = name
369
400
  changed.append("TASK_APP_NAME")
370
- secret_name = f"{name}-secret"
371
- if data.get("TASK_APP_SECRET_NAME") != secret_name:
372
- data["TASK_APP_SECRET_NAME"] = secret_name
373
- if "TASK_APP_NAME" not in changed:
374
- changed.append("TASK_APP_SECRET_NAME")
401
+ if data.get("TASK_APP_SECRET_NAME") != DEFAULT_TASK_APP_SECRET_NAME:
402
+ data["TASK_APP_SECRET_NAME"] = DEFAULT_TASK_APP_SECRET_NAME
403
+ changed.append("TASK_APP_SECRET_NAME")
404
+ elif data.get("TASK_APP_SECRET_NAME") != DEFAULT_TASK_APP_SECRET_NAME:
405
+ data["TASK_APP_SECRET_NAME"] = DEFAULT_TASK_APP_SECRET_NAME
406
+ changed.append("TASK_APP_SECRET_NAME")
375
407
  _write_state(data)
376
408
  if changed:
377
409
  print(f"Saved {', '.join(changed)} to {_state_path()}")
378
- if "TASK_APP_NAME" in changed or "TASK_APP_SECRET_NAME" in changed:
379
- print(f"TASK_APP_SECRET_NAME={data.get('TASK_APP_SECRET_NAME', '')}")
410
+ if "TASK_APP_SECRET_NAME" in changed:
411
+ print(f"TASK_APP_SECRET_NAME={DEFAULT_TASK_APP_SECRET_NAME}")
380
412
 
381
413
 
382
414
  def persist_api_key(key: str) -> None:
@@ -385,18 +417,24 @@ def persist_api_key(key: str) -> None:
385
417
  _write_state(data)
386
418
 
387
419
 
388
- def run_job(env: DemoEnv, config_toml_path: str, *, batch_size: Optional[int] = None, group_size: Optional[int] = None, model: Optional[str] = None) -> None:
420
+ def run_job(
421
+ env: DemoEnv,
422
+ config_toml_path: str,
423
+ *,
424
+ batch_size: int | None = None,
425
+ group_size: int | None = None,
426
+ model: str | None = None,
427
+ ) -> None:
389
428
  """Create and stream a short RL job using the backend API (placeholder: prints cURL to execute)."""
390
429
  backend = env.dev_backend_url.rstrip("/")
391
- if backend.endswith("/api"):
392
- api_base = backend
393
- else:
394
- api_base = backend + "/api"
430
+ api_base = backend if backend.endswith("/api") else backend + "/api"
395
431
  print("\nTo create an RL job, run:")
396
432
  print(
397
- "curl -s -X POST \"" + api_base + "/rl/jobs\" "
433
+ 'curl -s -X POST "' + api_base + '/rl/jobs" '
398
434
  "-H 'Content-Type: application/json' "
399
435
  f"-H 'Authorization: Bearer {env.synth_api_key}' "
400
436
  "-d '{" # intentionally not fully formed here for brevity in this scaffold
401
437
  )
402
- print(" NOTE: CLI implementation will build the full JSON body with inline TOML config and stream events.")
438
+ print(
439
+ " NOTE: CLI implementation will build the full JSON body with inline TOML config and stream events."
440
+ )
@@ -0,0 +1 @@
1
+ # Crafter demo task app
@@ -0,0 +1,53 @@
1
+ # FFT job config for Qwen/Qwen3-4B on Crafter SFT dataset
2
+
3
+ [algorithm]
4
+ type = "offline"
5
+ method = "supervised_finetune"
6
+ variety = "fft"
7
+
8
+
9
+ [job]
10
+ model = "Qwen/Qwen3-4B"
11
+ data = "ft_data/crafter_traces.jsonl"
12
+
13
+ [compute]
14
+ # Adjust as needed for your quota
15
+ gpu_type = "H100"
16
+ gpu_count = 1
17
+ nodes = 1
18
+
19
+ [data]
20
+ # Optional topology metadata (left empty for now)
21
+ topology = {}
22
+
23
+ # Optional local validation dataset path (JSONL). If set, the client will upload
24
+ # this file and wire up validation so the frontend can display val.loss.
25
+ # validation_path = "../ft_data/crafter_validation.jsonl"
26
+
27
+ [training]
28
+ mode = "sft_offline"
29
+ use_qlora = false
30
+
31
+ # Validation settings to emit val.loss on the frontend
32
+ [training.validation]
33
+ enabled = true
34
+ evaluation_strategy = "steps"
35
+ eval_steps = 20
36
+ save_best_model_at_end = true
37
+ metric_for_best_model = "val.loss"
38
+ greater_is_better = false
39
+
40
+ [hyperparameters]
41
+ # Minimal safe defaults; backend can override
42
+ n_epochs = 1
43
+ batch_size = 1
44
+ gradient_accumulation_steps = 64
45
+ sequence_length = 4096
46
+ learning_rate = 5e-6
47
+ warmup_ratio = 0.03
48
+ train_kind = "fft"
49
+
50
+ # Optional parallelism block example
51
+ #[hyperparameters.parallelism]
52
+ # tensor_parallel_size = 1
53
+ # pipeline_parallel_size = 1
@@ -0,0 +1,73 @@
1
+ # RL training starting from base Qwen/Qwen3-4B (TOML-only model selection)
2
+
3
+ [algorithm]
4
+ type = "online"
5
+ method = "policy_gradient"
6
+ variety = "gspo"
7
+
8
+
9
+ [services]
10
+ task_url = "https://synth-laboratories--grpo-crafter-task-app-final-warming--ceb5b2.modal.run"
11
+
12
+ [compute]
13
+ # Cluster shape for RL pipeline
14
+ gpu_type = "H100"
15
+ gpu_count = 8
16
+
17
+ [topology]
18
+ # Split GPUs across vLLM, training, and reference
19
+ # Must sum to compute.gpu_count
20
+ type = "single_node_split"
21
+ gpus_for_vllm = 4
22
+ gpus_for_training = 3
23
+ gpus_for_ref = 1
24
+ tensor_parallel = 4
25
+
26
+ [vllm]
27
+ # Serving tensor parallel size
28
+ tensor_parallel_size = 4
29
+ max_model_len = 8192
30
+
31
+ [reference]
32
+ # Required by trainer/runtime; ensures dedicated/scoped scoring server config exists
33
+ placement = "dedicated"
34
+ port = 8002
35
+ tp = 1
36
+ health_max_wait_s = 180
37
+ health_interval_ms = 300
38
+
39
+ [model]
40
+ # Base model start
41
+ base = "Qwen/Qwen3-4B"
42
+ label = "crafter-rl-from-base"
43
+
44
+ [rollout]
45
+ max_turns = 10
46
+ episodes_per_batch = 64
47
+ policy_name = "crafter"
48
+
49
+ [evaluation]
50
+ # Run baseline evaluation over the first 100 seeds every 20 training iterations
51
+ instances = 10
52
+ every_n_iters = 10
53
+ seeds = [
54
+ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,
55
+ ]
56
+
57
+ [training]
58
+ log_interval = 1
59
+ weight_sync_interval = 1
60
+ # Additional RL hyperparameters can go here
61
+
62
+ # Stepwise rewards (Crafter decision-level)
63
+ step_rewards_enabled = true
64
+ step_rewards_mode = "decision_stepwise" # "off" | "decision_stepwise" | "env_sparse"
65
+ step_rewards_beta = 0.0
66
+ step_rewards_indicator_lambda = 1.0
67
+ # Optional selector for decision scalar: "unique" | "absolute" (default unique)
68
+ event_rewards_kind = "unique"
69
+
70
+ [training.weight_sync]
71
+ enable = true
72
+ targets = ["policy"]
73
+ weight_sync_interval = 1
@@ -0,0 +1,184 @@
1
+ """Compatibility wrapper for the GRPO Crafter task app.
2
+
3
+ This module now delegates to the TaskAppConfig defined in the local example at
4
+ `examples/task_apps/crafter/task_app/grpo_crafter.py`. It is kept for legacy usage
5
+ (running the file directly or targeting `fastapi_app` from external tooling).
6
+ Prefer using `uvx synth-ai deploy --runtime uvicorn grpo-crafter` for local development and testing.
7
+ """
8
+
9
+ from __future__ import annotations
10
+
11
+ import argparse
12
+ import importlib.util
13
+ from pathlib import Path
14
+
15
+ from fastapi.exceptions import RequestValidationError
16
+ from fastapi.responses import JSONResponse
17
+ from starlette.requests import Request
18
+ from synth_ai.task.apps import ModalDeploymentConfig, registry
19
+ from synth_ai.task.auth import is_api_key_header_authorized, normalize_environment_api_key
20
+ from synth_ai.task.server import TaskAppConfig, create_task_app, run_task_app
21
+
22
+
23
+ def _load_build_config():
24
+ """Load the example's build_config, preferring package import with file fallback."""
25
+ # First try to import by package name (installed 'examples' package)
26
+ try:
27
+ module = importlib.import_module("examples.task_apps.crafter.task_app.grpo_crafter")
28
+ return module.build_config # type: ignore[attr-defined]
29
+ except Exception:
30
+ # Fallback: locate the file within the installed synth_ai distribution and exec it
31
+ import sys as _sys
32
+
33
+ import synth_ai
34
+
35
+ synth_ai_path = Path(synth_ai.__file__).resolve().parent.parent
36
+ module_path = (
37
+ synth_ai_path / "examples" / "task_apps" / "crafter" / "task_app" / "grpo_crafter.py"
38
+ )
39
+
40
+ if not module_path.exists():
41
+ raise ImportError(
42
+ f"Could not find task app module at {module_path}. Make sure you're running from the synth-ai repository."
43
+ ) from None
44
+
45
+ spec = importlib.util.spec_from_file_location(
46
+ "examples.task_apps.crafter.task_app.grpo_crafter", module_path
47
+ )
48
+ if spec is None or spec.loader is None:
49
+ raise ImportError(f"Could not load task app module at {module_path}") from None
50
+
51
+ module = importlib.util.module_from_spec(spec)
52
+ _sys.modules[spec.name] = module
53
+ spec.loader.exec_module(module)
54
+ return module.build_config # type: ignore[attr-defined]
55
+
56
+
57
+ build_config = _load_build_config()
58
+
59
+
60
+ APP_ID = "grpo-crafter-task-app"
61
+
62
+
63
+ def _build_base_config() -> TaskAppConfig:
64
+ # Lazily construct the base config to avoid heavy work at import time
65
+ return build_config()
66
+
67
+
68
+ try:
69
+ _REGISTERED_ENTRY = registry.get(APP_ID)
70
+ except Exception: # pragma: no cover - registry unavailable in some contexts
71
+ MODAL_DEPLOYMENT: ModalDeploymentConfig | None = None
72
+ ENV_FILES: tuple[str, ...] = ()
73
+ else:
74
+ MODAL_DEPLOYMENT = _REGISTERED_ENTRY.modal
75
+ ENV_FILES = tuple(_REGISTERED_ENTRY.env_files)
76
+
77
+
78
+ def build_task_app_config() -> TaskAppConfig:
79
+ """Return a fresh TaskAppConfig for this wrapper."""
80
+
81
+ base = _build_base_config()
82
+ return base.clone()
83
+
84
+
85
+ def fastapi_app():
86
+ """Return the FastAPI application for Modal or other ASGI hosts."""
87
+
88
+ app = create_task_app(build_task_app_config())
89
+
90
+ # Replace default health endpoints so we can permit soft auth failures and log 422s.
91
+ filtered_routes = []
92
+ for route in app.router.routes:
93
+ path = getattr(route, "path", None)
94
+ methods = getattr(route, "methods", set()) or set()
95
+ if path in {"/health", "/health/rollout"} and "GET" in methods:
96
+ continue
97
+ filtered_routes.append(route)
98
+ app.router.routes = filtered_routes
99
+
100
+ def _log_env_key_prefix(source: str, env_key: str | None) -> str | None:
101
+ if not env_key:
102
+ return None
103
+ prefix = env_key[: max(1, len(env_key) // 2)]
104
+ print(f"[{source}] expected ENVIRONMENT_API_KEY prefix: {prefix}")
105
+ return prefix
106
+
107
+ @app.get("/health")
108
+ async def health(request: Request):
109
+ env_key = normalize_environment_api_key()
110
+ if not env_key:
111
+ return JSONResponse(
112
+ status_code=503,
113
+ content={"status": "unhealthy", "detail": "Missing ENVIRONMENT_API_KEY"},
114
+ )
115
+ if not is_api_key_header_authorized(request):
116
+ prefix = _log_env_key_prefix("health", env_key)
117
+ content = {"status": "healthy", "authorized": False}
118
+ if prefix:
119
+ content["expected_api_key_prefix"] = prefix
120
+ return JSONResponse(status_code=200, content=content)
121
+ return {"status": "healthy", "authorized": True}
122
+
123
+ @app.get("/health/rollout")
124
+ async def health_rollout(request: Request):
125
+ env_key = normalize_environment_api_key()
126
+ if not env_key:
127
+ return JSONResponse(
128
+ status_code=503,
129
+ content={"status": "unhealthy", "detail": "Missing ENVIRONMENT_API_KEY"},
130
+ )
131
+ if not is_api_key_header_authorized(request):
132
+ prefix = _log_env_key_prefix("health/rollout", env_key)
133
+ content = {"status": "healthy", "authorized": False}
134
+ if prefix:
135
+ content["expected_api_key_prefix"] = prefix
136
+ return JSONResponse(status_code=200, content=content)
137
+ return {"ok": True, "authorized": True}
138
+
139
+ @app.exception_handler(RequestValidationError)
140
+ async def _on_validation_error(request: Request, exc: RequestValidationError):
141
+ try:
142
+ hdr = request.headers
143
+ snapshot = {
144
+ "path": str(request.url.path),
145
+ "have_x_api_key": bool(hdr.get("x-api-key")),
146
+ "have_x_api_keys": bool(hdr.get("x-api-keys")),
147
+ "have_authorization": bool(hdr.get("authorization")),
148
+ "errors": exc.errors()[:5],
149
+ }
150
+ print("[422] validation", snapshot, flush=True)
151
+ except Exception:
152
+ pass
153
+ return JSONResponse(
154
+ status_code=422,
155
+ content={"status": "invalid", "detail": exc.errors()[:5]},
156
+ )
157
+
158
+ return app
159
+
160
+
161
+ if __name__ == "__main__":
162
+ parser = argparse.ArgumentParser(description="Run the Crafter task app locally")
163
+ parser.add_argument("--host", default="0.0.0.0")
164
+ parser.add_argument("--port", type=int, default=8001)
165
+ parser.add_argument("--reload", action="store_true", help="Enable uvicorn autoreload")
166
+ parser.add_argument(
167
+ "--env-file",
168
+ action="append",
169
+ default=[],
170
+ help="Additional .env files to load before startup",
171
+ )
172
+ args = parser.parse_args()
173
+
174
+ default_env = Path(__file__).resolve().parents[4] / "backend" / ".env.dev"
175
+ env_files = [str(default_env)] if default_env.exists() else []
176
+ env_files.extend(args.env_file or [])
177
+
178
+ run_task_app(
179
+ build_task_app_config,
180
+ host=args.host,
181
+ port=args.port,
182
+ reload=args.reload,
183
+ env_files=env_files,
184
+ )
@@ -6,6 +6,7 @@ This module provides a local fallback for install_problem_bank_into_shared so
6
6
  the modal task app can import it without requiring an external math_rl package.
7
7
  """
8
8
 
9
+
9
10
  def install_problem_bank_into_shared() -> None:
10
11
  """No-op placeholder for installing the Hendrycks MATH problem bank.
11
12
 
@@ -13,5 +14,3 @@ def install_problem_bank_into_shared() -> None:
13
14
  into a shared directory. For the demo scaffold, it is a no-op.
14
15
  """
15
16
  return None
16
-
17
-
@@ -1,6 +1,7 @@
1
1
  from __future__ import annotations
2
2
 
3
3
  import os
4
+
4
5
  from fastapi import FastAPI
5
6
  from starlette.middleware.cors import CORSMiddleware
6
7
 
@@ -11,7 +12,7 @@ except Exception: # fallback path when imported from repo root
11
12
  try:
12
13
  from examples.rl.task_app import make_app as make_rl_app # type: ignore
13
14
  except Exception as e: # pragma: no cover
14
- raise ImportError(f"Unable to import RL task app: {e}")
15
+ raise ImportError(f"Unable to import RL task app: {e}") from e
15
16
 
16
17
 
17
18
  def create_app() -> FastAPI: