rio 0.4.0__py3-none-any.whl

rio-0.4.0.dist-info/METADATA

@@ -0,0 +1,90 @@
+Metadata-Version: 2.5
+Name: rio
+Version: 0.4.0
+Summary: rio_ai: a ported multi-provider LLM streaming SDK. rio_agent: a SKILL.state long-horizon agent runtime (arXiv:2608.26263). rio_coding: a SKILL.state coding agent.
+Project-URL: Repository, https://github.com/soasme/rio
+Author-email: Ju Lin <soasme@gmail.com>
+License-Expression: MIT
+License-File: LICENSE
+License-File: NOTICE
+Requires-Python: >=3.12
+Requires-Dist: anyio>=4.0
+Requires-Dist: httpx[socks]>=0.27
+Requires-Dist: packaging>=24.0
+Requires-Dist: pillow>=11.0
+Requires-Dist: pydantic>=2.11
+Requires-Dist: pygments>=2.18
+Requires-Dist: pyte>=0.8.2
+Requires-Dist: rich>=13.0
+Requires-Dist: textual-diff-view>=0.1.5
+Requires-Dist: textual>=8.2.8
+Requires-Dist: typer>=0.12
+Description-Content-Type: text/markdown
+
+# rio
+
+Four packages:
+
+- **`rio_ai`** -- a multi-provider LLM streaming SDK (Anthropic, Google
+  Gemini, Mistral, OpenAI Codex, OpenAI-compatible), ported from
+  [huggingface/tau](https://github.com/huggingface/tau)'s `tau_ai` module.
+  The message/tool/type vocabulary `tau_ai` depends on (`tau_agent.messages`,
+  `tau_agent.tools`, `tau_agent.types`, `tau_agent.provider`,
+  `tau_agent.provider_events` upstream) is folded in here too, so `rio_ai`
+  is self-contained. See [`NOTICE`](NOTICE) for the upstream MIT license.
+- **`rio_agent`** -- a from-scratch runtime implementing
+  [*SKILL.state: Scalable Long-Horizon Agent Skills*](https://arxiv.org/abs/2608.26263)
+  (Badhe, Tiwari & Chung; EMNLP), built on `rio_ai`.
+- **`rio_coding`** -- the coding backend (CLI, tools, skills, extensions), built on `rio_agent`.
+- **`rio_tui`** -- the terminal application, wired to `rio_coding`. It owns
+  session tabs, the prompt editor, shell, file navigation, diffs, and settings.
+
+## Coding agent
+
+```bash
+uv sync
+uv run rio setup --provider local --base-url http://localhost:8080/v1 --model my-model
+uv run rio                         # interactive terminal interface
+uv run rio -p "Inspect this project and explain its entry point"
+uv run rio -p --mode json "Add a regression test for the parser"
+uv run rio --mode rpc              # JSONL commands on stdin
+uv run rio sessions
+uv run rio --session SESSION_ID
+uv run rio export SESSION_ID --format html
+```
+
+Choose an existing provider with `--provider NAME --model MODEL`; `rio providers`
+lists the configured catalog. Credentials may come from the provider's environment
+variable or Rio's credential store (`rio login PROVIDER`; use `--method api-key`
+for an API key). A successful login remembers the provider for future launches.
+API keys and OAuth tokens are stored in `~/.rio/credentials.json` with owner-only
+permissions; the file is unencrypted. OAuth tokens refresh automatically.
+Use `rio logout PROVIDER` to remove saved credentials. Configuration and journals
+live under `~/.rio`.
+Use `--approve` to allow ambient project instructions and extensions for a run;
+project trust controls resource loading, not what shell commands can access.
+
+The coding tools are `read`, `write`, `edit`, `bash`, and `respond`. Each step
+updates structured state and executes one action. Sessions store snapshots for
+resume and checkpoint restoration; they never replay a transcript into the model.
+HTML exports show steps, final state, and estimated token footprints.
+
+See the [installed documentation](src/rio_coding/data/docs/README.md) for
+providers, skills, extensions, RPC, and the state-based runtime. The
+[offline example](src/rio_coding/data/examples/offline_session.py) runs a complete
+session with a fake provider and requires no credentials.
+
+## Development
+
+```bash
+uv sync
+uv run pytest
+uv run ruff check .
+```
+
+Tests use scripted providers (`rio_ai.FakeProvider`) and mocked HTTP; no live
+provider credentials are required. Headless Textual tests exercise terminal
+interaction, and tool integration tests work in temporary directories. See in
+particular `tests/test_rio_agent_loop.py::test_prompt_footprint_is_bounded_across_steps`,
+which asserts that per-step prompt size stays constant across many steps
+rather than growing -- a direct runtime check of the paper's core claim.

rio-0.4.0.dist-info/RECORD

@@ -0,0 +1,140 @@
+rio_agent/__init__.py,sha256=DC9rZDoQ4qI1gUamg1RbB9dZHtfSsiajm4F-6mxv18Y,1629
+rio_agent/errors.py,sha256=Zep_eWy0u4dTAY2vy534Uzl51pSsvFfwLvRWz2ioLJ8,939
+rio_agent/events.py,sha256=fYez8Yqf5Q_xQ1M7GR0WZxVfeNdRgj3NgyRCadf6aWo,2017
+rio_agent/harness.py,sha256=J9XVfofIIH5hvXHCLVkW0uyzTUuTKTraBLhn4gHpgmo,4105
+rio_agent/loop.py,sha256=ypgEm49I-5j15NjgSfbBy6T31AB9zL859_wnTftBQmI,8489
+rio_agent/observation.py,sha256=XDWkY80XYvAaOtBkmTaaGlNUsh_2G1-ioR2WVL9ZnaY,887
+rio_agent/prompt.py,sha256=93v8gJyLMj7dYkoFnReFnAFhQW0h8DKZG678jfyS5XE,4288
+rio_agent/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+rio_agent/skill.py,sha256=Ia_5duVHTSDAVIQO53oQsFKtr34g49-IUISSjonaRsQ,889
+rio_agent/state.py,sha256=DKTMV_Pffa7Jdh_MKA6BOyYlRASnxvmRr-weiuAJLQk,1873
+rio_ai/__init__.py,sha256=9KCNoV61_DmwbDkL7gftewJKEGST82iqjGGLH04Kr4I,2468
+rio_ai/_provider_events.py,sha256=qx9pdFfL6u6WXR5_xYccSX_C5SkZVTcUeA6ysqMxJP8,2321
+rio_ai/anthropic.py,sha256=5A7gSNLImEjWTKAsx8zMXIEDILH6sA0PWiLRvG7x_Bo,32553
+rio_ai/content.py,sha256=HIZ_NT61L7mX9nYY7sLuO5nA7fjUGFiXdAbH1mnc-zc,1683
+rio_ai/env.py,sha256=sSgw6mEuiEefWovVg7022jj8rAuUSgMXMXP8vDIpsv8,6218
+rio_ai/events.py,sha256=kKHwffbfjb4yXIPE1iGvcMnTKN1yPQ1zPKGLKI180zw,823
+rio_ai/fake.py,sha256=x0mU14Q0q9pXaVvnYmsDSwmcjOioy1DAF9GQpBHHRK4,1417
+rio_ai/google.py,sha256=e9huzlaXV1INGErp2nPcdnXHiXKP-5UCUqyphPBoXAI,20609
+rio_ai/http.py,sha256=9BvofnxhZqnp3a8lxKblRY6nfrvhz_4PAtDkATYCkOc,2387
+rio_ai/http_errors.py,sha256=kuSGGwkSCi9XLUFHAuR2xjOAXJPJbPmbeYanIiJ7I8Y,2020
+rio_ai/messages.py,sha256=1jCeqn8yovwMfVopZaJQUtKSuzvWIVEPpdaQaW_VDTo,8419
+rio_ai/mistral.py,sha256=1emOQywqR_CrL-oKXIrN1CWRgo4XNLKHx6_r75QS-Gw,19487
+rio_ai/model_limits.py,sha256=zxC4CbA6lrPhvmbkx_fp8gmJCSOUEzDLvizY-dNZtVk,2077
+rio_ai/openai_cache.py,sha256=jpyLf2jJndg3OVNdgJX0dTuOQU4zmqIlv9jdXQBynzE,628
+rio_ai/openai_codex.py,sha256=RwUFVzA4_FverridAD6kYb6mnpAtsfRg2EbU_PbsErM,40286
+rio_ai/openai_compatible.py,sha256=tJAhUdiuCXOEVL1PM2udYVExGLsYfkaSM6ksAXVHeHM,52499
+rio_ai/provider.py,sha256=pbyT8k6EYy5m6K7OmIBInqnv1XYfF6wU4aKYxwx_YLM,1062
+rio_ai/provider_events.py,sha256=1DlMfBAp5xIWCVnBEABQ4rXQj2d3YQOrBhs0nW188RQ,2438
+rio_ai/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+rio_ai/retry.py,sha256=b2hYU0k4bh73ihsUH-RK1gW_ZauSPH-FrsrTH3jz-ac,1885
+rio_ai/stream.py,sha256=KT5yVDfxZjlWfP2l4Cq7n3UH2jtqnWSOIGbZjv2pu-0,8454
+rio_ai/tool_call_ids.py,sha256=7-faY63oTCtTmdUBEQ9z3o9m0BREsIphCsEnCBWa0NI,928
+rio_ai/tools.py,sha256=GAZ6Em0ptvFyTmHDnYgF-GIUudqtPiqEgxIIgkF89pY,3549
+rio_ai/types.py,sha256=UOFf9T_uMjFMRvFsYS8RJvKPjIFMwHP2WeJFGtYsEZQ,333
+rio_coding/__init__.py,sha256=mkf3UE_wR3xk0Z-mbJCwJYDUItJIEXI23mV6JJuoBkA,8100
+rio_coding/__main__.py,sha256=DPbgfyZSWPIiTwG070jOOV_np2jNi4XEsU3_a8JKgyQ,84
+rio_coding/auth_commands.py,sha256=He4znCCEZmVM4bce43OU0T3t335C2GOEoLy6_Fy2_5k,2680
+rio_coding/branch_summary.py,sha256=hyPBzP08fnuNiD0mfMddJVeu7RkSk_fBK9G0nE0ITp4,3103
+rio_coding/built_in_extensions.py,sha256=5u8v81a0YjdZveUORIcD-uzuPkIV6F1OdCmqI5vXyCo,4075
+rio_coding/catalog_loader.py,sha256=38qVMwtuFryGX7Ja45s42dEdA5vrStksW9WMZqF58s0,29530
+rio_coding/cli.py,sha256=MRY537lBB-OmjTMtTvspAmAkZr8iFjk2MQ6LtS1IjT4,25325
+rio_coding/coding_skill.py,sha256=Xkqn8pCGnUH42kYoaZeBQh9-yO9GRVWh0A3d4gQPifk,4309
+rio_coding/commands.py,sha256=kTSBK0fQfZ_BwyDQDorfyYaNf1zXiThkQw1wagygI3Q,33331
+rio_coding/context.py,sha256=aRb8NrQxSEXBmtC92u3_Y5Zq9nPSnIgL37THHSAZIgE,3267
+rio_coding/credentials.py,sha256=5uGiUN--TB5Iw25E0jbzLSz7Tt_oUJcRrUuPq_WbcXc,9616
+rio_coding/diagnostics.py,sha256=-q9tIW1eWr905V6Gd9FVsByotFc0-6mrxTuYYsyajwk,6239
+rio_coding/events.py,sha256=scdrsJWHnSv_FbzjtD5FIvZYdNJS_eVeyzBSzbSy-G0,3465
+rio_coding/extension_installer.py,sha256=cDiqICrlBSFldqBL1LcoXR1uod-wyb6j01HQ40tPgc4,7438
+rio_coding/frontend_session.py,sha256=EwYsfJ-hs_MGAj1UF-DmmFzjtqdlkP_RmtMm1RmPpPU,9398
+rio_coding/image_processing.py,sha256=3172j-fRSjvZfUVNa_Q6ZmoJKRERdZS8FjeRFnwhnyg,9260
+rio_coding/local_backends.py,sha256=12ENUomVFRz6XjKVJVchqjiF84T3YgZdYBFBPzHQcJY,48797
+rio_coding/models_dev.py,sha256=gL-rI0h1OL7q2ZSFXdPKpIhWsWeqktn397UMWeZbotA,13111
+rio_coding/models_dev_store.py,sha256=DAejqmZRIb6xrsMr36a7G0qMrHuG1DXvDOChdxWaSzI,7187
+rio_coding/oauth.py,sha256=bwfElAQWKsW31PsNjnowkVigPomSgXG0V6Qqh-OKzkI,17462
+rio_coding/oauth_anthropic.py,sha256=YDLsyih0I0U41-o5XhiFKTkbhyoFtinDL0L0ygajJ2Y,10160
+rio_coding/oauth_device.py,sha256=D9P5FPDGak8KCF7cbDp20xWdr8FydAb_38a5hVqKY9c,3502
+rio_coding/oauth_github_copilot.py,sha256=M_vMnKHG-CK7qiGaYDbJZhaRZTVtx8U0sqAAgeB_CGw,10757
+rio_coding/oauth_registry.py,sha256=6zuyX6fZWEJ6FE5y_dYbGuiDcWYFzRYBVD2yifIcFl8,1998
+rio_coding/oauth_types.py,sha256=O1ZuCslKoEJvUi9WJlzZbax7iQ32FJo7wFT7Cz-nZMk,3556
+rio_coding/paths.py,sha256=tP--CPMsBGsuHHDxwe1cmS89dO6FM_mX6pQKPSMWm_Q,5284
+rio_coding/project_trust.py,sha256=cceOvdafJ35w8Bxx9Qat5nZPLWo2DcnwWsy_odChTZY,27518
+rio_coding/prompt_templates.py,sha256=5Wog4enRWnjdG2PxXH8Yfl4hKW-40wBdrTFBRuxOAc8,10227
+rio_coding/provider_catalog.py,sha256=RLTXavhKv5Mi-8AAM5oUYhEuIMbgH1fJVD6F5UAvLIk,4030
+rio_coding/provider_config.py,sha256=aw9uM5Qiz3-MxvUnyTqPASy1_wm4bYa2ZBT4lqytpvc,104036
+rio_coding/provider_runtime.py,sha256=SNt9oCeCv6xCPD24wErnNdEoHrcsRhHwlI2gFk0OaVs,22248
+rio_coding/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+rio_coding/reload.py,sha256=5MJ033lyL-jqHBAfEo1iR2c6y6qHrPJ2d07_tvEtJLM,789
+rio_coding/resources.py,sha256=l6aREEeAi6Umd5cY_7pz4zjwrOOT5J5UKZWzAWo7Mlg,12297
+rio_coding/rpc.py,sha256=h4rmBFkafQ6W5Yy9mHJyyByE-YQ1IYT-ZR4__cBJBn0,22723
+rio_coding/self_docs.py,sha256=jWy3exeyLABuCtobEf-nqsg5z_UnQrSpXfQk38xGWjs,613
+rio_coding/session.py,sha256=shgawEyN58mr175B1K1Io2vN3fHVfBUYb_0V-RTPiIM,29819
+rio_coding/session_export.py,sha256=-NM3ZnbPpy9yobFnlVMLBV6vWpzQyqrBLgjR4QpumYo,21888
+rio_coding/session_manager.py,sha256=PDTEYFHKaYAiTsrfhC_WdDX0gCKWVVVUf-bcKrlq-K4,14320
+rio_coding/session_preparation.py,sha256=MXYLfNgfUgyloGCwgOT9Q5_CWea1qQ36wxJExOjTgdk,2774
+rio_coding/session_runner.py,sha256=8THfX1xPzEKK0TJJ14lJq4kdRsluV_uiAvaF7fgqCoo,19211
+rio_coding/session_stats.py,sha256=RdfyxWp2bYmXwH9ROibMwRwJXh8lBpBdmeFr3r1dQTQ,3818
+rio_coding/session_usage.py,sha256=nKS-OqL5Ex30QcXST1W0r6vtrOb5RtfF1OqTA01Sxfg,7166
+rio_coding/shell_config.py,sha256=pu32bvupMZzLxd5OaQlg50ePzjML1vfJWCPanJpevUM,2843
+rio_coding/skills.py,sha256=yeWNZhfZVvFch5d4sEs0ZCNkomsCOr6UkrF2Fkno6mo,8959
+rio_coding/step_footprint.py,sha256=FO35WoocS0nDF5QbWb9m6hJpdgYvBQISiUoOV_k6MC8,5534
+rio_coding/system_prompt.py,sha256=bii7CgiDEmnhvnZseO2TFLPvSO90ux25j-k1zfKS_kw,13357
+rio_coding/thinking.py,sha256=aknZSnW7JthyGMP8i2GPy0bNbD799w6FyvCUaVoak8w,3044
+rio_coding/tools.py,sha256=EapKdE-tGdj8GDxJpAbVo7v7WB1G5dtx45cRIcj8HZg,51952
+rio_coding/update_check.py,sha256=CIrkv_p1GH5XxISHlZdTlkpKB8V_eKrw-hYyEVsu5pc,13075
+rio_coding/updater.py,sha256=Ud0rwdAhmu20ReQYekA-RVOn84bFgeO7Jy4o5xfpFeg,13491
+rio_coding/version.py,sha256=_jG2JtOXwfNuug0a0wfc4zKUvquFYUnDFH6FqijYfO4,409
+rio_coding/data/catalog.toml,sha256=WdFGv6xpyPkV0bqlre0CPK2aDf7fL3__YF6DXm4vgIo,233789
+rio_coding/data/models-dev-catalog.json,sha256=TMjIGXnQqIzO4W-SJZ19DMWmhwttCKOZG9gTxOeJA4Y,403853
+rio_coding/data/docs/README.md,sha256=UgaiPH8_leXVd-4CNe1cZycQTIe5qsiwDzLxbwSryns,781
+rio_coding/data/docs/architecture.md,sha256=yDrfpfTA1iw-DCbCd7rVenaY7qKCITEt9Rx4Tyh_vw8,5843
+rio_coding/data/docs/cli.md,sha256=9m5ZbnDLTJUvZah8lxxAY-sylb46oj2sNe9Fs5_r1yg,2637
+rio_coding/data/docs/extensions.md,sha256=td-Dpd-SvdePi3vN0YRvAtx4op6rl2M9TQS81QUUhjk,2436
+rio_coding/data/docs/local-inference.md,sha256=4vGaelExDf3e2GgA7ddNZ2xV99ClCm4LYEB0OKsrg7s,970
+rio_coding/data/docs/models.md,sha256=NBuNVt1STerRmcIx-5PLpXleSmCSWoRBwSeRaJZRNN4,2187
+rio_coding/data/docs/security.md,sha256=nBiIyaDphF4bEUxn08mAWlkcNWSRLMJI2K5ItCeP1bw,1452
+rio_coding/data/docs/skills.md,sha256=lgBUjiJ7w8J7iogVmIRaAGa-zCKYNuNZWTUM4dBRDhQ,2142
+rio_coding/data/examples/offline_session.py,sha256=wlhHKNueupqg8rh0MwqS6891IqsEVMbmzqJmm1jJpwk,1478
+rio_coding/data/release-notes/releases.json,sha256=EpVsjjsE8s9rJaAdznyx38p3pAOHMRj9eBcAkk2lLXY,1086
+rio_coding/extensions/__init__.py,sha256=Wksp3cLcH3uao4sxK8t7IOqrNfFt-sl8aG5-mgWnwjM,5756
+rio_coding/extensions/api.py,sha256=SAWymZX8EPRCT3hkV-sKycqLK8QGDkH4E6uC8yW98jc,42069
+rio_coding/extensions/loader.py,sha256=0FNF6Umx_UjICuxflW7BitUTtjvyhqZTNJZGnMXPQIA,14221
+rio_coding/extensions/provider_registry.py,sha256=yW0hnvp8u39EIBU2zy7Zo2P3-4DkHdXZ64hM2uq5HMk,25950
+rio_coding/extensions/providers.py,sha256=WKyuiMgtLSDzF_a3xMxpKvZRVZ8OaWxZAusdA6hE2d0,21826
+rio_coding/extensions/runtime.py,sha256=gNY_-HCY4wWN8m3irMc2diCI95Uj-9rOf-sY6QceN2E,51672
+rio_coding/extensions/startup.py,sha256=hWlRdqfobmNjDQL72owXCgP_-LFuGIUoCnszoN5CLYs,2438
+rio_coding/extensions/builtins/__init__.py,sha256=8RwaQsjE6IUr8qeUxamm_oYp-xeV-Tddp4_Ncn4BSBk,41
+rio_coding/extensions/builtins/llama_cpp/__init__.py,sha256=_RdD-RmlLjjwP5So428wdg3LbI4H-BMuSF9w5CfDiI4,2555
+rio_coding/extensions/builtins/llama_cpp/huggingface.py,sha256=oem3mZgAzgPWvlrLxtXsI64vLWlgfyGMohxtRe3ADZQ,7511
+rio_coding/extensions/builtins/llama_cpp/router.py,sha256=QgEsdFKYc82Nf9LaZmtWxnE35E-gmdeCMUws5WuVjTA,10639
+rio_coding/extensions/builtins/llama_cpp/service.py,sha256=8BPgxNcixo189EqWsvxdFc1mM9WGuesRF5H5vAFN8TU,71014
+rio_coding/extensions/builtins/llama_cpp/state.py,sha256=w4wklCBOB4ujGWhflmy6E0dhM8wbWPpgx07LfnK85QY,17333
+rio_coding/rendering/__init__.py,sha256=oibCD3w0YTu3--f2r2AwiKJ8Cc1WysgemHQHI1rIUJc,877
+rio_coding/rendering/base.py,sha256=g7Ag16orXE4OPAYCs8LR3vfvdFwhVImI_QMAEU0LTqs,1802
+rio_coding/rendering/json.py,sha256=ImYpMA8y9nJjkIAPpI1ZoQRo05e6iSFX3_gynJiyGlw,2654
+rio_coding/rendering/plain.py,sha256=_FSOaAbtJpP2fhHje2K4lcIIIzZE_uMQN4qnLRFs5LA,3909
+rio_coding/rendering/steps.py,sha256=lV4gBepQ2r42O3EXh3RhZLgg0-BOA7xZySN8o6BTYaE,4539
+rio_coding/session_store/__init__.py,sha256=wvn1HrS29bvJUsl_yRqyod4COzfgyP4sGJQOHlgN0Z4,1308
+rio_coding/session_store/entries.py,sha256=2xqd5VdCmqGWMCgoEhX_l3PoySyF7xrUFWKjjltNhSE,5564
+rio_coding/session_store/jsonl.py,sha256=PXilexk8J9d4Da92yECZTesHG51t-HMd43iqcrgewAo,1431
+rio_coding/session_store/storage.py,sha256=w2zBo7DBt2v8CF_11S00296imoSYODWBeVP7QX8lZ2k,7344
+rio_coding/session_store/tree.py,sha256=nbZ4nNVbehV52VVT0HKovc3mxTm3FsEEXLDK4_RasJg,3265
+rio_tui/__init__.py,sha256=GURtvbciFTnpf1TBIalK8hT7j3hvKP-V4-2okKYylIQ,191
+rio_tui/__main__.py,sha256=Z-FokKnWybtUCIyaU_-1RyMwMN2E1jZqcON1RtBmVC8,38
+rio_tui/app.py,sha256=SqK9NCINBFGEY8y_HTVid705GAIX1v_JwN3EBwmwZfc,8360
+rio_tui/app.tcss,sha256=ZYamNhJ2yii75DItVUYbiLJnG8RbyStNAtvvVKeV8Uw,4085
+rio_tui/bridge.py,sha256=s0m6miRd_zSY5nxVzr3d_97qun2b2TPcsWBAuTCLYu0,7653
+rio_tui/dialogs.py,sha256=a57UVyWOhqykBnO-J4i30NOO8yzvo5uFEMMLVSeGpAU,9298
+rio_tui/session.py,sha256=TQMNDyQnCz1UDizCHTqCusohaUVT76-aGQiF7NJcxoM,21705
+rio_tui/settings.py,sha256=CXmn7SN0ZmtQumz74EiLoGx2D1VI7HfU9yTq1G9RhyI,1533
+rio_tui/widgets/__init__.py,sha256=PsSHMD7Gm75fDIplUpoPWa0WVacHQIgzgtFoX0LajZE,40
+rio_tui/widgets/conversation.py,sha256=ZwvsLJs9E51Rho503xOLzJzw1bpRbVjOGTMuzW-B5lQ,4549
+rio_tui/widgets/prompt.py,sha256=nVss9GFaEypfOIt4gfVHZojU32on316lFpr9ngGJ5Q8,6566
+rio_tui/widgets/sidebar.py,sha256=NKL5pGMt6QHrDjv0BXVyhr9eBsMA7DM9X2R0kpcPb1Y,1943
+rio_tui/widgets/terminal.py,sha256=Ip0uKtdsWrLQphncVJigsFVeZeGQYTno6rM-5VIXKoU,6447
+rio-0.4.0.dist-info/METADATA,sha256=SY1fIe9X7VEMc2Z8Bt6u4L8EQsR_iI-DNLZ733ANtZo,4108
+rio-0.4.0.dist-info/WHEEL,sha256=zOwg4jB6zX2kU910N-cMawjivD6tO8NEWvE12je1bVk,87
+rio-0.4.0.dist-info/entry_points.txt,sha256=oGugo6Dvgb9fDiTC3zoUNQWZ_GRnkcUoCl51W00FoAU,43
+rio-0.4.0.dist-info/licenses/LICENSE,sha256=-PLaKGZFrtCpEyu6v0yJCWrR95hROnRQ3eEh0Fq2_Wg,1063
+rio-0.4.0.dist-info/licenses/NOTICE,sha256=lQVrQEsQH8CumOXq48YYUSXAMAMOz0CdqKjxT7rmYk4,1856
+rio-0.4.0.dist-info/RECORD,,

rio-0.4.0.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: hatchling 1.32.0
+Root-Is-Purelib: true
+Tag: py3-none-any

rio-0.4.0.dist-info/entry_points.txt

@@ -0,0 +1,2 @@
+[console_scripts]
+rio = rio_coding.cli:app

rio-0.4.0.dist-info/licenses/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Ju Lin
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

rio-0.4.0.dist-info/licenses/NOTICE

@@ -0,0 +1,37 @@
+rio_ai (src/rio_ai/) is a port of huggingface/tau's `tau_ai` module, plus the
+`tau_agent` message/tool/type vocabulary that `tau_ai` depends on
+(`messages.py`, `tools.py`, `types.py`, `provider.py`, `provider_events.py`).
+
+rio_coding (src/rio_coding/) is a port of huggingface/tau's `tau_coding`
+module, retargeted from `tau_agent`'s append-only conversational harness onto
+`rio_agent`'s SKILL.state runtime. Modules whose purpose was serving a growing
+transcript -- context compaction, transcript rendering and export, and the
+message-history session log -- were redesigned around execution state rather
+than carried over. `rio_coding/session_store/` derives from `tau_agent.session`.
+
+Upstream project: https://github.com/huggingface/tau
+Upstream license: MIT, reproduced below.
+
+---
+
+MIT License
+
+Copyright (c) 2026 Alejandro AO
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

rio_agent/__init__.py

@@ -0,0 +1,47 @@
+"""rio_agent: a SKILL.state long-horizon agent runtime.
+
+Fixed skill instructions, a structured mutable execution state, and the
+latest observation -- an action's result, a user message, or both -- are the
+only inputs to each step. The runtime discards the reasoning behind each
+step once it commits a valid state update, so per-step prompt size stays
+fixed instead of growing with the number of steps already taken.
+
+Exposes a loop interface shaped like `rio_ai`'s ported tau_agent-style
+`AgentHarness`/`run_agent_loop` (a stateful harness plus a bare async-
+generator loop function, both emitting a typed event stream), but backed
+internally by execution state rather than an append-only transcript.
+"""
+
+# ruff: noqa: F401 - this module intentionally defines the public facade
+
+from rio_agent.errors import (
+    ActionNotFoundError,
+    ProviderResponseError,
+    RetriesExhaustedError,
+    StateValidationError,
+)
+from rio_agent.events import (
+    ActionEndEvent,
+    ActionStartEvent,
+    ReasoningDiscardedEvent,
+    RunEndEvent,
+    RunStartEvent,
+    SkillEvent,
+    StateUpdateEvent,
+    StepEndEvent,
+    StepStartEvent,
+    ValidationErrorEvent,
+)
+from rio_agent.harness import (
+    EventListener,
+    Harness,
+    HarnessCancellationToken,
+    HarnessConfig,
+)
+from rio_agent.loop import run_skill_loop
+from rio_agent.observation import HarnessObservation
+from rio_agent.prompt import STEP_TOOL_NAME, build_step_messages, skill_step_tool
+from rio_agent.skill import HarnessSpec
+from rio_agent.state import apply_state_delta, validate_state_delta
+
+__all__ = [name for name in globals() if not name.startswith("_")]

rio_agent/errors.py

@@ -0,0 +1,29 @@
+"""Errors raised while validating and committing SKILL.state step output."""
+
+from __future__ import annotations
+
+
+class StateValidationError(Exception):
+    """A proposed state_delta failed deterministic runtime validation.
+
+    Validation happens in the runtime, not the model, so a malformed patch
+    never corrupts persistent state -- it triggers a rollback-retry cycle
+    instead (see `RetriesExhaustedError`).
+    """
+
+
+class ActionNotFoundError(Exception):
+    """The model proposed an action that isn't declared on the skill."""
+
+
+class RetriesExhaustedError(Exception):
+    """The rollback-retry cycle exceeded `max_retries` without a valid step."""
+
+
+class ProviderResponseError(Exception):
+    """The provider failed to produce an assistant message.
+
+    Distinct from a malformed step: the model never got to propose one, so
+    there is nothing to roll back and nothing a retry of the same prompt would
+    correct.
+    """

rio_agent/events.py

@@ -0,0 +1,95 @@
+"""Events emitted by `rio_agent.loop.run_skill_loop`.
+
+Shaped like `rio_ai`'s ported tau_agent-style agent events (a flat stream of
+small, typed records a UI or logger can subscribe to) but describing SKILL.
+state's step lifecycle instead of an append-only conversation turn.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+from rio_agent.observation import HarnessObservation
+from rio_ai.tools import AgentToolResult
+from rio_ai.types import JSONObject
+
+
+@dataclass(frozen=True, slots=True)
+class RunStartEvent:
+    skill: str
+
+
+@dataclass(frozen=True, slots=True)
+class StepStartEvent:
+    step: int
+    state: JSONObject
+    observation: HarnessObservation
+
+
+@dataclass(frozen=True, slots=True)
+class ReasoningDiscardedEvent:
+    """The model's reasoning for this step. Surfaced once for observability, then never sent back
+    to the model.
+    """
+
+    step: int
+    reasoning: str
+
+
+@dataclass(frozen=True, slots=True)
+class ValidationErrorEvent:
+    """A proposed state update or action failed runtime validation; a rollback-retry follows."""
+
+    step: int
+    attempt: int
+    error: str
+
+
+@dataclass(frozen=True, slots=True)
+class StateUpdateEvent:
+    """The proposed state update was validated and committed to form the new state."""
+
+    step: int
+    delta: JSONObject
+    state: JSONObject
+
+
+@dataclass(frozen=True, slots=True)
+class ActionStartEvent:
+    step: int
+    name: str
+    arguments: JSONObject
+
+
+@dataclass(frozen=True, slots=True)
+class ActionEndEvent:
+    step: int
+    name: str
+    result: AgentToolResult
+    is_error: bool
+
+
+@dataclass(frozen=True, slots=True)
+class StepEndEvent:
+    step: int
+    state: JSONObject
+    terminated: bool
+
+
+@dataclass(frozen=True, slots=True)
+class RunEndEvent:
+    steps: int
+    state: JSONObject
+
+
+type SkillEvent = (
+    RunStartEvent
+    | StepStartEvent
+    | ReasoningDiscardedEvent
+    | ValidationErrorEvent
+    | StateUpdateEvent
+    | ActionStartEvent
+    | ActionEndEvent
+    | StepEndEvent
+    | RunEndEvent
+)

rio_agent/harness.py

@@ -0,0 +1,125 @@
+"""Harness: a reusable stateful runtime around `run_skill_loop`.
+
+Deliberately mirrors the public shape of `rio_ai`'s ported tau_agent-style
+`AgentHarness` -- construct with a config, `subscribe()` an event listener,
+call `run()`/`prompt()` to drive it, `cancel()` to stop it -- so callers
+already familiar with that loop interface can pick this one up directly.
+The difference is entirely internal: instead of holding a growing message
+transcript, this harness holds only the current execution state and
+replays `run_skill_loop`'s fixed-size per-step prompt.
+"""
+
+from __future__ import annotations
+
+from collections.abc import AsyncIterator, Awaitable, Callable
+from contextlib import suppress
+from dataclasses import dataclass
+from inspect import isawaitable
+
+from rio_agent.events import RunEndEvent, SkillEvent, StateUpdateEvent
+from rio_agent.loop import run_skill_loop
+from rio_agent.observation import HarnessObservation
+from rio_agent.skill import HarnessSpec
+from rio_ai.provider import ModelProvider
+from rio_ai.types import JSONObject, JSONValue
+
+EventListener = Callable[[SkillEvent], Awaitable[None] | None]
+
+
+@dataclass(slots=True)
+class HarnessConfig:
+    provider: ModelProvider
+    model: str
+    skill: HarnessSpec
+    max_steps: int | None = None
+    max_retries: int = 2
+
+
+class HarnessCancellationToken:
+    def __init__(self) -> None:
+        self._cancelled = False
+
+    def cancel(self) -> None:
+        self._cancelled = True
+
+    def is_cancelled(self) -> bool:
+        return self._cancelled
+
+
+class Harness:
+    """Reusable stateful long-horizon agent runtime built on SKILL.state."""
+
+    def __init__(
+        self,
+        config: HarnessConfig,
+        *,
+        state: JSONObject | None = None,
+    ) -> None:
+        self._config = config
+        self._state: JSONObject = dict(state if state is not None else config.skill.initial_state)
+        self._listeners: list[EventListener] = []
+        self._current_signal: HarnessCancellationToken | None = None
+        self._running = False
+
+    @property
+    def state(self) -> dict[str, JSONValue]:
+        return dict(self._state)
+
+    @property
+    def config(self) -> HarnessConfig:
+        return self._config
+
+    @property
+    def is_running(self) -> bool:
+        return self._running
+
+    def subscribe(self, listener: EventListener) -> Callable[[], None]:
+        self._listeners.append(listener)
+
+        def unsubscribe() -> None:
+            with suppress(ValueError):
+                self._listeners.remove(listener)
+
+        return unsubscribe
+
+    def cancel(self) -> None:
+        if self._current_signal is not None:
+            self._current_signal.cancel()
+
+    def run(self, observation: HarnessObservation) -> AsyncIterator[SkillEvent]:
+        self._ensure_not_running()
+        self._running = True
+        return self._run(observation)
+
+    async def _run(self, observation: HarnessObservation) -> AsyncIterator[SkillEvent]:
+        signal = HarnessCancellationToken()
+        self._current_signal = signal
+        try:
+            async for event in run_skill_loop(
+                provider=self._config.provider,
+                model=self._config.model,
+                skill=self._config.skill,
+                observation=observation,
+                state=self._state,
+                max_steps=self._config.max_steps,
+                max_retries=self._config.max_retries,
+                signal=signal,
+            ):
+                if isinstance(event, (StateUpdateEvent, RunEndEvent)):
+                    self._state = dict(event.state)
+                await self._notify(event)
+                yield event
+        finally:
+            if self._current_signal is signal:
+                self._current_signal = None
+            self._running = False
+
+    async def _notify(self, event: SkillEvent) -> None:
+        for listener in list(self._listeners):
+            result = listener(event)
+            if isawaitable(result):
+                await result
+
+    def _ensure_not_running(self) -> None:
+        if self._running:
+            raise RuntimeError("Harness is already running")