AdityaProCoder commited on
Commit
78ab82d
·
1 Parent(s): af8c4cf

Add OpenEnv hackathon submission docs and structure guidance

Browse files
.env.example ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ # OpenAI API key
2
+ OPENAI_API_KEY=
3
+
4
+ # Google Vertex AI
5
+ VERTEX_PROJECT_ID=
6
+ VERTEX_LOCATION=us-central1
7
+
8
+ # Optional model/runtime overrides
9
+ MODEL_NAME=gpt-4o-mini
10
+ API_BASE_URL=https://api.openai.com/v1
.gitignore CHANGED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Environment files
2
+ .env
3
+ .env.*
4
+ !.env.example
5
+
6
+ # Virtual environments
7
+ .venv/
8
+ venv/
9
+
10
+ # Python cache/bytecode
11
+ __pycache__/
12
+ *.py[cod]
13
+ *$py.class
14
+
15
+ # Tool caches
16
+ .pytest_cache/
17
+ .mypy_cache/
18
+ .ruff_cache/
19
+
20
+ # Packaging/build artifacts
21
+ *.egg-info/
22
+ build/
23
+ dist/
Dockerfile CHANGED
@@ -1,29 +1,35 @@
1
  # OpenEnv/FastAPI runtime image for AutonomousReturns-v0
2
- FROM python:3.10-slim
3
 
4
  WORKDIR /app
5
 
6
  # Runtime safety + cleaner logs
7
  ENV PYTHONDONTWRITEBYTECODE=1 \
8
  PYTHONUNBUFFERED=1 \
 
9
  PIP_NO_CACHE_DIR=1 \
10
  ENABLE_WEB_INTERFACE=false \
11
  PORT=8000
12
 
13
  # Minimal OS deps + healthcheck utility
14
  RUN apt-get update && apt-get install -y --no-install-recommends \
15
- curl \
16
  && rm -rf /var/lib/apt/lists/*
17
 
18
- # Copy package metadata + source before install (required for successful build)
19
- COPY pyproject.toml README.md ./
 
 
 
 
 
20
  COPY autonomous_returns_v0 ./autonomous_returns_v0
21
  COPY server ./server
22
  COPY openenv.yaml ./openenv.yaml
 
23
 
24
- # Install package and dependencies
25
- RUN pip install --upgrade pip setuptools wheel && \
26
- pip install .
27
 
28
  EXPOSE 8000
29
 
@@ -32,4 +38,4 @@ HEALTHCHECK --interval=30s --timeout=3s --start-period=10s --retries=3 \
32
  CMD curl -fsS http://localhost:8000/health || exit 1
33
 
34
  # Start OpenEnv-compatible FastAPI app
35
- CMD ["uvicorn", "autonomous_returns_v0.server.app:app", "--host", "0.0.0.0", "--port", "8000"]
 
1
  # OpenEnv/FastAPI runtime image for AutonomousReturns-v0
2
+ FROM python:3.12-slim
3
 
4
  WORKDIR /app
5
 
6
  # Runtime safety + cleaner logs
7
  ENV PYTHONDONTWRITEBYTECODE=1 \
8
  PYTHONUNBUFFERED=1 \
9
+ PATH="/app/.venv/bin:${PATH}" \
10
  PIP_NO_CACHE_DIR=1 \
11
  ENABLE_WEB_INTERFACE=false \
12
  PORT=8000
13
 
14
  # Minimal OS deps + healthcheck utility
15
  RUN apt-get update && apt-get install -y --no-install-recommends \
16
+ curl ca-certificates \
17
  && rm -rf /var/lib/apt/lists/*
18
 
19
+ # Install uv for lockfile-based environment sync
20
+ RUN pip install --upgrade pip setuptools wheel uv
21
+
22
+ # Copy metadata/lock first to maximize Docker layer caching
23
+ COPY pyproject.toml uv.lock README.md ./
24
+
25
+ # Copy package source
26
  COPY autonomous_returns_v0 ./autonomous_returns_v0
27
  COPY server ./server
28
  COPY openenv.yaml ./openenv.yaml
29
+ COPY run_baseline.py run_server.py ./
30
 
31
+ # Install dependencies and project exactly from lockfile
32
+ RUN uv sync --frozen --no-dev
 
33
 
34
  EXPOSE 8000
35
 
 
38
  CMD curl -fsS http://localhost:8000/health || exit 1
39
 
40
  # Start OpenEnv-compatible FastAPI app
41
+ CMD ["uvicorn", "server.app:app", "--host", "0.0.0.0", "--port", "8000"]
README.md ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ # AutonomousReturns-v0
2
+
3
+ OpenEnv-compatible reverse-logistics environment for e-commerce returns triage.
4
+
5
+ - Package docs: `autonomous_returns_v0/README.md`
6
+ - Server entrypoint: `server.app:app`
7
+ - CLI scripts: `server`, `run-server`, `baseline`
autonomous_returns_v0/.env.example CHANGED
@@ -1,9 +1,6 @@
1
  # OpenAI API key (get from https://platform.openai.com/api-keys)
2
  OPENAI_API_KEY=your_openai_key_here
3
 
4
- # Anthropic API key (get from https://console.anthropic.com/settings/keys)
5
- ANTHROPIC_API_KEY=your_anthropic_key_here
6
-
7
  # Google Vertex AI (use `gcloud auth application-default login` for auth)
8
  VERTEX_PROJECT_ID=your-gcp-project-id
9
  VERTEX_LOCATION=us-central1
 
1
  # OpenAI API key (get from https://platform.openai.com/api-keys)
2
  OPENAI_API_KEY=your_openai_key_here
3
 
 
 
 
4
  # Google Vertex AI (use `gcloud auth application-default login` for auth)
5
  VERTEX_PROJECT_ID=your-gcp-project-id
6
  VERTEX_LOCATION=us-central1
autonomous_returns_v0/README.md CHANGED
@@ -137,7 +137,8 @@ Optional:
137
 
138
  ```bash
139
  pip install ".[openai]"
140
- pip install ".[anthropic]"
 
141
  pip install ".[all]"
142
  ```
143
 
@@ -164,7 +165,7 @@ uv run run-server
164
  Runtime contract validation:
165
 
166
  ```bash
167
- uvicorn autonomous_returns_v0.server.app:app --host 127.0.0.1 --port 8000
168
  openenv validate --url http://127.0.0.1:8000
169
  ```
170
 
@@ -220,10 +221,10 @@ Deterministic local baseline (this branch, current implementation):
220
 
221
  ## Optional LLM Baseline
222
 
223
- OpenAI/Anthropic baseline runner is available but non-deterministic over time:
224
 
225
  - `OPENAI_API_KEY`
226
- - `ANTHROPIC_API_KEY`
227
 
228
  Example:
229
 
 
137
 
138
  ```bash
139
  pip install ".[openai]"
140
+ pip install ".[vertex]"
141
+ pip install ".[web]"
142
  pip install ".[all]"
143
  ```
144
 
 
165
  Runtime contract validation:
166
 
167
  ```bash
168
+ uvicorn server.app:app --host 127.0.0.1 --port 8000
169
  openenv validate --url http://127.0.0.1:8000
170
  ```
171
 
 
221
 
222
  ## Optional LLM Baseline
223
 
224
+ OpenAI/Google baseline runner is available but non-deterministic over time:
225
 
226
  - `OPENAI_API_KEY`
227
+ - `VERTEX_PROJECT_ID`
228
 
229
  Example:
230
 
autonomous_returns_v0/baseline.py CHANGED
@@ -5,7 +5,7 @@ Features:
5
  - OpenEnv-compatible typed actions (`ReturnsAction`)
6
  - Robust handling of delayed-reward tail steps (`WAIT`)
7
  - Reproducible benchmark runner across easy/medium/hard tasks
8
- - OpenAI/Anthropic/Vertex AI model support with env-var credentials
9
  """
10
 
11
  from __future__ import annotations
@@ -97,21 +97,6 @@ class BaselineAgent:
97
  if not self.model:
98
  self.model = os.environ.get("MODEL_NAME", "gpt-4o-mini")
99
 
100
- elif self.provider == "anthropic":
101
- try:
102
- anthropic_mod = import_module("anthropic")
103
- Anthropic = getattr(anthropic_mod, "Anthropic")
104
- except Exception as exc:
105
- raise ImportError(
106
- "anthropic package required. Install with: pip install anthropic"
107
- ) from exc
108
-
109
- api_key = os.environ.get("ANTHROPIC_API_KEY")
110
- if not api_key:
111
- raise ValueError("ANTHROPIC_API_KEY environment variable not set")
112
- self.client = Anthropic(api_key=api_key)
113
- if not self.model:
114
- self.model = "claude-3-5-haiku-20241022"
115
  elif self.provider == "google":
116
  project_id = os.environ.get("VERTEX_PROJECT_ID")
117
  if not project_id:
@@ -121,7 +106,7 @@ class BaselineAgent:
121
  if not self.model:
122
  self.model = "gemini-1.5-pro"
123
  else:
124
- raise ValueError("Unknown provider. Use 'openai', 'anthropic', or 'google'.")
125
 
126
  def _call_llm(self, prompt: str) -> str:
127
  system_msg = LLM_SYSTEM_PROMPT
@@ -139,15 +124,6 @@ class BaselineAgent:
139
  text = response.choices[0].message.content or ""
140
  return text.strip()
141
 
142
- if self.provider == "anthropic":
143
- response = self.client.messages.create(
144
- model=self.model,
145
- max_tokens=self.config.max_tokens,
146
- system=system_msg,
147
- messages=[{"role": "user", "content": prompt}],
148
- )
149
- return response.content[0].text.strip()
150
-
151
  init(project=self._vertex_project, location=self._vertex_location)
152
  model = GenerativeModel(
153
  self.model,
@@ -318,17 +294,6 @@ def run_openai_baseline(
318
  ) -> dict:
319
  agent = BaselineAgent(BaselineRunConfig(provider="openai", model=model))
320
  return agent.run_benchmark(seeds=seeds, verbose=verbose)
321
-
322
-
323
- def run_anthropic_baseline(
324
- model: str = "claude-3-5-haiku-20241022",
325
- seeds: Iterable[int] = (42,),
326
- verbose: bool = False,
327
- ) -> dict:
328
- agent = BaselineAgent(BaselineRunConfig(provider="anthropic", model=model))
329
- return agent.run_benchmark(seeds=seeds, verbose=verbose)
330
-
331
-
332
  if __name__ == "__main__":
333
  result = run_openai_baseline(verbose=True)
334
- print(result)
 
5
  - OpenEnv-compatible typed actions (`ReturnsAction`)
6
  - Robust handling of delayed-reward tail steps (`WAIT`)
7
  - Reproducible benchmark runner across easy/medium/hard tasks
8
+ - OpenAI/Vertex AI model support with env-var credentials
9
  """
10
 
11
  from __future__ import annotations
 
97
  if not self.model:
98
  self.model = os.environ.get("MODEL_NAME", "gpt-4o-mini")
99
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
100
  elif self.provider == "google":
101
  project_id = os.environ.get("VERTEX_PROJECT_ID")
102
  if not project_id:
 
106
  if not self.model:
107
  self.model = "gemini-1.5-pro"
108
  else:
109
+ raise ValueError("Unknown provider. Use 'openai' or 'google'.")
110
 
111
  def _call_llm(self, prompt: str) -> str:
112
  system_msg = LLM_SYSTEM_PROMPT
 
124
  text = response.choices[0].message.content or ""
125
  return text.strip()
126
 
 
 
 
 
 
 
 
 
 
127
  init(project=self._vertex_project, location=self._vertex_location)
128
  model = GenerativeModel(
129
  self.model,
 
294
  ) -> dict:
295
  agent = BaselineAgent(BaselineRunConfig(provider="openai", model=model))
296
  return agent.run_benchmark(seeds=seeds, verbose=verbose)
 
 
 
 
 
 
 
 
 
 
 
297
  if __name__ == "__main__":
298
  result = run_openai_baseline(verbose=True)
299
+ print(result)
pyproject.toml CHANGED
@@ -11,7 +11,6 @@ requires-python = ">=3.10"
11
  dependencies = [
12
  "openenv-core[core]>=0.2.2",
13
  "fastapi>=0.115.0",
14
- "gradio>=6.0.0",
15
  "pydantic>=2.0.0",
16
  "uvicorn>=0.24.0",
17
  "numpy>=1.20",
@@ -22,20 +21,20 @@ dependencies = [
22
 
23
  [project.optional-dependencies]
24
  openai = ["openai>=1.0"]
25
- anthropic = ["anthropic>=0.18"]
26
  vertex = ["google-cloud-aiplatform>=1.145.0"]
27
- all = ["openai>=1.0", "anthropic>=0.18", "google-cloud-aiplatform>=1.145.0"]
 
28
  dev = [
29
  "pytest>=8.0.0",
30
  "pytest-cov>=4.0.0",
31
  ]
32
 
33
  [project.scripts]
34
- server = "autonomous_returns_v0.server.app:main"
35
  run-server = "run_server:main"
36
  baseline = "run_baseline:main"
37
 
38
  [tool.setuptools]
39
  include-package-data = true
40
- packages = ["autonomous_returns_v0", "autonomous_returns_v0.server"]
41
  py-modules = ["run_baseline", "run_server"]
 
11
  dependencies = [
12
  "openenv-core[core]>=0.2.2",
13
  "fastapi>=0.115.0",
 
14
  "pydantic>=2.0.0",
15
  "uvicorn>=0.24.0",
16
  "numpy>=1.20",
 
21
 
22
  [project.optional-dependencies]
23
  openai = ["openai>=1.0"]
 
24
  vertex = ["google-cloud-aiplatform>=1.145.0"]
25
+ web = ["gradio>=6.0.0"]
26
+ all = ["openai>=1.0", "google-cloud-aiplatform>=1.145.0", "gradio>=6.0.0"]
27
  dev = [
28
  "pytest>=8.0.0",
29
  "pytest-cov>=4.0.0",
30
  ]
31
 
32
  [project.scripts]
33
+ server = "server.app:main"
34
  run-server = "run_server:main"
35
  baseline = "run_baseline:main"
36
 
37
  [tool.setuptools]
38
  include-package-data = true
39
+ packages = ["autonomous_returns_v0", "server"]
40
  py-modules = ["run_baseline", "run_server"]
run_baseline.py CHANGED
@@ -28,15 +28,13 @@ def _parse_int_csv(value: str) -> list[int]:
28
 
29
 
30
  def _auto_detect_provider() -> str:
31
- """Auto-detect LLM provider based on available API keys. Priority: openai > anthropic > google."""
32
  if os.getenv("OPENAI_API_KEY"):
33
  return "openai"
34
- if os.getenv("ANTHROPIC_API_KEY"):
35
- return "anthropic"
36
  if os.getenv("GEMINI_API_KEY"):
37
  return "google"
38
  raise ValueError(
39
- "No LLM API key found. Set OPENAI_API_KEY, ANTHROPIC_API_KEY, or GEMINI_API_KEY in .env"
40
  )
41
 
42
 
@@ -62,14 +60,14 @@ def build_parser() -> argparse.ArgumentParser:
62
  parser = argparse.ArgumentParser(
63
  description=(
64
  "AutonomousReturns-v0 baseline CLI. "
65
- "Supports deterministic local baseline and optional LLM baseline."
66
  )
67
  )
68
  parser.add_argument(
69
  "--mode",
70
  choices=("local", "llm"),
71
  default="local",
72
- help="Execution mode: local (deterministic) or llm (auto/OpenAI/Anthropic/Google).",
73
  )
74
  parser.add_argument(
75
  "--tasks",
@@ -83,7 +81,7 @@ def build_parser() -> argparse.ArgumentParser:
83
  )
84
  parser.add_argument(
85
  "--provider",
86
- choices=("auto", "openai", "anthropic", "google"),
87
  default="auto",
88
  help="LLM provider (used only in --mode llm). Use 'auto' to detect from API keys.",
89
  )
 
28
 
29
 
30
  def _auto_detect_provider() -> str:
31
+ """Auto-detect LLM provider based on available API keys. Priority: openai > google."""
32
  if os.getenv("OPENAI_API_KEY"):
33
  return "openai"
 
 
34
  if os.getenv("GEMINI_API_KEY"):
35
  return "google"
36
  raise ValueError(
37
+ "No LLM API key found. Set OPENAI_API_KEY or GEMINI_API_KEY in .env"
38
  )
39
 
40
 
 
60
  parser = argparse.ArgumentParser(
61
  description=(
62
  "AutonomousReturns-v0 baseline CLI. "
63
+ "Supports deterministic local baseline and optional LLM baseline (OpenAI/Google)."
64
  )
65
  )
66
  parser.add_argument(
67
  "--mode",
68
  choices=("local", "llm"),
69
  default="local",
70
+ help="Execution mode: local (deterministic) or llm (auto/OpenAI/Google).",
71
  )
72
  parser.add_argument(
73
  "--tasks",
 
81
  )
82
  parser.add_argument(
83
  "--provider",
84
+ choices=("auto", "openai", "google"),
85
  default="auto",
86
  help="LLM provider (used only in --mode llm). Use 'auto' to detect from API keys.",
87
  )
run_server.py CHANGED
@@ -1,4 +1,4 @@
1
- """Single-command launcher for local OpenEnv + Gradio Web UI."""
2
 
3
  from __future__ import annotations
4
 
@@ -15,7 +15,7 @@ def main() -> None:
15
  os.environ.setdefault("ENV_README_PATH", os.path.abspath(os.path.join(os.path.dirname(__file__), "README.md")))
16
  port = int(os.getenv("PORT", "8000"))
17
 
18
- app_module = import_module("autonomous_returns_v0.server.app")
19
  app = getattr(app_module, "build_app")()
20
 
21
  print(f"Web UI: http://localhost:{port}/web/")
 
1
+ """Single-command launcher for local OpenEnv API + optional web UI."""
2
 
3
  from __future__ import annotations
4
 
 
15
  os.environ.setdefault("ENV_README_PATH", os.path.abspath(os.path.join(os.path.dirname(__file__), "README.md")))
16
  port = int(os.getenv("PORT", "8000"))
17
 
18
+ app_module = import_module("server.app")
19
  app = getattr(app_module, "build_app")()
20
 
21
  print(f"Web UI: http://localhost:{port}/web/")
server/Dockerfile CHANGED
@@ -32,4 +32,4 @@ HEALTHCHECK --interval=30s --timeout=3s --start-period=10s --retries=3 \
32
  CMD curl -fsS http://localhost:8000/health || exit 1
33
 
34
  # OpenEnv-compatible FastAPI server
35
- CMD ["uvicorn", "autonomous_returns_v0.server.app:app", "--host", "0.0.0.0", "--port", "8000"]
 
32
  CMD curl -fsS http://localhost:8000/health || exit 1
33
 
34
  # OpenEnv-compatible FastAPI server
35
+ CMD ["uvicorn", "server.app:app", "--host", "0.0.0.0", "--port", "8000"]
server/__init__.py CHANGED
@@ -5,5 +5,5 @@ This package enables import paths such as:
5
  - `server.app:app`
6
 
7
  The actual FastAPI app is provided in `server/app.py` (root-level wrapper)
8
- and delegates to `autonomous_returns_v0.server.app`.
9
  """
 
5
  - `server.app:app`
6
 
7
  The actual FastAPI app is provided in `server/app.py` (root-level wrapper)
8
+ and builds the runtime app directly from `autonomous_returns_v0`.
9
  """
server/app.py CHANGED
@@ -1,35 +1,75 @@
1
- """
2
- Root server app wrapper for AutonomousReturns-v0.
3
 
4
- This file exists to support OpenEnv layouts that expect `server.app:app`
5
- at repository root. It delegates to the package implementation at
6
- `autonomous_returns_v0.server.app`.
7
- """
8
 
9
  import os
10
- from importlib import import_module
 
11
 
 
12
 
13
- def _build_runtime_app():
14
- module = import_module("autonomous_returns_v0.server.app")
15
- build_app = getattr(module, "build_app")
16
- return build_app()
17
 
18
 
19
- app = _build_runtime_app()
 
 
 
 
 
20
 
21
 
22
- def main(host: str = "0.0.0.0", port: int = 8000) -> None:
 
 
 
 
 
 
 
 
 
23
  """
24
- Root entrypoint expected by OpenEnv local validators and local UI launch.
 
 
25
  """
26
- os.environ["ENABLE_WEB_INTERFACE"] = "true"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
27
  run_port = int(os.getenv("PORT", str(port)))
28
  uvicorn = import_module("uvicorn")
29
- uvicorn.run(_build_runtime_app(), host=host, port=run_port)
30
 
31
 
32
- __all__ = ["app", "main"]
33
 
34
  if __name__ == "__main__":
35
  main()
 
1
+ """FastAPI/OpenEnv application entrypoint for AutonomousReturns-v0."""
 
2
 
3
+ from __future__ import annotations
 
 
 
4
 
5
  import os
6
+ from importlib import import_module, util
7
+ from typing import Any
8
 
9
+ from autonomous_returns_v0 import AutonomousReturnsEnv, ReturnsAction, ReturnsObservation
10
 
11
+ _openenv_env_server = import_module("openenv.core.env_server")
12
+ create_app = getattr(_openenv_env_server, "create_app")
13
+ create_fastapi_app = getattr(_openenv_env_server, "create_fastapi_app")
 
14
 
15
 
16
+ def _web_requested() -> bool:
17
+ return os.getenv("ENABLE_WEB_INTERFACE", "false").strip().lower() in {
18
+ "1",
19
+ "true",
20
+ "yes",
21
+ }
22
 
23
 
24
+ def _gradio_available() -> bool:
25
+ return util.find_spec("gradio") is not None
26
+
27
+
28
+ def _env_factory() -> AutonomousReturnsEnv:
29
+ task = os.getenv("AUTONOMOUS_RETURNS_TASK", "easy").strip().lower() or "easy"
30
+ return AutonomousReturnsEnv(task=task)
31
+
32
+
33
+ def build_app() -> Any:
34
  """
35
+ Build the OpenEnv-compatible app.
36
+
37
+ If web UI is requested but Gradio is not installed, gracefully fall back to API-only.
38
  """
39
+ if _web_requested() and _gradio_available():
40
+ from server.gradio_builder import gradio_builder
41
+
42
+ return create_app(
43
+ env=_env_factory,
44
+ action_cls=ReturnsAction,
45
+ observation_cls=ReturnsObservation,
46
+ env_name="autonomous_returns_v0",
47
+ gradio_builder=gradio_builder,
48
+ )
49
+
50
+ if _web_requested() and not _gradio_available():
51
+ print(
52
+ "ENABLE_WEB_INTERFACE=true but gradio is not installed; serving API-only "
53
+ "(install optional dependency with `.[web]`)."
54
+ )
55
+
56
+ return create_fastapi_app(
57
+ env=_env_factory,
58
+ action_cls=ReturnsAction,
59
+ observation_cls=ReturnsObservation,
60
+ )
61
+
62
+
63
+ app = build_app()
64
+
65
+
66
+ def main(host: str = "0.0.0.0", port: int = 8000) -> None:
67
  run_port = int(os.getenv("PORT", str(port)))
68
  uvicorn = import_module("uvicorn")
69
+ uvicorn.run(build_app(), host=host, port=run_port)
70
 
71
 
72
+ __all__ = ["app", "build_app", "main"]
73
 
74
  if __name__ == "__main__":
75
  main()
server/gradio_builder.py CHANGED
@@ -20,9 +20,9 @@ from typing import Any
20
 
21
  import gradio as gr
22
 
23
- from ..baseline import BaselineAgent, BaselineRunConfig
24
- from ..models import DispositionAction, ReturnsObservation
25
- from ..policies import HeuristicPolicy
26
 
27
 
28
  def _build_header(task: str, ledger: float, pending_count: int) -> str:
@@ -478,15 +478,15 @@ def gradio_builder(
478
  return await _render_ui(web_manager, False)
479
 
480
  async def on_action_discount_15():
481
- await web_manager.step_environment({"action": "discount_15"})
482
  return await _render_ui(web_manager, False)
483
 
484
  async def on_action_discount_30():
485
- await web_manager.step_environment({"action": "discount_30"})
486
  return await _render_ui(web_manager, False)
487
 
488
  async def on_action_discount_50():
489
- await web_manager.step_environment({"action": "discount_50"})
490
  return await _render_ui(web_manager, False)
491
 
492
  async def on_action_refurbish():
@@ -535,14 +535,12 @@ def gradio_builder(
535
  provider = str(provider).strip().lower()
536
  if provider == "openai" and not os.getenv("OPENAI_API_KEY"):
537
  return await _render_ui(web_manager, reveal)
538
- if provider == "anthropic" and not os.getenv("ANTHROPIC_API_KEY"):
539
- return await _render_ui(web_manager, reveal)
540
  if provider == "google" and not os.getenv("VERTEX_PROJECT_ID"):
541
  return await _render_ui(web_manager, reveal)
542
 
543
  resolved_model = (model or "").strip() or (
544
  "gpt-4o-mini" if provider == "openai" else (
545
- "claude-3-5-haaku-20241022" if provider == "anthropic" else "gemini-2.0-flash"
546
  )
547
  )
548
  config = BaselineRunConfig(provider=provider, model=resolved_model, temperature=float(temp))
 
20
 
21
  import gradio as gr
22
 
23
+ from autonomous_returns_v0.baseline import BaselineAgent, BaselineRunConfig
24
+ from autonomous_returns_v0.models import DispositionAction, ReturnsObservation
25
+ from autonomous_returns_v0.policies import HeuristicPolicy
26
 
27
 
28
  def _build_header(task: str, ledger: float, pending_count: int) -> str:
 
478
  return await _render_ui(web_manager, False)
479
 
480
  async def on_action_discount_15():
481
+ await web_manager.step_environment({"action": "resell_discount_15"})
482
  return await _render_ui(web_manager, False)
483
 
484
  async def on_action_discount_30():
485
+ await web_manager.step_environment({"action": "resell_discount_30"})
486
  return await _render_ui(web_manager, False)
487
 
488
  async def on_action_discount_50():
489
+ await web_manager.step_environment({"action": "resell_discount_50"})
490
  return await _render_ui(web_manager, False)
491
 
492
  async def on_action_refurbish():
 
535
  provider = str(provider).strip().lower()
536
  if provider == "openai" and not os.getenv("OPENAI_API_KEY"):
537
  return await _render_ui(web_manager, reveal)
 
 
538
  if provider == "google" and not os.getenv("VERTEX_PROJECT_ID"):
539
  return await _render_ui(web_manager, reveal)
540
 
541
  resolved_model = (model or "").strip() or (
542
  "gpt-4o-mini" if provider == "openai" else (
543
+ "gemini-2.0-flash"
544
  )
545
  )
546
  config = BaselineRunConfig(provider=provider, model=resolved_model, temperature=float(temp))
spaces/Dockerfile CHANGED
@@ -25,4 +25,4 @@ EXPOSE 7860
25
  HEALTHCHECK --interval=30s --timeout=3s --start-period=15s --retries=3 \
26
  CMD curl -fsS http://localhost:7860/health || exit 1
27
 
28
- CMD ["uvicorn", "autonomous_returns_v0.server.app:app", "--host", "0.0.0.0", "--port", "7860"]
 
25
  HEALTHCHECK --interval=30s --timeout=3s --start-period=15s --retries=3 \
26
  CMD curl -fsS http://localhost:7860/health || exit 1
27
 
28
+ CMD ["uvicorn", "server.app:app", "--host", "0.0.0.0", "--port", "7860"]
spaces/README.md CHANGED
@@ -46,7 +46,7 @@ The project includes an OpenEnv manifest at `openenv.yaml` with:
46
 
47
  - `spec_version: 1`
48
  - runtime: `fastapi`
49
- - app: `autonomous_returns_v0.server.app:app`
50
  - typed action/observation/state model references
51
  - task definitions and scoring metadata
52
 
@@ -58,7 +58,7 @@ From the project root (`returnenv/autonomous_returns_v0`):
58
  - `pip install .`
59
 
60
  2. Run the server:
61
- - `uvicorn autonomous_returns_v0.server.app:app --host 0.0.0.0 --port 8000`
62
 
63
  3. Health check:
64
  - `GET /health`
@@ -67,7 +67,7 @@ From the project root (`returnenv/autonomous_returns_v0`):
67
 
68
  This Space uses the Dockerfile in `spaces/Dockerfile` and launches:
69
 
70
- - `uvicorn autonomous_returns_v0.server.app:app --host 0.0.0.0 --port 7860`
71
 
72
  The container exposes and serves on port `7860`, matching Hugging Face Space configuration.
73
  `ENABLE_WEB_INTERFACE=true` is enabled for this Space runtime, so both:
@@ -100,7 +100,7 @@ It supports async usage and sync wrapper usage through the OpenEnv client base.
100
  Two baselines are included:
101
 
102
  1. `BaselineAgent` (LLM-based)
103
- - Supports OpenAI (`OPENAI_API_KEY`) and Anthropic (`ANTHROPIC_API_KEY`).
104
 
105
  2. `LocalBaselineRunner` (deterministic, no external API)
106
  - Reproducible task scoring over fixed seeds.
@@ -111,12 +111,12 @@ The custom Gradio tab (`Custom`) includes:
111
 
112
  - Manual action controls for all 8 disposition actions
113
  - Heuristic autoplay (deterministic, no API key required)
114
- - LLM autoplay (`openai` or `anthropic`) using runtime env vars only
115
 
116
  LLM autoplay requires:
117
 
118
  - `OPENAI_API_KEY` for OpenAI mode
119
- - `ANTHROPIC_API_KEY` for Anthropic mode
120
 
121
  No API keys are entered in the UI.
122
 
 
46
 
47
  - `spec_version: 1`
48
  - runtime: `fastapi`
49
+ - app: `server.app:app`
50
  - typed action/observation/state model references
51
  - task definitions and scoring metadata
52
 
 
58
  - `pip install .`
59
 
60
  2. Run the server:
61
+ - `uvicorn server.app:app --host 0.0.0.0 --port 8000`
62
 
63
  3. Health check:
64
  - `GET /health`
 
67
 
68
  This Space uses the Dockerfile in `spaces/Dockerfile` and launches:
69
 
70
+ - `uvicorn server.app:app --host 0.0.0.0 --port 7860`
71
 
72
  The container exposes and serves on port `7860`, matching Hugging Face Space configuration.
73
  `ENABLE_WEB_INTERFACE=true` is enabled for this Space runtime, so both:
 
100
  Two baselines are included:
101
 
102
  1. `BaselineAgent` (LLM-based)
103
+ - Supports OpenAI (`OPENAI_API_KEY`) and Google Vertex (`VERTEX_PROJECT_ID`).
104
 
105
  2. `LocalBaselineRunner` (deterministic, no external API)
106
  - Reproducible task scoring over fixed seeds.
 
111
 
112
  - Manual action controls for all 8 disposition actions
113
  - Heuristic autoplay (deterministic, no API key required)
114
+ - LLM autoplay (`openai` or `google`) using runtime env vars only
115
 
116
  LLM autoplay requires:
117
 
118
  - `OPENAI_API_KEY` for OpenAI mode
119
+ - `VERTEX_PROJECT_ID` for Google mode
120
 
121
  No API keys are entered in the UI.
122
 
tests/test_gradio_builder.py CHANGED
@@ -1,9 +1,11 @@
1
  from __future__ import annotations
2
 
3
- import gradio as gr
4
  from openenv.core.env_server.types import EnvironmentMetadata
5
 
6
- from autonomous_returns_v0.server.gradio_builder import gradio_builder
 
 
7
 
8
 
9
  class _DummyEpisodeState:
 
1
  from __future__ import annotations
2
 
3
+ import pytest
4
  from openenv.core.env_server.types import EnvironmentMetadata
5
 
6
+ gr = pytest.importorskip("gradio")
7
+
8
+ from server.gradio_builder import gradio_builder
9
 
10
 
11
  class _DummyEpisodeState:
tests/test_server_web_launch.py CHANGED
@@ -3,7 +3,7 @@ from __future__ import annotations
3
  from fastapi import FastAPI
4
  from fastapi.testclient import TestClient
5
 
6
- from autonomous_returns_v0.server.app import build_app
7
 
8
 
9
  def test_build_app_returns_fastapi_instance() -> None:
 
3
  from fastapi import FastAPI
4
  from fastapi.testclient import TestClient
5
 
6
+ from server.app import build_app
7
 
8
 
9
  def test_build_app_returns_fastapi_instance() -> None:
uv.lock ADDED
The diff for this file is too large to render. See raw diff