Merge branch 'trace-recorder' into hints_retrieve

Author: ollmer

.github/workflows/code_format.yml

@@ -18,17 +18,19 @@ jobs:
       - name: Checkout Repository
         uses: actions/checkout@v4
 
-      - name: Set up Python
-        uses: actions/setup-python@v5
+      - name: Install uv
+        uses: astral-sh/setup-uv@v4
         with:
-          python-version: '3.11'
-          cache: 'pip' # caching pip dependencies
+          enable-cache: true
+
+      - name: Set up Python
+        run: uv python install 3.11
 
-      - name: Pip install
-        run: pip install -r requirements.txt
+      - name: Install dependencies
+        run: uv sync --frozen --extra dev
 
-      - name: Pip list
-        run: pip list
+      - name: List packages
+        run: uv pip list
 
       - name: Code Formatting
-        run: black . --check --diff
+        run: uv run black src/ --check --diff

.github/workflows/darglint.yml

@@ -18,17 +18,19 @@ jobs:
       - name: Checkout Repository
         uses: actions/checkout@v4
 
-      - name: Set up Python
-        uses: actions/setup-python@v5
+      - name: Install uv
+        uses: astral-sh/setup-uv@v4
         with:
-          python-version: '3.12'
-          cache: 'pip' # caching pip dependencies
+          enable-cache: true
+
+      - name: Set up Python
+        run: uv python install 3.12 # this fails in 3.11
 
-      - name: Pip install
-        run: pip install darglint
+      - name: Install dependencies
+        run: uv sync --frozen --extra dev
 
-      - name: Pip list
-        run: pip list
+      - name: List packages
+        run: uv pip list
 
       - name: Darglint checks
-        run: darglint -v 2 -z short src/
+        run: uv run darglint -v 2 -z short src/

.github/workflows/python_version_compatibility.yml

@@ -32,9 +32,12 @@ jobs:
     - name: Check Python ${{ matrix.python-version }}
       continue-on-error: true
       run: |
-        export PATH="$HOME/.cargo/bin:$PATH"
-        if uvx --python ${{ matrix.python-version }} --from python --with-requirements requirements.txt python -c "print('✅ Compatible')"; then
-          echo "✅ Python ${{ matrix.python-version }} works"
-        else
-          echo "❌ Python ${{ matrix.python-version }} incompatible"
-        fi
+          export PATH="$HOME/.cargo/bin:$PATH"
+          uv python install ${{ matrix.python-version }}
+          if uv sync --frozen --python ${{ matrix.python-version }}; then
+            uv run -p ${{ matrix.python-version }} python -c "import sys; print('✅ Compatible:', sys.version)"
+            echo "✅ Python ${{ matrix.python-version }} works"
+          else
+            echo "❌ Python ${{ matrix.python-version }} incompatible"
+            exit 1
+          fi

.github/workflows/unit_tests.yml

@@ -23,23 +23,25 @@ jobs:
       - name: Set up Git user
         run: git config --global user.email "[email protected]" && git config --global user.name "GitHub Actions"
 
+      - name: Install uv
+        uses: astral-sh/setup-uv@v4
+        with:
+          enable-cache: true
+
       - name: Set up Python
-        uses: actions/setup-python@v5
-        with: # python at least 3.11
-          python-version: '3.11'
-          cache: 'pip' # caching pip dependencies
+        run: uv python install 3.11
 
       - name: Install AgentLab
-        run: pip install -e .
+        run: uv sync --frozen --extra dev
 
-      - name: Pip list
-        run: pip list
+      - name: List packages
+        run: uv pip list
 
       - name: Install Playwright
-        run: playwright install chromium --with-deps
+        run: uv run playwright install chromium --with-deps
 
       - name: Download WebArena / VisualWebArena ressource files
-        run: python -c 'import nltk; nltk.download("punkt_tab")'
+        run: uv run python -c 'import nltk; nltk.download("punkt_tab")'
 
       - name: Fetch MiniWob
         uses: actions/checkout@v4
@@ -59,9 +61,9 @@ jobs:
         run: curl -I "http://localhost:8080/miniwob/" || echo "MiniWob not reachable"
 
       - name: Pre-download nltk ressources
-        run: python -c "import nltk; nltk.download('punkt_tab')"
+        run: uv run python -c "import nltk; nltk.download('punkt_tab')"
 
       - name: Run AgentLab Unit Tests
         env:
           MINIWOB_URL: "http://localhost:8080/miniwob/"
-        run: pytest -n 5 --durations=10 -m 'not pricy' -v tests/
+        run: uv run pytest -n 5 --durations=10 -m 'not pricy' -v tests/

.readthedocs.yaml

@@ -32,4 +32,8 @@ sphinx:
 # See https://docs.readthedocs.io/en/stable/guides/reproducible-builds.html
 python:
   install:
+    - method: pip
+      path: .
+      extra_requirements:
+        - dev
     - requirements: docs/source/requirements.txt

docs/source/requirements.txt

@@ -1,2 +1,2 @@
-agentlab
-sphinx-rtd-theme
+sphinx-rtd-theme
+sphinx

main.py

@@ -15,9 +15,9 @@
     AGENT_4o,
     AGENT_4o_MINI,
     AGENT_o3_MINI,
-    AGENT_o1_MINI,
     AGENT_37_SONNET,
     AGENT_CLAUDE_SONNET_35,
+    AGENT_GPT5_MINI,
 )
 from agentlab.experiments.study import Study
 

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "agentlab"
-dynamic = ["version", "dependencies"]
+dynamic = ["version"]
 description = "Main package for developing agents and experiments"
 authors = [
     {name = "Rim Assouel", email = "[email protected]"},
@@ -27,13 +27,55 @@ classifiers = [
     "Topic :: Scientific/Engineering :: Artificial Intelligence",
     "License :: OSI Approved :: Apache Software License",
 ]
+dependencies = [
+    "pydantic~=2.9",
+    "dask",
+    "distributed",
+    "browsergym>=0.7.1",
+    "joblib>=1.2.0",
+    "openai>=1.7,<2",
+    "langchain_community",
+    "tiktoken",
+    "huggingface_hub",
+    "contexttimer",
+    "ipython",
+    "pyyaml>=6",
+    "pandas",
+    "gradio>=5.5",
+    "gitpython",
+    "requests",
+    "matplotlib",
+    "ray[default]",
+    "python-slugify",
+    "pillow",
+    "gymnasium>=0.27",
+    "torch>=2.2.2",
+    "safetensors>=0.4.0",
+    "transformers>=4.38.2",
+    "anthropic>=0.62.0",
+    "litellm>=1.75.3",
+    "python-dotenv>=1.1.1",
+]
+
+[project.optional-dependencies]
+dev = [
+    "black[jupyter]>=24.2.0",
+    "blacken-docs",
+    "pre-commit",
+    "pytest==7.3.2",
+    "flaky",
+    "pytest-xdist",
+    "pytest-playwright",
+]
+# tapeagents = [
+#     "tapeagents[converters]",
+# ]
 
 [project.urls]
 "Homepage" = "https://github.com/ServiceNow/AgentLab"
 
 [tool.setuptools.dynamic]
 version = {attr = "agentlab.__version__"}
-dependencies = {file = ["requirements.txt"]}
 
 [tool.black]
 line-length = 100
@@ -54,6 +96,13 @@ exclude = '''
 )/
 '''
 
+[dependency-groups]
+dev = [
+    "darglint>=1.8.1",
+    "ipykernel>=6.30.1",
+    "pip>=25.2",
+]
+
 
 [project.scripts]
 agentlab-assistant = "agentlab.ui_assistant:main"

src/agentlab/agents/generic_agent/__init__.py

@@ -26,6 +26,7 @@
     AGENT_o3_MINI,
     FLAGS_GPT_4o,
     GenericAgentArgs,
+    AGENT_GPT5_MINI,
 )
 
 __all__ = [
@@ -46,4 +47,5 @@
     "AGENT_4o_VISION",
     "AGENT_4o_MINI_VISION",
     "AGENT_CLAUDE_SONNET_35_VISION",
+    "AGENT_GPT5_MINI",
 ]

src/agentlab/agents/generic_agent/agent_configs.py

@@ -270,8 +270,12 @@
     chat_model_args=CHAT_MODEL_ARGS_DICT["openrouter/anthropic/claude-3.7-sonnet"],
     flags=FLAGS_GPT_4o,
 )
+# AGENT_o3_MINI = GenericAgentArgs(
+#     chat_model_args=CHAT_MODEL_ARGS_DICT["openai/o3-mini-2025-01-31"],
+#     flags=FLAGS_GPT_4o,
+# )
 AGENT_o3_MINI = GenericAgentArgs(
-    chat_model_args=CHAT_MODEL_ARGS_DICT["openai/o3-mini-2025-01-31"],
+    chat_model_args=CHAT_MODEL_ARGS_DICT["openrouter/openai/o3-mini"],
     flags=FLAGS_GPT_4o,
 )
 
@@ -302,6 +306,18 @@
     chat_model_args=CHAT_MODEL_ARGS_DICT["openrouter/meta-llama/llama-4-maverick"],
     flags=BASE_FLAGS,
 )
+GPT5_MINI_FLAGS = BASE_FLAGS.copy()
+GPT5_MINI_FLAGS.action = dp.ActionFlags(  # action should not be str to work with agentlab-assistant
+    action_set=HighLevelActionSetArgs(
+        subsets=["bid"],
+        multiaction=False,
+    )
+)
+
+AGENT_GPT5_MINI = GenericAgentArgs(
+    chat_model_args=CHAT_MODEL_ARGS_DICT["openai/gpt-5-mini-2025-08-07"],
+    flags=GPT5_MINI_FLAGS,
+)
 
 DEFAULT_RS_FLAGS = GenericPromptFlags(
     flag_group="default_rs",