Expand test suite

.github/workflows/checks.yml

@@ -26,7 +26,7 @@ jobs:
         with:
           enable-cache: true
 
-      - run: uv sync --all-extras --all-packages --group lint
+      - run: uv sync --all-packages --group lint
 
       - uses: pre-commit/action@v3.0.0
         with:
@@ -40,6 +40,11 @@ jobs:
   test:
     runs-on: ubuntu-latest
     timeout-minutes: 10
+    strategy:
+      fail-fast: false
+      matrix:
+        pydantic-ai-version: [min, latest]
+    name: test (pydantic-ai ${{ matrix.pydantic-ai-version }})
     steps:
       - uses: actions/checkout@v4
 
@@ -48,5 +53,10 @@ jobs:
           enable-cache: true
 
       - run: mkdir .coverage
-      - run: uv sync --all-extras --group dev
-      - run: uv run coverage run -m pytest --durations=100 -n auto --dist=loadgroup
+      - if: matrix.pydantic-ai-version == 'min'
+        run: uv sync --group dev --resolution lowest-direct
+      - if: matrix.pydantic-ai-version == 'latest'
+        run: uv sync --group dev --upgrade-package pydantic-ai
+        env:
+          UV_FROZEN: "0"
+      - run: make test

.pre-commit-config.yaml

@@ -26,13 +26,6 @@ repos:
 
   - repo: local
     hooks:
-      - id: format
-        name: Format
-        entry: make
-        args: [format]
-        language: system
-        types: [python]
-        pass_filenames: false
       - id: lint
         name: Lint
         entry: make

Makefile

@@ -1,3 +1,7 @@
+.PHONY: install
+install: ## Install git hooks for local development
+	uv run pre-commit install
+
 .PHONY: format
 format: ## Format the code
 	uv run ruff format

pyproject.toml

@@ -20,24 +20,15 @@ version = { attr = "cragents._version.__version__" }
 
 [dependency-groups]
 dev = [
-    "anyio>=4.11.0",
-    "coverage>=7.12.0",
+    "anyio>=4.12.1",
+    "coverage>=7.13.2",
     "inline-snapshot>=0.31.1",
-    "pyright>=1.1.406",
-    "pytest>=9.0.1",
+    "pre-commit>=4.5.1",
+    "pyright>=1.1.408",
+    "pytest>=9.0.2",
     "pytest-xdist>=3.8.0",
     "trio>=0.32.0",
 ]
-example = [
-    "chromadb>=1.3.7",
-    "datasets>=4.4.2",
-    "langchain-text-splitters>=1.1.0",
-    "markdownify>=1.2.2",
-    "playwright>=1.57.0",
-    "requests-cache>=1.2.1",
-    "rich>=14.2.0",
-    "transformers>=4.57.3",
-]
 lint = [
     "pyright>=1.1.406",
     "ruff>=0.14.1",

tests/test_.py → tests/test_agent.py

@@ -1,7 +1,8 @@
 import pytest
 from inline_snapshot import snapshot
 from pydantic_ai import ToolOutput
-from pydantic_ai.models.openai import OpenAIChatModel
+from pydantic_ai.models.openai import OpenAIChatModel, OpenAIChatModelSettings
+from pydantic_ai.models.test import TestModel
 from pydantic_ai.providers.openai import OpenAIProvider
 
 from cragents import Anchor, Constrain, CRAgent, Free, Think, UseTools, vllm_model_profile
@@ -24,6 +25,9 @@
 ]
 
 
+# ── end-to-end set_guide output type tests ────────────────────────────────────
+
+
 async def test_default_agent_output():
     agent = CRAgent(model)
     await agent.set_guide(generation_sequence)
@@ -134,3 +138,102 @@ async def test_mixed_output_type():
             }
         }
     )
+
+
+# ── set_guide error handling and model settings ───────────────────────────────
+
+
+async def test_set_guide_requires_openai_model():
+    agent = CRAgent(TestModel())
+    with pytest.raises(RuntimeError, match="OpenAIChatModel required"):
+        await agent.set_guide([Anchor("hi ")])
+
+
+async def test_set_guide_creates_model_settings_when_none():
+    agent = CRAgent(model)
+    assert agent.model_settings is None
+    await agent.set_guide([Anchor("hi ")])
+    assert agent.model_settings is not None
+    assert "extra_body" in agent.model_settings
+
+
+async def test_set_guide_preserves_existing_model_settings():
+    agent = CRAgent(model, model_settings=OpenAIChatModelSettings(temperature=0.5))
+    await agent.set_guide([Anchor("hi ")])
+    assert agent.model_settings["temperature"] == 0.5
+    assert "extra_body" in agent.model_settings
+
+
+async def test_set_guide_overwrites_on_second_call():
+    agent = CRAgent(model)
+    await agent.set_guide([Anchor("first ")])
+    first_grammar = agent.model_settings["extra_body"]["structured_outputs"]["grammar"]
+    await agent.set_guide([Anchor("second ")])
+    second_grammar = agent.model_settings["extra_body"]["structured_outputs"]["grammar"]
+    assert first_grammar != second_grammar
+    assert "second" in second_grammar
+
+
+# ── set_guide UseTools schema handling ────────────────────────────────────────
+
+
+async def test_set_guide_explicit_use_tools_schema_not_overwritten():
+    explicit_schema = {"type": "number"}
+    agent = CRAgent(model)
+    await agent.set_guide([UseTools(json_schema=explicit_schema)])
+    grammar = agent.model_settings["extra_body"]["structured_outputs"]["grammar"]
+    assert '"type": "number"' in grammar
+
+
+async def test_set_guide_use_tools_with_registered_tool():
+    agent = CRAgent(model)
+
+    @agent.tool_plain
+    def my_tool(x: int) -> str:
+        return str(x)
+
+    await agent.set_guide([UseTools()])
+    grammar = agent.model_settings["extra_body"]["structured_outputs"]["grammar"]
+    # The tool's parameter schema (containing "x") should appear in the grammar
+    assert '"x"' in grammar
+
+
+async def test_set_guide_use_tools_tool_names():
+    agent = CRAgent(model)
+    await agent.set_guide([UseTools(json_schema={"type": "string"}, tool_names=["alpha", "beta"])])
+    grammar = agent.model_settings["extra_body"]["structured_outputs"]["grammar"]
+    assert 'FUNCTION_NAME: ("alpha" | "beta")' in grammar
+
+
+async def test_set_guide_merges_toolset_with_anyof_output():
+    # When the output schema already has anyOf (multiple output types) and the agent
+    # also has registered tools, anyOf = toolset_schemas + return_schema["anyOf"]
+    agent = CRAgent(model, output_type=[ToolOutput(bool), ToolOutput(int)])
+
+    @agent.tool_plain
+    def helper(x: str) -> str:
+        return x
+
+    await agent.set_guide([UseTools()])
+    grammar = agent.model_settings["extra_body"]["structured_outputs"]["grammar"]
+    assert "tool_schema" in grammar
+    assert "anyOf" in grammar
+
+
+# ── vllm_model_profile ─────────────────────────────────────────────────────────
+
+
+def test_vllm_profile_strict_tool_definition():
+    assert vllm_model_profile.openai_supports_strict_tool_definition is False
+
+
+def test_vllm_profile_tool_choice_required():
+    assert vllm_model_profile.openai_supports_tool_choice_required is False
+
+
+def test_vllm_profile_json_object_output():
+    assert vllm_model_profile.supports_json_object_output is False
+
+
+def test_vllm_profile_json_schema_output():
+    assert vllm_model_profile.supports_json_schema_output is True