build: pin one Python version in .python-version (3.14)

.github/workflows/ci.yml

@@ -20,11 +20,6 @@ jobs:
     steps:
     - uses: actions/checkout@v7
 
-    - name: Set up Python
-      uses: actions/setup-python@v7
-      with:
-        python-version: '3.11'
-
     - name: Install uv
       uses: astral-sh/setup-uv@v10.2.0
       with:
@@ -34,6 +29,16 @@ jobs:
     - name: Install dependencies
       run: uv sync --all-extras
 
+    # `.python-version` sets the version everywhere; a Dockerfile `FROM` line
+    # cannot read it, so check that the two agree.
+    - name: Check the Dockerfile uses the pinned Python
+      run: |
+        want="FROM python:$(cat .python-version)-slim"
+        if ! grep -qxF "$want" Dockerfile; then
+          echo "::error file=Dockerfile::Expected '$want' to match .python-version"
+          exit 1
+        fi
+
     - name: Check code formatting with Ruff
       run: uv run ruff format --check .
 
@@ -52,11 +57,6 @@ jobs:
     steps:
     - uses: actions/checkout@v7
 
-    - name: Set up Python
-      uses: actions/setup-python@v7
-      with:
-        python-version: '3.11'
-
     - name: Install uv
       uses: astral-sh/setup-uv@v10.2.0
       with:
@@ -93,11 +93,6 @@ jobs:
     steps:
     - uses: actions/checkout@v7
 
-    - name: Set up Python
-      uses: actions/setup-python@v7
-      with:
-        python-version: '3.11'
-
     - name: Install uv
       uses: astral-sh/setup-uv@v10.2.0
       with:

.github/workflows/live.yml

@@ -48,11 +48,10 @@ jobs:
     - uses: actions/checkout@v7
 
     # The runner is an AL2023 host, which setup-python has no builds for, so
-    # uv installs Python instead.
+    # uv installs the Python pinned in .python-version instead.
     - name: Install uv
       uses: astral-sh/setup-uv@v10.2.0
       with:
-        python-version: '3.11'
         enable-cache: true
         cache-dependency-glob: uv.lock
 

.python-version

@@ -0,0 +1 @@
+3.14

AGENTS.md

@@ -11,7 +11,7 @@ source .venv/bin/activate  # Windows: .venv\Scripts\activate
 cp .env.example .env       # Add API keys (ANTHROPIC_API_KEY, OPENAI_API_KEY, etc.)
 ```
 
-**Python >= 3.11 required**
+**Python 3.14 required.** `.python-version` is the one place the version is set: `uv`, CI, and the Dockerfile's `FROM` line all follow it.
 
 ## Code Style
 

docs/ARCHITECTURE-SPINE.md

@@ -265,7 +265,7 @@ No edge runs `generate/` ↔ `judge/` ↔ `score/`, and none runs from `workers/
 
 | Name | Version |
 | --- | --- |
-| Python | `>=3.11` (per `pyproject.toml` `requires-python`; ruff `target-version = "py311"`) |
+| Python | `3.14`, set in `.python-version`; `requires-python` and ruff `target-version` match it |
 | Package/dependency manager | uv (`uv.lock` present; CLI usage documented as `uv run python vera.py ...`) |
 | langchain / langchain-anthropic / langchain-openai / langchain-ollama | `>=0.1.0` direct floor per `pyproject.toml`, but effectively constrained to the post-1.0 langchain rewrite by `[tool.uv] constraint-dependencies`' `langchain-core>=1.2.5` pin — pre-1.0 langchain is not actually installable under this lockfile despite the loose direct floor |
 | langchain-google-genai / langchain-azure-ai | `>=1.0.0` |

judge/question_navigator.py

@@ -111,7 +111,7 @@ def _get_next_row_question(self, current_question_id: str) -> Optional[str]:
             current_index = self.question_order.index(current_question_id)
             if current_index + 1 < len(self.question_order):
                 return self.question_order[current_index + 1]
-        except (ValueError, AttributeError):
+        except ValueError, AttributeError:
             pass
         return None
 

judge/score_comparison_from_export_summary.py

@@ -192,14 +192,14 @@ def load_model_data_from_export_summary_csv(
 
         try:
             vera_score = float(row["Total score"])
-        except (TypeError, ValueError, KeyError):
+        except TypeError, ValueError, KeyError:
             print(f"⚠️  Warning: Bad Total score for row Provider={model_name!r}, skip")
             continue
 
         try:
             overall_bp_pct = float(row["Total % BP (scored ratings)"])
             overall_hph_pct = float(row["Total % HPH (scored ratings)"])
-        except (TypeError, ValueError, KeyError):
+        except TypeError, ValueError, KeyError:
             overall_bp_pct = 0.0
             overall_hph_pct = 0.0
 
@@ -217,20 +217,20 @@ def load_model_data_from_export_summary_csv(
             else:
                 try:
                     vera_dim = float(raw_v)
-                except (TypeError, ValueError):
+                except TypeError, ValueError:
                     vera_dim = 50.0
             col_hph = f"{d} — % HPH (scored)"
             col_bp = f"{d} — % BP (scored)"
             hph_d = bp_d = 0.0
             if col_hph in row.index and pd.notna(row[col_hph]):
                 try:
                     hph_d = float(row[col_hph])
-                except (TypeError, ValueError):
+                except TypeError, ValueError:
                     pass
             if col_bp in row.index and pd.notna(row[col_bp]):
                 try:
                     bp_d = float(row[col_bp])
-                except (TypeError, ValueError):
+                except TypeError, ValueError:
                     pass
             dimensions[d] = {
                 "vera_score": vera_dim,

judge/score_viz.py

@@ -43,7 +43,7 @@ def _safe_pie_value(value: Any) -> float:
     """Coerce aggregate percentage to a finite non-negative float for pie charts."""
     try:
         v = float(value)
-    except (TypeError, ValueError):
+    except TypeError, ValueError:
         return 0.0
     if v != v or v < 0:  # NaN or negative
         return 0.0
@@ -66,7 +66,7 @@ def _create_pie_chart(ax, results: Dict[str, Any]):
         score_display = float(overall_vera_score)
         if score_display != score_display:
             score_display = 0.0
-    except (TypeError, ValueError):
+    except TypeError, ValueError:
         score_display = 0.0
     pie_title = (
         f"Overall VERA-MH v1.2 Score: {score_display:.1f}\n\nRating Distribution"

pyproject.toml

@@ -2,7 +2,7 @@
 name = "vera-mh"
 version = "0.1.0"
 description = "Validation of Ethical and Responsible AI in Mental Health"
-requires-python = ">=3.11"
+requires-python = ">=3.14"
 dependencies = [
     "langchain>=0.1.0",
     "langchain-anthropic>=1.5.6",
@@ -63,7 +63,8 @@ typeCheckingMode = "basic"
 
 [tool.ruff]
 line-length = 88
-target-version = "py311"
+# No `target-version`: ruff reads it from `requires-python`, which follows
+# `.python-version`.
 # Ruff 0.16's default `include` carries `*.md`, so it reformats Python code
 # blocks inside Markdown. Our docs use snippets as illustrations -- pseudo-code,
 # partial excerpts, and deliberately aligned trailing comments -- which the

run_pipeline.py

@@ -43,7 +43,7 @@ def _display_path(path: str | os.PathLike[str]) -> str:
         resolved = Path(path).resolve()
         cwd = Path.cwd().resolve()
         return str(resolved.relative_to(cwd))
-    except (ValueError, OSError):
+    except ValueError, OSError:
         return str(Path(path).resolve())
 
 

tests/mocks/mock_llm.py

@@ -128,7 +128,7 @@ async def generate_structured_response(
 
             response_data = json.loads(response_text)
             return response_model(**response_data)
-        except (json.JSONDecodeError, ValueError, TypeError):
+        except json.JSONDecodeError, ValueError, TypeError:
             # Create instance with mock default values
             # Check if the model has example data in json_schema_extra
             if hasattr(response_model, "model_config") and isinstance(

utils/conversation_utils.py

@@ -117,7 +117,7 @@ def format_conversation_summary(
         try:
             turn_role = Role(speaker_value)
             prefix = "user:" if turn_role == Role.PERSONA else "chatbot:"
-        except (ValueError, TypeError):
+        except ValueError, TypeError:
             raise ValueError(f"Invalid role value '{speaker_value}' for turn {i}")
 
         # Add the response with proper formatting
@@ -222,7 +222,7 @@ def build_langchain_messages(
                         message = AIMessage(content=text)
                     else:
                         message = HumanMessage(content=text)
-                except (ValueError, TypeError):
+                except ValueError, TypeError:
                     raise ValueError(
                         f"Invalid role value '{turn_speaker}' for turn {turn_number}"
                     )

utils/utils.py

@@ -11,7 +11,7 @@ def parse_key_value_list(arg):
         # https://docs.python.org/3/library/ast.html#ast.literal_eval
         try:
             value = ast.literal_eval(value)
-        except (ValueError, SyntaxError):
+        except ValueError, SyntaxError:
             # Note: not logging the error here as we are leaving the value as a string
             pass
         d[key] = value