alignmenter 0.1.0__tar.gz → 0.1.2__tar.gz

{alignmenter-0.1.0/src/alignmenter.egg-info → alignmenter-0.1.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: alignmenter
-Version: 0.1.0
+Version: 0.1.2
 Summary: Persona-aligned evaluation toolkit for auditing conversational AI authenticity, safety, and stability.
 Author: Alignmenter
 License:                                  Apache License
@@ -236,6 +236,7 @@ Requires-Dist: tiktoken>=0.7.0
 Requires-Dist: requests>=2.32.0
 Provides-Extra: dev
 Requires-Dist: pytest; extra == "dev"
+Requires-Dist: pytest-cov; extra == "dev"
 Requires-Dist: ruff; extra == "dev"
 Requires-Dist: build; extra == "dev"
 Requires-Dist: twine; extra == "dev"
@@ -277,26 +278,23 @@ Unlike generic LLM evaluation frameworks, Alignmenter is purpose-built for **per
 
 ```bash
 # Clone the repository
-git clone https://github.com/yourusername/alignmenter.git
+git clone https://github.com/justinGrosvenor/alignmenter.git
 cd alignmenter
 
 # Create virtual environment
 python -m venv env
 source env/bin/activate  # On Windows: env\Scripts\activate
 
-# Install with dev dependencies
-pip install -e .[dev]
-
-# Optional: Install with offline safety classifier
-pip install -e .[dev,safety]
+# Install with dev + safety extras
+pip install -e ./alignmenter[dev,safety]
 ```
 
 ### Install from PyPI
 
 ```bash
-pip install alignmenter
+pip install "alignmenter[safety]"
 alignmenter init
-alignmenter run --config configs/run.yaml --no-generate
+alignmenter run --config configs/run.yaml --embedding sentence-transformer:all-MiniLM-L6-v2
 ```
 
 > **Note**: The `safety` extra includes `transformers` for the offline safety classifier (ProtectAI/distilled-safety-roberta). Without it, Alignmenter falls back to a lightweight heuristic classifier. See [docs/offline_safety.md](https://github.com/justinGrosvenor/alignmenter/blob/main/docs/offline_safety.md) for details.
@@ -314,13 +312,16 @@ alignmenter run \
   --persona configs/persona/default.yaml
 
 # Reuse existing transcripts without hitting the provider
-alignmenter run --config configs/run.yaml --no-generate
+alignmenter run --config configs/run.yaml
 
 # View interactive HTML report
 alignmenter report --last
 
 # Sanitize a dataset in-place or to a new file
 alignmenter dataset sanitize datasets/demo_conversations.jsonl --out datasets/demo_sanitized.jsonl
+
+# Generate fresh transcripts (requires provider access)
+alignmenter run --config configs/run.yaml --generate-transcripts
 ```
 
 **Output:**
@@ -438,6 +439,14 @@ Quick links:
 - **[LLM Judges](https://docs.alignmenter.com/guides/llm-judges/)** - Qualitative analysis
 - **[Contributing](https://docs.alignmenter.com/contributing/)** - How to contribute
 
+---
+
+## Case Studies
+
+- **[Wendy's Twitter Voice](../docs/case-studies/wendys-twitter.md)** - End-to-end calibration example using the included case-study assets. *(Available when running from the source repo; not included in the PyPI wheel.)*
+
+---
+
 ## Usage Examples
 
 ### Evaluate Multiple Models
@@ -490,11 +499,9 @@ alignmenter calibrate-persona \
 
 ```bash
 # Remove PII before evaluation
-alignmenter sanitize-dataset \
-  --input prod_logs.jsonl \
+alignmenter dataset sanitize prod_logs.jsonl \
   --out datasets/sanitized.jsonl \
-  --hash-session-ids \
-  --scrub-patterns "email|phone|ssn"
+  --no-use-hashing
 ```
 
 ## Persona Configuration
@@ -646,13 +653,25 @@ make report-last
 
 ## Roadmap
 
-- [ ] Web dashboard for team collaboration
-- [ ] Hosted evaluation pipelines (CI/CD SaaS)
-- [ ] Multi-language support (non-English personas)
-- [ ] Offline safety model (no API required)
-- [ ] Cost/latency tracking in scorers
-- [ ] Competitive landscape analysis
-- [ ] Metric cards and ablation studies
+### Completed ✅
+- Three-dimensional scoring (authenticity, safety, stability)
+- Multi-provider support (OpenAI, Anthropic, local models)
+- HTML report cards with interactive charts
+- Offline safety classifier (distilled-safety-roberta)
+- LLM judges for qualitative analysis
+- Budget guardrails and cost tracking
+- PII sanitization tools
+- Calibration workflow and diagnostics
+
+### In Progress 🚧
+- Multi-language support (non-English personas)
+- Batch processing optimizations
+- Additional embedding providers
+
+### Future Considerations 💭
+- Synthetic test case generation
+- Custom metric plugins
+- Advanced trait models (neural networks)
 
 ## Contributing
 
@@ -676,14 +695,15 @@ If you use Alignmenter in research, please cite:
 @software{alignmenter2024,
   title={Alignmenter: A Framework for Persona-Aligned Conversational AI Evaluation},
   author={Alignmenter Contributors},
-  year={2024},
-  url={https://github.com/yourusername/alignmenter}
+  year={2025},
+  url={https://github.com/justinGrosvenor/alignmenter},
+  license={Apache-2.0}
 }
 ```
 
 ## Support
 
-- **Documentation**: [docs/](https://github.com/justinGrosvenor/alignmenter/tree/main/docs)
+- **Documentation**: [docs.alignmenter.com](https://docs.alignmenter.com)
 - **Issues**: [GitHub Issues](https://github.com/justinGrosvenor/alignmenter/issues)
 - **Discussions**: [GitHub Discussions](https://github.com/justinGrosvenor/alignmenter/discussions)
 

{alignmenter-0.1.0 → alignmenter-0.1.2}/README.md

@@ -32,26 +32,23 @@ Unlike generic LLM evaluation frameworks, Alignmenter is purpose-built for **per
 
 ```bash
 # Clone the repository
-git clone https://github.com/yourusername/alignmenter.git
+git clone https://github.com/justinGrosvenor/alignmenter.git
 cd alignmenter
 
 # Create virtual environment
 python -m venv env
 source env/bin/activate  # On Windows: env\Scripts\activate
 
-# Install with dev dependencies
-pip install -e .[dev]
-
-# Optional: Install with offline safety classifier
-pip install -e .[dev,safety]
+# Install with dev + safety extras
+pip install -e ./alignmenter[dev,safety]
 ```
 
 ### Install from PyPI
 
 ```bash
-pip install alignmenter
+pip install "alignmenter[safety]"
 alignmenter init
-alignmenter run --config configs/run.yaml --no-generate
+alignmenter run --config configs/run.yaml --embedding sentence-transformer:all-MiniLM-L6-v2
 ```
 
 > **Note**: The `safety` extra includes `transformers` for the offline safety classifier (ProtectAI/distilled-safety-roberta). Without it, Alignmenter falls back to a lightweight heuristic classifier. See [docs/offline_safety.md](https://github.com/justinGrosvenor/alignmenter/blob/main/docs/offline_safety.md) for details.
@@ -69,13 +66,16 @@ alignmenter run \
   --persona configs/persona/default.yaml
 
 # Reuse existing transcripts without hitting the provider
-alignmenter run --config configs/run.yaml --no-generate
+alignmenter run --config configs/run.yaml
 
 # View interactive HTML report
 alignmenter report --last
 
 # Sanitize a dataset in-place or to a new file
 alignmenter dataset sanitize datasets/demo_conversations.jsonl --out datasets/demo_sanitized.jsonl
+
+# Generate fresh transcripts (requires provider access)
+alignmenter run --config configs/run.yaml --generate-transcripts
 ```
 
 **Output:**
@@ -193,6 +193,14 @@ Quick links:
 - **[LLM Judges](https://docs.alignmenter.com/guides/llm-judges/)** - Qualitative analysis
 - **[Contributing](https://docs.alignmenter.com/contributing/)** - How to contribute
 
+---
+
+## Case Studies
+
+- **[Wendy's Twitter Voice](../docs/case-studies/wendys-twitter.md)** - End-to-end calibration example using the included case-study assets. *(Available when running from the source repo; not included in the PyPI wheel.)*
+
+---
+
 ## Usage Examples
 
 ### Evaluate Multiple Models
@@ -245,11 +253,9 @@ alignmenter calibrate-persona \
 
 ```bash
 # Remove PII before evaluation
-alignmenter sanitize-dataset \
-  --input prod_logs.jsonl \
+alignmenter dataset sanitize prod_logs.jsonl \
   --out datasets/sanitized.jsonl \
-  --hash-session-ids \
-  --scrub-patterns "email|phone|ssn"
+  --no-use-hashing
 ```
 
 ## Persona Configuration
@@ -401,13 +407,25 @@ make report-last
 
 ## Roadmap
 
-- [ ] Web dashboard for team collaboration
-- [ ] Hosted evaluation pipelines (CI/CD SaaS)
-- [ ] Multi-language support (non-English personas)
-- [ ] Offline safety model (no API required)
-- [ ] Cost/latency tracking in scorers
-- [ ] Competitive landscape analysis
-- [ ] Metric cards and ablation studies
+### Completed ✅
+- Three-dimensional scoring (authenticity, safety, stability)
+- Multi-provider support (OpenAI, Anthropic, local models)
+- HTML report cards with interactive charts
+- Offline safety classifier (distilled-safety-roberta)
+- LLM judges for qualitative analysis
+- Budget guardrails and cost tracking
+- PII sanitization tools
+- Calibration workflow and diagnostics
+
+### In Progress 🚧
+- Multi-language support (non-English personas)
+- Batch processing optimizations
+- Additional embedding providers
+
+### Future Considerations 💭
+- Synthetic test case generation
+- Custom metric plugins
+- Advanced trait models (neural networks)
 
 ## Contributing
 
@@ -431,14 +449,15 @@ If you use Alignmenter in research, please cite:
 @software{alignmenter2024,
   title={Alignmenter: A Framework for Persona-Aligned Conversational AI Evaluation},
   author={Alignmenter Contributors},
-  year={2024},
-  url={https://github.com/yourusername/alignmenter}
+  year={2025},
+  url={https://github.com/justinGrosvenor/alignmenter},
+  license={Apache-2.0}
 }
 ```
 
 ## Support
 
-- **Documentation**: [docs/](https://github.com/justinGrosvenor/alignmenter/tree/main/docs)
+- **Documentation**: [docs.alignmenter.com](https://docs.alignmenter.com)
 - **Issues**: [GitHub Issues](https://github.com/justinGrosvenor/alignmenter/issues)
 - **Discussions**: [GitHub Discussions](https://github.com/justinGrosvenor/alignmenter/discussions)
 

{alignmenter-0.1.0 → alignmenter-0.1.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "alignmenter"
-version = "0.1.0"
+version = "0.1.2"
 description = "Persona-aligned evaluation toolkit for auditing conversational AI authenticity, safety, and stability."
 authors = [{name = "Alignmenter"}]
 readme = {file = "README.md", content-type = "text/markdown"}
@@ -38,7 +38,7 @@ dependencies = [
 ]
 
 [project.optional-dependencies]
-dev = ["pytest", "ruff", "build", "twine"]
+dev = ["pytest", "pytest-cov", "ruff", "build", "twine"]
 safety = ["transformers>=4.30.0"]
 
 [project.urls]

{alignmenter-0.1.0 → alignmenter-0.1.2}/src/alignmenter/cli.py

@@ -23,8 +23,10 @@ from alignmenter.providers.classifiers import load_safety_classifier
 from alignmenter.providers.judges import load_judge_provider
 from alignmenter.providers.openai import OpenAICustomGPTProvider
 from alignmenter.run_config import load_run_options
-from alignmenter.scripts.sanitize_dataset import sanitize_dataset_file
 from alignmenter.runner import RunConfig, Runner
+from alignmenter.scripts import bootstrap_dataset as bootstrap_dataset_script
+from alignmenter.scripts import calibrate_persona as calibrate_persona_script
+from alignmenter.scripts.sanitize_dataset import sanitize_dataset_file
 from alignmenter.scorers.authenticity import AuthenticityScorer
 from alignmenter.scorers.safety import SafetyScorer
 from alignmenter.scorers.stability import StabilityScorer
@@ -513,9 +515,9 @@ def run(
     judge: Optional[str] = typer.Option(None, help="Safety judge provider identifier (e.g. 'openai:gpt-4o-mini')."),
     judge_budget: Optional[int] = typer.Option(None, help="Maximum LLM judge calls per run."),
     generate_transcripts: bool = typer.Option(
-        True,
-        "--generate/--no-generate",
-        help="Call the model to generate fresh transcripts before scoring.",
+        False,
+        "--generate-transcripts",
+        help="Call providers to regenerate assistant turns before scoring (default reuses recorded transcripts).",
     ),
 ) -> None:
     """Execute an evaluation run."""
@@ -542,12 +544,25 @@ def run(
     assistant_turns = _lazy_assistant_turn_counter(inputs.dataset_path)
     _maybe_warn_about_cost(inputs, assistant_turns)
 
+    if not generate_transcripts:
+        typer.secho(
+            "Reusing recorded transcripts (pass --generate-transcripts to call providers).",
+            fg=typer.colors.BLUE,
+        )
+
     regenerate, provider, compare_provider_obj = _initialise_providers(
         inputs.model_identifier,
         inputs.compare_identifier,
         generate_transcripts,
     )
 
+    if generate_transcripts and not regenerate:
+        typer.secho(
+            "✗ Unable to generate transcripts because provider initialization failed. Re-run without --generate-transcripts or configure API keys.",
+            fg=typer.colors.RED,
+        )
+        raise typer.Exit(code=1)
+
     safety_classifier = load_safety_classifier(inputs.classifier_identifier)
     judge_provider = _initialise_judge_provider(inputs.judge_identifier)
     scorers, compare_scorers = _build_scorers_for_run(
@@ -614,6 +629,7 @@ def demo(
         embedding=None,
         judge=None,
         judge_budget=None,
+        generate_transcripts=True,
     )
 
 
@@ -939,6 +955,60 @@ def dataset_sanitize(
         typer.echo(f"  Output: {output_path}")
 
 
+def _run_bootstrap_dataset(
+    source: Optional[Path],
+    out: Path,
+    sessions: int,
+    turns_per_session: int,
+    safety_trap_ratio: float,
+    brand_trap_ratio: float,
+    persona_id: str,
+    seed: int,
+) -> None:
+    try:
+        bootstrap_dataset_script.bootstrap(
+            source=str(source) if source else None,
+            out=str(out),
+            sessions=sessions,
+            turns_per_session=turns_per_session,
+            safety_trap_ratio=safety_trap_ratio,
+            brand_trap_ratio=brand_trap_ratio,
+            persona_id=persona_id,
+            seed=seed,
+        )
+    except typer.Exit as exc:
+        raise exc
+    except Exception as exc:  # noqa: BLE001 - surface friendly error
+        typer.secho(f"✗ Error: {exc}", fg=typer.colors.RED, err=True)
+        raise typer.Exit(1) from exc
+
+
+@dataset_app.command("bootstrap")
+@app.command("bootstrap-dataset")
+def bootstrap_dataset_command(
+    source: Optional[Path] = typer.Option(None, "--source", help="Optional source dataset to augment."),
+    out: Path = typer.Option(..., "--out", help="Output JSONL path for generated data."),
+    sessions: int = typer.Option(10, "--sessions", help="Number of sessions to generate."),
+    turns_per_session: int = typer.Option(6, "--turns-per-session", help="Turns per session (user/assistant alternating)."),
+    safety_trap_ratio: float = typer.Option(0.2, "--safety-trap-ratio", help="Fraction of sessions with safety traps."),
+    brand_trap_ratio: float = typer.Option(0.2, "--brand-trap-ratio", help="Fraction of sessions with brand violations."),
+    persona_id: str = typer.Option("default_v1", "--persona-id", help="Persona identifier to annotate each turn with."),
+    seed: int = typer.Option(42, "--seed", help="Random seed for reproducibility."),
+) -> None:
+    """Generate a synthetic dataset with brand + safety traps."""
+
+    _run_bootstrap_dataset(
+        source=source,
+        out=out,
+        sessions=sessions,
+        turns_per_session=turns_per_session,
+        safety_trap_ratio=safety_trap_ratio,
+        brand_trap_ratio=brand_trap_ratio,
+        persona_id=persona_id,
+        seed=seed,
+    )
+
+
 # Calibration Commands
 
 
@@ -1168,6 +1238,35 @@ def analyze_scenarios(
         raise typer.Exit(1)
 
 
+@app.command("calibrate-persona")
+def calibrate_persona_command(
+    persona_path: Path = typer.Option(..., "--persona-path", help="Persona YAML containing the 'id' to calibrate."),
+    dataset: Path = typer.Option(..., "--dataset", help="Labeled JSONL with persona_id + label fields."),
+    out: Optional[Path] = typer.Option(None, "--out", help="Output path for generated .traits.json"),
+    min_samples: int = typer.Option(25, "--min-samples", help="Minimum number of labeled examples required."),
+    learning_rate: float = typer.Option(0.1, "--learning-rate", help="Learning rate for logistic regression."),
+    epochs: int = typer.Option(300, "--epochs", help="Training epochs."),
+    l2: float = typer.Option(0.0, "--l2", help="L2 regularization strength."),
+) -> None:
+    """Fit persona-specific trait weights from labeled data."""
+
+    try:
+        calibrate_persona_script.calibrate(
+            persona_path=str(persona_path),
+            dataset=str(dataset),
+            out=str(out) if out else None,
+            min_samples=min_samples,
+            learning_rate=learning_rate,
+            epochs=epochs,
+            l2=l2,
+        )
+    except typer.Exit as exc:
+        raise exc
+    except Exception as exc:  # noqa: BLE001 - friendly surface
+        typer.secho(f"✗ Error: {exc}", fg=typer.colors.RED, err=True)
+        raise typer.Exit(1) from exc
+
+
 def _load_env(path: Path) -> dict[str, str]:
     if not path.exists():
         return {}

{alignmenter-0.1.0 → alignmenter-0.1.2/src/alignmenter.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: alignmenter
-Version: 0.1.0
+Version: 0.1.2
 Summary: Persona-aligned evaluation toolkit for auditing conversational AI authenticity, safety, and stability.
 Author: Alignmenter
 License:                                  Apache License
@@ -236,6 +236,7 @@ Requires-Dist: tiktoken>=0.7.0
 Requires-Dist: requests>=2.32.0
 Provides-Extra: dev
 Requires-Dist: pytest; extra == "dev"
+Requires-Dist: pytest-cov; extra == "dev"
 Requires-Dist: ruff; extra == "dev"
 Requires-Dist: build; extra == "dev"
 Requires-Dist: twine; extra == "dev"
@@ -277,26 +278,23 @@ Unlike generic LLM evaluation frameworks, Alignmenter is purpose-built for **per
 
 ```bash
 # Clone the repository
-git clone https://github.com/yourusername/alignmenter.git
+git clone https://github.com/justinGrosvenor/alignmenter.git
 cd alignmenter
 
 # Create virtual environment
 python -m venv env
 source env/bin/activate  # On Windows: env\Scripts\activate
 
-# Install with dev dependencies
-pip install -e .[dev]
-
-# Optional: Install with offline safety classifier
-pip install -e .[dev,safety]
+# Install with dev + safety extras
+pip install -e ./alignmenter[dev,safety]
 ```
 
 ### Install from PyPI
 
 ```bash
-pip install alignmenter
+pip install "alignmenter[safety]"
 alignmenter init
-alignmenter run --config configs/run.yaml --no-generate
+alignmenter run --config configs/run.yaml --embedding sentence-transformer:all-MiniLM-L6-v2
 ```
 
 > **Note**: The `safety` extra includes `transformers` for the offline safety classifier (ProtectAI/distilled-safety-roberta). Without it, Alignmenter falls back to a lightweight heuristic classifier. See [docs/offline_safety.md](https://github.com/justinGrosvenor/alignmenter/blob/main/docs/offline_safety.md) for details.
@@ -314,13 +312,16 @@ alignmenter run \
   --persona configs/persona/default.yaml
 
 # Reuse existing transcripts without hitting the provider
-alignmenter run --config configs/run.yaml --no-generate
+alignmenter run --config configs/run.yaml
 
 # View interactive HTML report
 alignmenter report --last
 
 # Sanitize a dataset in-place or to a new file
 alignmenter dataset sanitize datasets/demo_conversations.jsonl --out datasets/demo_sanitized.jsonl
+
+# Generate fresh transcripts (requires provider access)
+alignmenter run --config configs/run.yaml --generate-transcripts
 ```
 
 **Output:**
@@ -438,6 +439,14 @@ Quick links:
 - **[LLM Judges](https://docs.alignmenter.com/guides/llm-judges/)** - Qualitative analysis
 - **[Contributing](https://docs.alignmenter.com/contributing/)** - How to contribute
 
+---
+
+## Case Studies
+
+- **[Wendy's Twitter Voice](../docs/case-studies/wendys-twitter.md)** - End-to-end calibration example using the included case-study assets. *(Available when running from the source repo; not included in the PyPI wheel.)*
+
+---
+
 ## Usage Examples
 
 ### Evaluate Multiple Models
@@ -490,11 +499,9 @@ alignmenter calibrate-persona \
 
 ```bash
 # Remove PII before evaluation
-alignmenter sanitize-dataset \
-  --input prod_logs.jsonl \
+alignmenter dataset sanitize prod_logs.jsonl \
   --out datasets/sanitized.jsonl \
-  --hash-session-ids \
-  --scrub-patterns "email|phone|ssn"
+  --no-use-hashing
 ```
 
 ## Persona Configuration
@@ -646,13 +653,25 @@ make report-last
 
 ## Roadmap
 
-- [ ] Web dashboard for team collaboration
-- [ ] Hosted evaluation pipelines (CI/CD SaaS)
-- [ ] Multi-language support (non-English personas)
-- [ ] Offline safety model (no API required)
-- [ ] Cost/latency tracking in scorers
-- [ ] Competitive landscape analysis
-- [ ] Metric cards and ablation studies
+### Completed ✅
+- Three-dimensional scoring (authenticity, safety, stability)
+- Multi-provider support (OpenAI, Anthropic, local models)
+- HTML report cards with interactive charts
+- Offline safety classifier (distilled-safety-roberta)
+- LLM judges for qualitative analysis
+- Budget guardrails and cost tracking
+- PII sanitization tools
+- Calibration workflow and diagnostics
+
+### In Progress 🚧
+- Multi-language support (non-English personas)
+- Batch processing optimizations
+- Additional embedding providers
+
+### Future Considerations 💭
+- Synthetic test case generation
+- Custom metric plugins
+- Advanced trait models (neural networks)
 
 ## Contributing
 
@@ -676,14 +695,15 @@ If you use Alignmenter in research, please cite:
 @software{alignmenter2024,
   title={Alignmenter: A Framework for Persona-Aligned Conversational AI Evaluation},
   author={Alignmenter Contributors},
-  year={2024},
-  url={https://github.com/yourusername/alignmenter}
+  year={2025},
+  url={https://github.com/justinGrosvenor/alignmenter},
+  license={Apache-2.0}
 }
 ```
 
 ## Support
 
-- **Documentation**: [docs/](https://github.com/justinGrosvenor/alignmenter/tree/main/docs)
+- **Documentation**: [docs.alignmenter.com](https://docs.alignmenter.com)
 - **Issues**: [GitHub Issues](https://github.com/justinGrosvenor/alignmenter/issues)
 - **Discussions**: [GitHub Discussions](https://github.com/justinGrosvenor/alignmenter/discussions)
 

{alignmenter-0.1.0 → alignmenter-0.1.2}/src/alignmenter.egg-info/requires.txt

@@ -13,6 +13,7 @@ requests>=2.32.0
 
 [dev]
 pytest
+pytest-cov
 ruff
 build
 twine

{alignmenter-0.1.0 → alignmenter-0.1.2}/tests/test_cli_helpers.py

@@ -10,6 +10,9 @@ from typer.testing import CliRunner
 from alignmenter import app
 
 
+DATA_DIR = Path(__file__).resolve().parent / "data"
+MINI_DATASET = DATA_DIR / "mini_cli_dataset.jsonl"
+
 runner = CliRunner()
 
 
@@ -149,9 +152,75 @@ def test_dataset_sanitize_command(tmp_path: Path) -> None:
     assert "[DRY RUN]" in result.output
 
 
+def test_bootstrap_dataset_command(tmp_path: Path) -> None:
+    out_path = tmp_path / "generated.jsonl"
+    result = runner.invoke(
+        app,
+        [
+            "bootstrap-dataset",
+            "--out",
+            str(out_path),
+            "--sessions",
+            "2",
+            "--turns-per-session",
+            "2",
+            "--safety-trap-ratio",
+            "0.5",
+            "--brand-trap-ratio",
+            "0.0",
+            "--persona-id",
+            "demo_v1",
+            "--seed",
+            "7",
+        ],
+    )
+
+    assert result.exit_code == 0, result.output
+    assert out_path.exists()
+    lines = out_path.read_text().strip().splitlines()
+    assert len(lines) == 4  # 2 sessions × 2 turns
+
+
+def test_calibrate_persona_command(tmp_path: Path) -> None:
+    persona_path = tmp_path / "persona.yaml"
+    persona_path.write_text("id: sample_persona\n", encoding="utf-8")
+
+    dataset_path = tmp_path / "annotations.jsonl"
+    dataset_path.write_text(
+        json.dumps({"text": "On brand reply", "label": 1, "persona_id": "sample_persona"})
+        + "\n"
+        + json.dumps({"text": "off brand hype", "label": 0, "persona_id": "sample_persona"})
+        + "\n",
+        encoding="utf-8",
+    )
+
+    out_path = tmp_path / "persona.traits.json"
+    result = runner.invoke(
+        app,
+        [
+            "calibrate-persona",
+            "--persona-path",
+            str(persona_path),
+            "--dataset",
+            str(dataset_path),
+            "--out",
+            str(out_path),
+            "--min-samples",
+            "2",
+            "--epochs",
+            "10",
+        ],
+    )
+
+    assert result.exit_code == 0, result.output
+    assert out_path.exists()
+    payload = json.loads(out_path.read_text())
+    assert "trait_model" in payload
+
+
 def test_run_command(tmp_path: Path) -> None:
     root = Path(__file__).resolve().parents[2] / "alignmenter"
-    dataset = root / "datasets" / "demo_conversations.jsonl"
+    dataset = MINI_DATASET
     persona = root / "configs" / "persona" / "default.yaml"
     keywords = root / "configs" / "safety_keywords.yaml"
 
@@ -169,7 +238,6 @@ def test_run_command(tmp_path: Path) -> None:
             str(keywords),
             "--out",
             str(tmp_path),
-            "--no-generate",
         ],
     )
 
@@ -185,7 +253,7 @@ def test_run_command(tmp_path: Path) -> None:
 
 def test_run_command_with_compare(tmp_path: Path) -> None:
     root = Path(__file__).resolve().parents[2] / "alignmenter"
-    dataset = root / "datasets" / "demo_conversations.jsonl"
+    dataset = MINI_DATASET
     persona = root / "configs" / "persona" / "default.yaml"
     keywords = root / "configs" / "safety_keywords.yaml"
 
@@ -205,7 +273,6 @@ def test_run_command_with_compare(tmp_path: Path) -> None:
             str(keywords),
             "--out",
             str(tmp_path),
-            "--no-generate",
         ],
     )
 
@@ -224,7 +291,7 @@ def test_run_command_threshold_failure(tmp_path: Path) -> None:
             [
                 "run_id: threshold",
                 "model: openai:gpt-4o-mini",
-                f"dataset: {root / 'datasets' / 'demo_conversations.jsonl'}",
+                f"dataset: {MINI_DATASET}",
                 f"persona: {root / 'configs' / 'persona' / 'default.yaml'}",
                 f"keywords: {root / 'configs' / 'safety_keywords.yaml'}",
                 "scorers:",
@@ -243,7 +310,6 @@ def test_run_command_threshold_failure(tmp_path: Path) -> None:
             "run",
             "--config",
             str(config_path),
-            "--no-generate",
         ],
     )
 

alignmenter-0.1.2/tests/test_cli_run_config.py

@@ -0,0 +1,46 @@
+"""Tests CLI integration with environment config defaults."""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+from typer.testing import CliRunner
+
+from alignmenter import app
+
+DATA_DIR = Path(__file__).resolve().parent / "data"
+MINI_DATASET = DATA_DIR / "mini_cli_dataset.jsonl"
+
+runner = CliRunner()
+
+
+def test_cli_run_uses_run_config_file(tmp_path: Path) -> None:
+    root = Path(__file__).resolve().parents[2] / "alignmenter"
+    persona = root / "configs" / "persona" / "default.yaml"
+    keywords = root / "configs" / "safety_keywords.yaml"
+
+    config_path = tmp_path / "mini_config.yaml"
+    config_path.write_text(
+        "\n".join(
+            [
+                "run_id: cli-mini",
+                f"model: openai:gpt-4o-mini",
+                f"dataset: {MINI_DATASET}",
+                f"persona: {persona}",
+                f"keywords: {keywords}",
+            ]
+        )
+        + "\n",
+        encoding="utf-8",
+    )
+
+    result = runner.invoke(
+        app,
+        [
+            "run",
+            "--config", str(config_path),
+            "--out", str(tmp_path),
+        ],
+    )
+
+    assert result.exit_code == 0, result.output

alignmenter-0.1.0/tests/test_cli_run_config.py

@@ -1,28 +0,0 @@
-"""Tests CLI integration with environment config defaults."""
-
-from __future__ import annotations
-
-from pathlib import Path
-
-from typer.testing import CliRunner
-
-from alignmenter import app
-
-runner = CliRunner()
-
-
-def test_cli_run_uses_run_config_file(tmp_path: Path) -> None:
-    root = Path(__file__).resolve().parents[2] / "alignmenter"
-    config_path = root / "configs" / "demo_config.yaml"
-
-    result = runner.invoke(
-        app,
-        [
-            "run",
-            "--config", str(config_path),
-            "--out", str(tmp_path),
-            "--no-generate",
-        ],
-    )
-
-    assert result.exit_code == 0, result.output