remdb 0.3.0__py3-none-any.whl → 0.3.127__py3-none-any.whl

rem/cli/commands/experiments.py

@@ -514,6 +514,159 @@ def show(name: str, base_path: Optional[str]):
         raise click.Abort()
 
 
+# =============================================================================
+# VIBES MODE HELPER
+# =============================================================================
+
+
+def _run_vibes_mode(
+    config: Any,
+    dataset_df: Any,
+    task_fn: Any,
+    base_path: str,
+    limit: Optional[int],
+    evaluator_schema_path: Path,
+) -> None:
+    """Run experiment in vibes mode - execute agent and export for AI evaluation.
+
+    Vibes mode runs the agent on each example and saves results to a JSONL file.
+    The AI assistant (e.g., Claude Code) then acts as the judge using the
+    evaluator schema to evaluate results.
+
+    Args:
+        config: ExperimentConfig object
+        dataset_df: Polars DataFrame with ground truth examples
+        task_fn: Function to run agent on each example
+        base_path: Base directory for experiments
+        limit: Optional limit on number of examples to process
+        evaluator_schema_path: Path to the evaluator schema YAML file
+    """
+    from rem.utils.date_utils import format_timestamp_for_experiment, utc_now, to_iso
+    import json
+
+    # Apply limit if specified
+    if limit:
+        dataset_df = dataset_df.head(limit)
+        click.echo(f"  (Limited to {limit} examples)")
+
+    # Create results directory
+    timestamp = format_timestamp_for_experiment()
+    results_dir = Path(base_path) / config.name / "results" / timestamp
+    results_dir.mkdir(parents=True, exist_ok=True)
+
+    click.echo(f"\n⏳ Running agent on {len(dataset_df)} examples...")
+    click.echo(f"   Results will be saved to: {results_dir}")
+    click.echo()
+
+    # Run agent on each example and collect results
+    results = []
+    records = dataset_df.to_dicts()
+
+    for i, record in enumerate(records, 1):
+        example_id = record.get("id", i)
+        click.echo(f"  [{i}/{len(records)}] Processing example {example_id}...", nl=False)
+
+        try:
+            # Prepare input for agent
+            input_text = record.get("text", record.get("input", record.get("query", "")))
+            example_input = {"query": input_text} if isinstance(input_text, str) else input_text
+
+            # Run agent
+            output = task_fn({"input": example_input})
+
+            result = {
+                "id": example_id,
+                "input": input_text,
+                "ground_truth": record.get("ground_truth", record.get("expected_output", "")),
+                "category": record.get("category", ""),
+                "agent_output": output,
+                "status": "success",
+            }
+            click.echo(" ✓")
+
+        except Exception as e:
+            result = {
+                "id": example_id,
+                "input": record.get("text", record.get("input", "")),
+                "ground_truth": record.get("ground_truth", record.get("expected_output", "")),
+                "category": record.get("category", ""),
+                "agent_output": None,
+                "status": "error",
+                "error": str(e),
+            }
+            click.echo(f" ✗ ({e})")
+
+        results.append(result)
+
+    # Save results to JSONL
+    results_file = results_dir / "vibes-results.jsonl"
+    with open(results_file, "w") as f:
+        for result in results:
+            f.write(json.dumps(result) + "\n")
+
+    # Copy evaluator schema to results dir for easy reference
+    import shutil
+    evaluator_copy = results_dir / "evaluator-schema.yaml"
+    shutil.copy(evaluator_schema_path, evaluator_copy)
+
+    # Save run metadata
+    run_info = {
+        "experiment": config.name,
+        "agent": config.agent_schema_ref.name,
+        "evaluator": config.evaluator_schema_ref.name,
+        "mode": "vibes",
+        "timestamp": timestamp,
+        "total_examples": len(records),
+        "successful": len([r for r in results if r["status"] == "success"]),
+        "failed": len([r for r in results if r["status"] == "error"]),
+        "completed_at": to_iso(utc_now()),
+    }
+
+    run_info_file = results_dir / "run-info.json"
+    with open(run_info_file, "w") as f:
+        json.dump(run_info, f, indent=2)
+
+    # Print summary and instructions
+    success_count = run_info["successful"]
+    fail_count = run_info["failed"]
+
+    click.echo(f"\n{'=' * 60}")
+    click.echo(f"VIBES MODE COMPLETE")
+    click.echo(f"{'=' * 60}")
+    click.echo(f"\nResults: {success_count} successful, {fail_count} failed")
+    click.echo(f"\nFiles saved to: {results_dir}/")
+    click.echo(f"  - vibes-results.jsonl    (agent outputs)")
+    click.echo(f"  - evaluator-schema.yaml  (evaluation criteria)")
+    click.echo(f"  - run-info.json          (run metadata)")
+
+    click.echo(f"\n{'=' * 60}")
+    click.echo(f"NEXT STEP: Ask your AI assistant to evaluate")
+    click.echo(f"{'=' * 60}")
+    click.echo(f"""
+Copy this prompt to Claude Code or your AI assistant:
+
+    Please evaluate the experiment results in:
+    {results_dir}/
+
+    Read the vibes-results.jsonl file and evaluate each example
+    using the evaluator schema in evaluator-schema.yaml.
+
+    For each example, provide:
+    1. extracted_classification
+    2. exact_match (vs ground_truth)
+    3. semantic_match
+    4. reasoning_quality_score
+    5. overall_score
+    6. pass/fail
+
+    Then provide summary metrics:
+    - Exact match accuracy
+    - Semantic match accuracy
+    - Average overall score
+    - Pass rate
+""")
+
+
 # =============================================================================
 # RUN COMMAND
 # =============================================================================
@@ -524,6 +677,8 @@ def show(name: str, base_path: Optional[str]):
 @click.option("--base-path", help="Base directory for experiments (default: EXPERIMENTS_HOME or 'experiments')")
 @click.option("--version", help="Git tag version to load (e.g., 'experiments/my-exp/v1.0.0')")
 @click.option("--dry-run", is_flag=True, help="Test on small subset without saving")
+@click.option("--only-vibes", is_flag=True, help="Run agent locally, export results for AI evaluation (no Phoenix)")
+@click.option("--limit", "-n", type=int, help="Limit number of examples to evaluate (useful with --only-vibes)")
 @click.option("--update-prompts", is_flag=True, help="Update prompts in Phoenix before running")
 @click.option("--phoenix-url", help="Phoenix server URL (overrides PHOENIX_BASE_URL env var)")
 @click.option("--phoenix-api-key", help="Phoenix API key (overrides PHOENIX_API_KEY env var)")
@@ -532,14 +687,45 @@ def run(
     base_path: Optional[str],
     version: Optional[str],
     dry_run: bool,
+    only_vibes: bool,
+    limit: Optional[int],
     update_prompts: bool,
     phoenix_url: Optional[str],
     phoenix_api_key: Optional[str],
 ):
-    """Run an experiment using Phoenix provider.
+    """Run an experiment using Phoenix provider or local vibes mode.
 
     Loads configuration, executes agent and evaluator, saves results.
 
+    Vibes Mode (--only-vibes):
+        Run agent locally without Phoenix infrastructure. Agent outputs are saved
+        to a JSONL file along with the evaluator schema. Your AI assistant (e.g.,
+        Claude Code) then acts as the judge to evaluate results.
+
+        This enables seamless switching between:
+        - Local evaluation: Quick iteration with AI-as-judge
+        - Phoenix evaluation: Production metrics and dashboards
+
+        Usage:
+            rem experiments run my-experiment --only-vibes
+            rem experiments run my-experiment --only-vibes --limit 5
+
+        The command will:
+        1. Run the agent on each ground-truth example
+        2. Save results to results/{timestamp}/vibes-results.jsonl
+        3. Print the evaluator prompt and schema
+        4. Instruct you to ask your AI assistant to evaluate
+
+        Example workflow with Claude Code:
+            $ rem experiments run mental-health-classifier --only-vibes --limit 3
+            # ... agent runs ...
+            # Results saved to: .experiments/mental-health-classifier/results/20241203-143022/
+
+            # Then ask Claude Code:
+            "Please evaluate the experiment results in
+             .experiments/mental-health-classifier/results/20241203-143022/
+             using the evaluator schema provided"
+
     Phoenix Connection:
         Commands respect PHOENIX_BASE_URL and PHOENIX_API_KEY environment variables.
         Defaults to localhost:6006 for local development.
@@ -562,6 +748,12 @@ def run(
         # Run experiment with latest schemas
         rem experiments run hello-world-validation
 
+        # Quick local evaluation (vibes mode)
+        rem experiments run hello-world-validation --only-vibes
+
+        # Vibes mode with limited examples
+        rem experiments run hello-world-validation --only-vibes --limit 5
+
         # Run specific version
         rem experiments run hello-world-validation \\
             --version experiments/hello-world-validation/v1.0.0
@@ -578,8 +770,7 @@ def run(
     from rem.services.git import GitService
     from rem.services.phoenix import PhoenixClient
     from rem.agentic.providers.phoenix import create_evaluator_from_schema
-    from datetime import datetime
-    import pandas as pd
+    from rem.utils.date_utils import utc_now, to_iso, format_timestamp_for_experiment
     import os
 
     try:
@@ -615,36 +806,22 @@ def run(
             click.echo(f"  Mode: DRY RUN (no data will be saved)")
         click.echo()
 
-        # Load agent schema from Git or filesystem
+        # Load agent schema using centralized schema loader
         agent_name = config.agent_schema_ref.name
         agent_version = config.agent_schema_ref.version
 
         click.echo(f"Loading agent schema: {agent_name} (version: {agent_version or 'latest'})")
 
-        # Try Git first, fallback to filesystem
-        agent_schema = None
-        try:
-            git_svc = GitService()
-            agent_schema = git_svc.load_schema(agent_name, version=agent_version)
-            click.echo(f"✓ Loaded agent schema from Git")
-        except Exception as e:
-            logger.debug(f"Git not available, trying filesystem: {e}")
-
-            # Fallback to local filesystem
-            from rem.services.fs import FS
-            fs = FS()
+        from rem.utils.schema_loader import load_agent_schema
 
-            schema_path = f"schemas/agents/{agent_name}.yaml"
-            try:
-                agent_schema = fs.read(schema_path)
-                click.echo(f"✓ Loaded agent schema from filesystem")
-            except Exception as fs_error:
-                logger.error(f"Failed to load agent schema: Git: {e}, FS: {fs_error}")
-                click.echo(f"Error: Could not load agent schema '{agent_name}'")
-                click.echo(f"  Tried Git: {e}")
-                click.echo(f"  Tried filesystem: {schema_path}")
-                click.echo(f"  Make sure the schema exists")
-                raise click.Abort()
+        try:
+            agent_schema = load_agent_schema(agent_name)
+            click.echo(f"✓ Loaded agent schema: {agent_name}")
+        except FileNotFoundError as e:
+            logger.error(f"Failed to load agent schema: {e}")
+            click.echo(f"Error: Could not load agent schema '{agent_name}'")
+            click.echo(f"  {e}")
+            raise click.Abort()
 
         # Create agent function from schema
         from rem.agentic.providers.pydantic_ai import create_agent
@@ -683,73 +860,97 @@ def run(
                 return {"output": serialized}
             return serialized if isinstance(serialized, dict) else {"output": str(serialized)}
 
-        # Load evaluator schema
+        # Load evaluator schema using centralized schema loader
         evaluator_name = config.evaluator_schema_ref.name
         evaluator_version = config.evaluator_schema_ref.version
 
-        # Resolve evaluator path (evaluators are organized by agent name)
-        evaluator_schema_path = f"rem/schemas/evaluators/{agent_name}/{evaluator_name}.yaml"
-
         click.echo(f"Loading evaluator: {evaluator_name} for agent {agent_name}")
 
-        try:
-            evaluator_fn = create_evaluator_from_schema(
-                evaluator_schema_path=evaluator_schema_path,
-                model_name=None,  # Use default from schema
-            )
-            click.echo(f"✓ Loaded evaluator schema")
-        except Exception as e:
-            logger.warning(f"Failed to load evaluator: {e}")
-            click.echo(f"Error: Could not load evaluator schema")
-            click.echo(f"  Path: {evaluator_schema_path}")
-            click.echo(f"  Make sure the schema exists")
+        # Find evaluator schema file path
+        from rem.utils.schema_loader import get_evaluator_schema_path
+
+        evaluator_schema_path = get_evaluator_schema_path(evaluator_name)
+        if not evaluator_schema_path or not evaluator_schema_path.exists():
+            click.echo(f"Error: Could not find evaluator schema '{evaluator_name}'")
+            raise click.Abort()
+
+        click.echo(f"✓ Found evaluator schema: {evaluator_schema_path}")
+
+        # For Phoenix mode, also load evaluator function
+        evaluator_fn = None
+        if not only_vibes:
+            # Try multiple evaluator path patterns (agent-specific, then generic)
+            evaluator_paths_to_try = [
+                f"{agent_name}/{evaluator_name}",  # e.g., hello-world/default
+                f"{agent_name}-{evaluator_name}",  # e.g., hello-world-default
+                evaluator_name,                     # e.g., default (generic)
+            ]
+
+            evaluator_load_error = None
+
+            for evaluator_path in evaluator_paths_to_try:
+                try:
+                    evaluator_fn = create_evaluator_from_schema(
+                        evaluator_schema_path=evaluator_path,
+                        model_name=None,  # Use default from schema
+                    )
+                    click.echo(f"✓ Loaded evaluator function: {evaluator_path}")
+                    break
+                except FileNotFoundError as e:
+                    evaluator_load_error = e
+                    logger.debug(f"Evaluator not found at {evaluator_path}: {e}")
+                    continue
+                except Exception as e:
+                    evaluator_load_error = e
+                    logger.warning(f"Failed to load evaluator from {evaluator_path}: {e}")
+                    continue
+
+        if evaluator_fn is None and not only_vibes:
+            click.echo(f"Error: Could not load evaluator function '{evaluator_name}'")
+            click.echo(f"  Tried paths: {evaluator_paths_to_try}")
+            if evaluator_load_error:
+                click.echo(f"  Last error: {evaluator_load_error}")
             raise click.Abort()
 
-        # Load dataset
+        # Load dataset using Polars
+        import polars as pl
+
         click.echo(f"Loading dataset: {list(config.datasets.keys())[0]}")
         dataset_ref = list(config.datasets.values())[0]
 
         if dataset_ref.location.value == "git":
-            # Load from Git
+            # Load from Git (local filesystem)
             dataset_path = Path(base_path) / name / dataset_ref.path
             if not dataset_path.exists():
                 click.echo(f"Error: Dataset not found: {dataset_path}")
                 raise click.Abort()
 
             if dataset_ref.format == "csv":
-                dataset_df = pd.read_csv(dataset_path)
+                dataset_df = pl.read_csv(dataset_path)
             elif dataset_ref.format == "parquet":
-                dataset_df = pd.read_parquet(dataset_path)
+                dataset_df = pl.read_parquet(dataset_path)
             elif dataset_ref.format == "jsonl":
-                dataset_df = pd.read_json(dataset_path, lines=True)
+                dataset_df = pl.read_ndjson(dataset_path)
             else:
                 click.echo(f"Error: Format '{dataset_ref.format}' not yet supported")
                 raise click.Abort()
         elif dataset_ref.location.value in ["s3", "hybrid"]:
             # Load from S3 using FS provider
             from rem.services.fs import FS
+            from io import BytesIO
 
             fs = FS()
 
             try:
                 if dataset_ref.format == "csv":
                     content = fs.read(dataset_ref.path)
-                    from io import StringIO
-                    dataset_df = pd.read_csv(StringIO(content))
+                    dataset_df = pl.read_csv(BytesIO(content.encode() if isinstance(content, str) else content))
                 elif dataset_ref.format == "parquet":
-                    # For parquet, we need binary read
-                    import tempfile
-                    with tempfile.NamedTemporaryFile(suffix=".parquet", delete=False) as tmp:
-                        tmp_path = tmp.name
-                        # Download via FS
-                        content_bytes = fs.read(dataset_ref.path)
-                        tmp.write(content_bytes)
-                    dataset_df = pd.read_parquet(tmp_path)
-                    Path(tmp_path).unlink()  # Clean up temp file
+                    content_bytes = fs.read(dataset_ref.path)
+                    dataset_df = pl.read_parquet(BytesIO(content_bytes if isinstance(content_bytes, bytes) else content_bytes.encode()))
                 elif dataset_ref.format == "jsonl":
                     content = fs.read(dataset_ref.path)
-                    from io import StringIO
-                    dataset_df = pd.read_json(StringIO(content), lines=True)
+                    dataset_df = pl.read_ndjson(BytesIO(content.encode() if isinstance(content, str) else content))
                 else:
                     click.echo(f"Error: Format '{dataset_ref.format}' not yet supported")
                     raise click.Abort()
@@ -772,6 +973,18 @@ def run(
             # TODO: Implement prompt updating
             click.echo("⚠  --update-prompts not yet implemented")
 
+        # Vibes mode: run agent and export for AI evaluation
+        if only_vibes:
+            _run_vibes_mode(
+                config=config,
+                dataset_df=dataset_df,
+                task_fn=task_fn,
+                base_path=base_path,
+                limit=limit,
+                evaluator_schema_path=evaluator_schema_path,
+            )
+            return
+
         # Run experiment via Phoenix
         if not dry_run:
             # Create Phoenix client with optional overrides
@@ -793,13 +1006,13 @@ def run(
 
             client = PhoenixClient(config=phoenix_config)
 
-            experiment_name = f"{config.name}-{datetime.now().strftime('%Y%m%d-%H%M%S')}"
+            experiment_name = f"{config.name}-{format_timestamp_for_experiment()}"
 
             click.echo(f"\n⏳ Running experiment: {experiment_name}")
             click.echo(f"   This may take several minutes...")
 
             experiment = client.run_experiment(
-                dataset=dataset_df,  # type: ignore[arg-type]
+                dataset=dataset_df,
                 task=task_fn,
                 evaluators=[evaluator_fn],
                 experiment_name=experiment_name,
@@ -809,12 +1022,15 @@ def run(
                     "evaluator": config.evaluator_schema_ref.name,
                     "experiment_config": config.name,
                     **config.metadata
-                }
+                },
+                # Smart column detection for DataFrame -> Phoenix Dataset conversion
+                input_keys=["input"] if "input" in dataset_df.columns else None,
+                output_keys=["expected_output"] if "expected_output" in dataset_df.columns else None,
             )
 
             # Update experiment status
             config.status = ExperimentStatus.COMPLETED
-            config.last_run_at = datetime.now()
+            config.last_run_at = utc_now()
             if not version:  # Only save if not loading from Git
                 config.save(base_path)
 
@@ -835,7 +1051,7 @@ def run(
                         "agent": config.agent_schema_ref.name,
                         "evaluator": config.evaluator_schema_ref.name,
                         "dataset_size": len(dataset_df),
-                        "completed_at": datetime.now().isoformat(),
+                        "completed_at": to_iso(utc_now()),
                         "phoenix_url": getattr(experiment, "url", None),
                         "task_runs": len(exp_data.get("task_runs", [])),
                     }
@@ -1015,20 +1231,24 @@ def dataset_add(
             --output-keys expected_label,expected_type
     """
     from rem.services.phoenix import PhoenixClient
-    import pandas as pd
+    import polars as pl
 
     try:
         client = PhoenixClient()
 
-        # Load CSV
-        df = pd.read_csv(from_csv)
+        # Load CSV with Polars
+        df = pl.read_csv(from_csv)
+        records = df.to_dicts()
 
         # Extract data
-        inputs = cast(list[dict[str, Any]], df[input_keys.split(",")].to_dict("records"))
-        outputs = cast(list[dict[str, Any]], df[output_keys.split(",")].to_dict("records"))
+        input_cols = input_keys.split(",")
+        output_cols = output_keys.split(",")
+        inputs = [{k: row.get(k) for k in input_cols} for row in records]
+        outputs = [{k: row.get(k) for k in output_cols} for row in records]
         metadata = None
         if metadata_keys:
-            metadata = cast(list[dict[str, Any]], df[metadata_keys.split(",")].to_dict("records"))
+            meta_cols = metadata_keys.split(",")
+            metadata = [{k: row.get(k) for k in meta_cols} for row in records]
 
         # Add to dataset
         dataset = client.add_examples_to_dataset(
@@ -1269,12 +1489,12 @@ def trace_list(
         rem experiments trace list --project rem-agents --days 7 --limit 50
     """
     from rem.services.phoenix import PhoenixClient
-    from datetime import datetime, timedelta
+    from rem.utils.date_utils import days_ago
 
     try:
         client = PhoenixClient()
 
-        start_time = datetime.now() - timedelta(days=days)
+        start_time = days_ago(days)
 
         traces_df = client.get_traces(
             project_name=project,

rem/cli/commands/process.py

@@ -12,12 +12,12 @@ from rem.services.content import ContentService
 
 @click.command(name="ingest")
 @click.argument("file_path", type=click.Path(exists=True))
-@click.option("--user-id", required=True, help="User ID to own the file")
+@click.option("--user-id", default=None, help="User ID to scope file privately (default: public/shared)")
 @click.option("--category", help="Optional file category")
 @click.option("--tags", help="Optional comma-separated tags")
 def process_ingest(
     file_path: str,
-    user_id: str,
+    user_id: str | None,
     category: str | None,
     tags: str | None,
 ):
@@ -32,8 +32,9 @@ def process_ingest(
     5. Creates a File entity record.
 
     Examples:
-        rem process ingest sample.pdf --user-id user-123
-        rem process ingest contract.docx --user-id user-123 --category legal --tags contract,2023
+        rem process ingest sample.pdf
+        rem process ingest contract.docx --category legal --tags contract,2023
+        rem process ingest agent.yaml  # Auto-detects kind=agent, saves to schemas table
     """
     import asyncio
     from ...services.content import ContentService
@@ -56,7 +57,8 @@ def process_ingest(
 
             tag_list = tags.split(",") if tags else None
 
-            logger.info(f"Ingesting file: {file_path} for user: {user_id}")
+            scope_msg = f"user: {user_id}" if user_id else "public"
+            logger.info(f"Ingesting file: {file_path} ({scope_msg})")
             result = await service.ingest_file(
                 file_uri=file_path,
                 user_id=user_id,
@@ -65,11 +67,15 @@ def process_ingest(
                 is_local_server=True, # CLI is local
             )
 
-            if result.get("processing_status") == "completed":
-                logger.success(f"File ingested successfully: {result['file_name']}")
+            # Handle schema ingestion (agents/evaluators)
+            if result.get("schema_name"):
+                logger.success(f"Schema ingested: {result['schema_name']} (kind={result.get('kind', 'agent')})")
+                logger.info(f"Version: {result.get('version', '1.0.0')}")
+            # Handle file ingestion
+            elif result.get("processing_status") == "completed":
+                logger.success(f"File ingested: {result['file_name']}")
                 logger.info(f"File ID: {result['file_id']}")
                 logger.info(f"Resources created: {result['resources_created']}")
-                logger.info(f"Status: {result['processing_status']}")
             else:
                 logger.error(f"Ingestion failed: {result.get('message', 'Unknown error')}")
                 sys.exit(1)
@@ -192,15 +198,13 @@ def process_uri(uri: str, output: str, save: str | None):
 
 
 @click.command(name="files")
-@click.option("--tenant-id", required=True, help="Tenant ID")
-@click.option("--user-id", help="Filter by user ID")
+@click.option("--user-id", default=None, help="User ID (default: from settings)")
 @click.option("--status", type=click.Choice(["pending", "processing", "completed", "failed"]), help="Filter by status")
 @click.option("--extractor", help="Run files through custom extractor (e.g., cv-parser-v1)")
 @click.option("--limit", type=int, help="Max files to process")
 @click.option("--provider", help="Optional LLM provider override")
 @click.option("--model", help="Optional model override")
 def process_files(
-    tenant_id: str,
     user_id: Optional[str],
     status: Optional[str],
     extractor: Optional[str],
@@ -217,19 +221,22 @@ def process_files(
 
         \b
         # List completed files
-        rem process files --tenant-id acme-corp --status completed
+        rem process files --status completed
 
         \b
         # Extract from CV files
-        rem process files --tenant-id acme-corp --extractor cv-parser-v1 --limit 10
+        rem process files --extractor cv-parser-v1 --limit 10
 
         \b
         # Extract with provider override
-        rem process files --tenant-id acme-corp --extractor contract-analyzer-v1 \\
+        rem process files --extractor contract-analyzer-v1 \\
             --provider anthropic --model claude-sonnet-4-5
     """
+    from ...settings import settings
+    effective_user_id = user_id or settings.test.effective_user_id
+
     logger.warning("Not implemented yet")
-    logger.info(f"Would process files for tenant: {tenant_id}")
+    logger.info(f"Would process files for user: {effective_user_id}")
 
     if user_id:
         logger.info(f"Filter: user_id={user_id}")

rem/cli/commands/scaffold.py

@@ -0,0 +1,47 @@
+"""
+Scaffold command - generate project structure for REM-based applications.
+
+TODO: Implement this command to generate:
+- my_app/main.py (entry point with create_app)
+- my_app/models.py (example CoreModel subclass)
+- my_app/routers/ (example FastAPI router)
+- schemas/agents/ (example agent schema)
+- schemas/evaluators/ (example evaluator)
+- sql/migrations/ (empty migrations directory)
+- pyproject.toml (with remdb dependency)
+- README.md (basic usage instructions)
+
+Usage:
+    rem scaffold my-app
+    rem scaffold my-app --with-examples  # Include example models/routers/tools
+"""
+
+import click
+
+
+@click.command()
+@click.argument("name")
+@click.option("--with-examples", is_flag=True, help="Include example code")
+def scaffold(name: str, with_examples: bool) -> None:
+    """
+    Generate a new REM-based project structure.
+
+    NAME is the project directory name to create.
+    """
+    click.echo(f"TODO: Scaffold command not yet implemented")
+    click.echo(f"Would create project: {name}")
+    click.echo(f"With examples: {with_examples}")
+    click.echo()
+    click.echo("For now, manually create this structure:")
+    click.echo(f"""
+{name}/
+├── {name.replace('-', '_')}/
+│   ├── main.py           # Entry point (create_app + extensions)
+│   ├── models.py         # Custom models (inherit CoreModel)
+│   └── routers/          # Custom FastAPI routers
+├── schemas/
+│   ├── agents/           # Custom agent YAML schemas
+│   └── evaluators/       # Custom evaluator schemas
+├── sql/migrations/       # Custom SQL migrations
+└── pyproject.toml
+""")