sembl 0.1.0__py3-none-any.whl

sembl/__init__.py

@@ -0,0 +1,6 @@
+"""
+Sembl — Work Order generator for AI software development.
+Turn messy repo intent into scoped execution contracts.
+"""
+
+__version__ = "0.1.0"

sembl/cli.py

@@ -0,0 +1,381 @@
+"""
+cli.py
+
+Sembl CLI — the first machine.
+
+Commands:
+  sembl generate  — repo + task → Work Order
+  sembl show      — display the latest Work Order
+  sembl list      — list all Work Orders in this repo
+"""
+
+import os
+import json
+import sys
+from pathlib import Path
+
+import click
+from rich.console import Console
+from rich.panel import Panel
+from rich.table import Table
+from rich.text import Text
+from rich import box
+
+from .repo_probe import probe_repo
+from .generator import generate_work_order
+from .output import write_work_order
+
+console = Console()
+
+
+@click.group()
+@click.version_option("0.1.0", prog_name="sembl")
+def main():
+    """Sembl — turn messy repo intent into scoped AI Work Orders."""
+    pass
+
+
+# ── sembl generate ────────────────────────────────────────────────────────────
+
+@main.command()
+@click.option("--repo",     "-r", default=".", show_default=True,
+              help="Path to the repository.")
+@click.option("--task",     "-t", required=True,
+              help="The task or request to turn into a Work Order.")
+@click.option("--provider", "-p", default="openai",
+              type=click.Choice(["openai", "anthropic", "gemini", "nvidia"], case_sensitive=False),
+              show_default=True, help="LLM provider.")
+@click.option("--model",    "-m", default=None,
+              help="Model name. Defaults to gpt-4o (openai), claude-sonnet-4-6 (anthropic), gemini-2.5-flash (gemini), or mistralai/mistral-medium-3.5-128b (nvidia).")
+@click.option("--api-key",  default=None,
+              help="API key. Otherwise Sembl reads the selected provider env var.")
+@click.option("--no-graphify", is_flag=True, default=False,
+              help="Skip graphify even if available.")
+@click.option("--no-crg",      is_flag=True, default=False,
+              help="Skip code-review-graph even if available.")
+@click.option("--require-graph-context", is_flag=True, default=False,
+              help="Fail instead of using direct-probe fallback when graphify/CRG context is unavailable.")
+def generate(repo, task, provider, model, api_key, no_graphify, no_crg, require_graph_context):
+    """Generate a Work Order from a repo and a task description."""
+
+    repo_path = str(Path(repo).resolve())
+
+    # ── Step 1: Probe the repo ────────────────────────────────────────────
+    console.print()
+    console.print(Panel(
+        f"[bold]Task:[/bold] {task}",
+        title="[bold blue]Sembl — Work Order Generator[/bold blue]",
+        border_style="blue"
+    ))
+    console.print()
+
+    with console.status("[blue]Probing repo...[/blue]"):
+        probe = probe_repo(
+            repo_path,
+            task,
+            use_graphify=not no_graphify,
+            use_crg=not no_crg,
+        )
+
+    _print_probe_summary(probe, no_graphify, no_crg)
+    if require_graph_context and probe.context_basis != "graph_pipeline":
+        console.print(
+            "\n[red]Graph context required but unavailable.[/red]\n"
+            "Run graphify/code-review-graph for this repo, check PATH/venv resolution, "
+            "or rerun without [bold]--require-graph-context[/bold] to allow direct-probe fallback.\n"
+        )
+        sys.exit(1)
+
+    # ── Step 2: Generate Work Order ───────────────────────────────────────
+    _check_api_key(provider, api_key)
+
+    with console.status(f"[blue]Generating Work Order via {provider}...[/blue]"):
+        try:
+            wo = generate_work_order(
+                task=task,
+                probe=probe,
+                model_provider=provider,
+                model=model,
+                api_key=api_key,
+            )
+        except Exception as e:
+            console.print(_format_generation_error(e))
+            sys.exit(1)
+
+    # ── Step 3: Write output files ────────────────────────────────────────
+    with console.status("[blue]Writing output files...[/blue]"):
+        out_dir = write_work_order(wo, repo_path)
+
+    # ── Step 4: Print summary ─────────────────────────────────────────────
+    _print_work_order_summary(wo, out_dir)
+
+
+# ── sembl show ────────────────────────────────────────────────────────────────
+
+@main.command()
+@click.option("--repo", "-r", default=".", show_default=True,
+              help="Path to the repository.")
+@click.option("--id", "wo_id", default=None,
+              help="Work Order ID. Defaults to latest.")
+@click.option("--file", "output_file", default="work-order",
+              type=click.Choice(["work-order", "executor-prompt", "validation-plan"]),
+              show_default=True, help="Which file to show.")
+def show(repo, wo_id, output_file):
+    """Show a Work Order. Defaults to the latest one."""
+    repo_path = Path(repo).resolve()
+    wo_dir = _find_wo_dir(repo_path, wo_id)
+    if not wo_dir:
+        console.print("[red]No Work Orders found in this repo.[/red]")
+        console.print("Run [bold]sembl generate[/bold] to create one.")
+        sys.exit(1)
+
+    target = wo_dir / f"{output_file}.md"
+    if not target.exists():
+        console.print(f"[red]File not found:[/red] {target}")
+        sys.exit(1)
+
+    content = target.read_text(encoding="utf-8")
+    from rich.markdown import Markdown
+    console.print(Markdown(content))
+
+
+# ── sembl list ────────────────────────────────────────────────────────────────
+
+@main.command("list")
+@click.option("--repo", "-r", default=".", show_default=True,
+              help="Path to the repository.")
+def list_orders(repo):
+    """List all Work Orders in this repo."""
+    repo_path = Path(repo).resolve()
+    sembl_dir = repo_path / ".sembl" / "work-orders"
+
+    if not sembl_dir.exists():
+        console.print("[yellow]No Work Orders found.[/yellow] Run [bold]sembl generate[/bold].")
+        return
+
+    dirs = sorted(sembl_dir.iterdir(), key=lambda d: d.stat().st_mtime, reverse=True)
+    if not dirs:
+        console.print("[yellow]No Work Orders found.[/yellow]")
+        return
+
+    table = Table(box=box.ROUNDED, border_style="blue", show_header=True)
+    table.add_column("Work Order ID", style="bold cyan")
+    table.add_column("Task Type", style="green")
+    table.add_column("Risk", style="yellow")
+    table.add_column("Created", style="dim")
+    table.add_column("Goal", style="white")
+
+    for d in dirs:
+        json_file = d / "work-order.json"
+        if not json_file.exists():
+            continue
+        try:
+            data = json.loads(json_file.read_text(encoding="utf-8"))
+            risk = data.get("risk_level", "?").upper()
+            risk_style = {"LOW": "green", "MEDIUM": "yellow", "HIGH": "red"}.get(risk, "white")
+            table.add_row(
+                data.get("id", d.name),
+                data.get("task_type", "?"),
+                Text(risk, style=risk_style),
+                data.get("created_at", "?")[:19].replace("T", " "),
+                data.get("clarified_goal", "")[:60] + ("…" if len(data.get("clarified_goal","")) > 60 else ""),
+            )
+        except Exception:
+            table.add_row(d.name, "?", "?", "?", "?")
+
+    console.print()
+    console.print(table)
+    console.print()
+
+
+# ── Helpers ──────────────────────────────────────────────────────────────────
+
+def _print_probe_summary(probe, skip_graphify: bool, skip_crg: bool):
+    table = Table(box=box.SIMPLE, show_header=False, padding=(0, 1))
+    table.add_column("Key", style="dim")
+    table.add_column("Value", style="white")
+
+    table.add_row("Project", probe.project_name or "unknown")
+    table.add_row("Type", probe.project_type or "unknown")
+    table.add_row("Languages", ", ".join(probe.primary_languages) or "unknown")
+    table.add_row("Frameworks", ", ".join(probe.framework_hints) or "none detected")
+    table.add_row("Branch", probe.git_branch or "unknown")
+    table.add_row("Dirty", "yes" if probe.git_is_dirty else "no")
+    table.add_row("Context basis", "graph pipeline" if probe.context_basis == "graph_pipeline" else "direct fallback")
+    if probe.graph_context_sources:
+        table.add_row("Graph sources", ", ".join(probe.graph_context_sources))
+
+    gstatus = "available" if probe.graphify_available else ("skipped" if skip_graphify else "not found")
+    cstatus = "available" if probe.crg_available else ("skipped" if skip_crg else "not found")
+    table.add_row("Graphify", f"[green]{gstatus}[/green]" if probe.graphify_available else f"[dim]{gstatus}[/dim]")
+    table.add_row("code-review-graph", f"[green]{cstatus}[/green]" if probe.crg_available else f"[dim]{cstatus}[/dim]")
+
+    if probe.crg_available and probe.crg_node_count:
+        table.add_row("CRG graph", f"{probe.crg_node_count:,} nodes · {probe.crg_edge_count:,} edges")
+    if probe.crg_blast_radius:
+        table.add_row("CRG impact summaries", str(len(probe.crg_blast_radius)))
+
+    console.print(Panel(table, title="[bold]Repo probe[/bold]", border_style="dim"))
+    console.print()
+
+
+def _print_work_order_summary(wo, out_dir: Path):
+    risk_color = {"low": "green", "medium": "yellow", "high": "red"}.get(wo.risk_level, "white")
+
+    lines = [
+        f"[bold]{wo.id}[/bold]",
+        f"",
+        f"[bold]Goal:[/bold] {wo.clarified_goal}",
+        f"[bold]Outcome:[/bold] {wo.user_visible_outcome}",
+        f"[bold]Task type:[/bold] {wo.task_type}  |  "
+        f"[bold]Risk:[/bold] [{risk_color}]{wo.risk_level.upper()}[/{risk_color}]",
+        f"",
+        f"[bold]Acceptance criteria:[/bold] {len(wo.acceptance_criteria)} items",
+        f"[bold]Validation commands:[/bold] {len(wo.validation_commands)} commands",
+        f"[bold]Stop conditions:[/bold] {len(wo.stop_conditions)} triggers",
+        f"",
+        f"[dim]Output:[/dim] {out_dir}",
+        f"",
+        f"  work-order.md       — read this",
+        f"  executor-prompt.md  — paste into your agent",
+        f"  validation-plan.md  — run this after",
+        f"  work-order.json     — machine-readable",
+    ]
+
+    console.print(Panel(
+        "\n".join(lines),
+        title="[bold green]Work Order generated[/bold green]",
+        border_style="green"
+    ))
+    console.print()
+
+
+def _find_wo_dir(repo_path: Path, wo_id: str | None) -> Path | None:
+    sembl_dir = repo_path / ".sembl" / "work-orders"
+    if not sembl_dir.exists():
+        return None
+    if wo_id:
+        target = sembl_dir / wo_id
+        return target if target.exists() else None
+    # Latest
+    dirs = sorted(sembl_dir.iterdir(), key=lambda d: d.stat().st_mtime, reverse=True)
+    return dirs[0] if dirs else None
+
+
+def _check_api_key(provider: str, api_key: str | None):
+    env_keys = {
+        "openai": "OPENAI_API_KEY",
+        "anthropic": "ANTHROPIC_API_KEY",
+        "gemini": "GEMINI_API_KEY",
+        "nvidia": "NVIDIA_API_KEY",
+    }
+    env_key = env_keys[provider.lower()]
+    if not api_key and not os.environ.get(env_key):
+        console.print(f"\n[red]No API key found.[/red]")
+        console.print(f"Set [bold]{env_key}[/bold] or pass [bold]--api-key[/bold].\n")
+        sys.exit(1)
+
+
+def _format_generation_error(error: Exception) -> str:
+    message = str(error)
+    provider = getattr(error, "provider", None)
+    code = getattr(error, "code", None)
+    status_code = getattr(error, "status_code", None)
+    lower_message = message.lower()
+
+    if provider == "gemini":
+        return _format_gemini_error(message, status_code, code)
+    if provider == "nvidia":
+        return _format_nvidia_error(message, status_code, code)
+
+    if code == "insufficient_quota" or "insufficient_quota" in lower_message:
+        return (
+            "\n[red]Generation failed:[/red] OpenAI quota is exhausted for this API project/org.\n\n"
+            "The request reached OpenAI, but the API account has no available quota or has hit a spend limit.\n"
+            "Check billing: https://platform.openai.com/settings/organization/billing\n"
+            "Check limits:  https://platform.openai.com/settings/organization/limits\n\n"
+            "ChatGPT subscriptions and OpenAI API billing are separate. After adding credits or raising the limit, "
+            "rerun the same Sembl command.\n"
+        )
+
+    if status_code == 401 or code == "invalid_api_key" or "invalid_api_key" in lower_message:
+        return (
+            "\n[red]Generation failed:[/red] OpenAI rejected the API key.\n\n"
+            "Set a valid OPENAI_API_KEY for the selected project, then rerun the same Sembl command.\n"
+        )
+
+    if "rate_limit" in lower_message or code == "rate_limit_exceeded":
+        return (
+            "\n[red]Generation failed:[/red] OpenAI rate limit reached.\n\n"
+            "Wait briefly, reduce concurrency, or use a lower-throughput model/project before retrying.\n"
+        )
+
+    if status_code == 403 or "permission" in lower_message or "model_not_found" in lower_message:
+        return (
+            "\n[red]Generation failed:[/red] The API key does not have access to the requested model or project.\n\n"
+            "Check the model name, project, organization, and key scope, then rerun the command.\n"
+        )
+
+    return f"\n[red]Generation failed:[/red] {message}"
+
+
+def _format_gemini_error(message: str, status_code: int | None, code: str | None) -> str:
+    code_text = str(code or "").upper()
+    lower_message = message.lower()
+
+    if status_code == 400 or code_text == "INVALID_ARGUMENT":
+        return (
+            "\n[red]Generation failed:[/red] Gemini rejected the request.\n\n"
+            f"{message}\n\n"
+            "Check the model name and request shape, then rerun the same Sembl command.\n"
+        )
+
+    if status_code == 401 or status_code == 403 or code_text in {"UNAUTHENTICATED", "PERMISSION_DENIED"}:
+        return (
+            "\n[red]Generation failed:[/red] Gemini rejected the API key or project access.\n\n"
+            "Set a valid GEMINI_API_KEY with access to the selected model, then rerun the same Sembl command.\n"
+        )
+
+    if status_code == 429 or code_text == "RESOURCE_EXHAUSTED" or "quota" in lower_message:
+        return (
+            "\n[red]Generation failed:[/red] Gemini quota or rate limit reached.\n\n"
+            "Wait briefly, lower the request volume, or check the Google AI Studio / Google Cloud quota for this key.\n"
+        )
+
+    if status_code == 404 or "not found" in lower_message:
+        return (
+            "\n[red]Generation failed:[/red] Gemini model was not found or is unavailable for this API key.\n\n"
+            "Try --model gemini-2.5-flash, then rerun the same Sembl command.\n"
+        )
+
+    return f"\n[red]Generation failed:[/red] Gemini API error: {message}"
+
+
+def _format_nvidia_error(message: str, status_code: int | None, code: str | None) -> str:
+    lower_message = message.lower()
+
+    if status_code == 401 or status_code == 403 or "unauthorized" in lower_message:
+        return (
+            "\n[red]Generation failed:[/red] NVIDIA rejected the API key or model access.\n\n"
+            "Set a valid NVIDIA_API_KEY with access to the selected model, then rerun the same Sembl command.\n"
+        )
+
+    if status_code == 429 or "rate" in lower_message or "quota" in lower_message:
+        return (
+            "\n[red]Generation failed:[/red] NVIDIA quota or rate limit reached.\n\n"
+            "Wait briefly, choose another available NVIDIA model, or check the key limits in the NVIDIA API catalog.\n"
+        )
+
+    if status_code == 404 or "model" in lower_message and "not found" in lower_message:
+        return (
+            "\n[red]Generation failed:[/red] NVIDIA model was not found or is unavailable for this key.\n\n"
+            "Try --model mistralai/mistral-medium-3.5-128b or another model shown in the NVIDIA catalog.\n"
+        )
+
+    if "invalid json" in lower_message or "llm returned invalid json" in lower_message:
+        return (
+            "\n[red]Generation failed:[/red] NVIDIA returned text that was not valid Work Order JSON.\n\n"
+            "Retry once, or try a stronger instruction-following model from the NVIDIA catalog.\n"
+        )
+
+    return f"\n[red]Generation failed:[/red] NVIDIA API error: {message}"