agent-trajectory-diff 0.1.0__py3-none-any.whl

agentdiff/reporters/terminal.py

@@ -0,0 +1,107 @@
+from rich.console import Console
+from rich.panel import Panel
+from rich.table import Table
+from rich.text import Text
+
+from agentdiff.models.report import DiffReport, StepDiffStatus
+from agentdiff.models.step import StepStatus
+
+
+def render_diff_table(report: DiffReport) -> Table:
+    """Renders the step-by-step diff as a Rich Table."""
+    table = Table(
+        title="Trajectory Comparison Details",
+        show_header=True,
+        header_style="bold magenta",
+    )
+    table.add_column("Index", style="dim", width=6)
+    table.add_column("Step Name", width=25)
+    table.add_column("Status", width=12)
+    table.add_column("Baseline (Lat/Tkn/Cost)", width=28)
+    table.add_column("Candidate (Lat/Tkn/Cost)", width=28)
+
+    for idx, sd in enumerate(report.step_diffs):
+        status_text = Text(sd.diff_status.value.upper())
+        row_style = ""
+
+        if sd.diff_status == StepDiffStatus.MATCHED:
+            status_text.stylize("dim green")
+            row_style = "dim"
+        elif sd.diff_status == StepDiffStatus.ADDED:
+            status_text.stylize("bold cyan")
+            row_style = "bold cyan"
+        elif sd.diff_status == StepDiffStatus.REMOVED:
+            status_text.stylize("bold red")
+            row_style = "bold red"
+        elif sd.diff_status == StepDiffStatus.MODIFIED:
+            status_text.stylize("bold yellow")
+            row_style = "bold yellow"
+
+        base_info = "-"
+        if sd.baseline_step:
+            s = sd.baseline_step
+            base_info = f"{s.latency_ms:.0f}ms / {s.tokens.total_tokens}t / ${s.tokens.estimated_cost_usd:.4f}"
+            if s.status != StepStatus.SUCCESS:
+                base_info += f" ({s.status.value.upper()})"
+
+        cand_info = "-"
+        if sd.candidate_step:
+            s = sd.candidate_step
+            cand_info = f"{s.latency_ms:.0f}ms / {s.tokens.total_tokens}t / ${s.tokens.estimated_cost_usd:.4f}"
+            if s.status != StepStatus.SUCCESS:
+                cand_info += f" ({s.status.value.upper()})"
+
+        table.add_row(
+            str(idx + 1),
+            sd.step_name,
+            status_text,
+            base_info,
+            cand_info,
+            style=row_style,
+        )
+
+    return table
+
+
+def print_report(report: DiffReport):
+    """Outputs the complete DiffReport to the terminal."""
+    console = Console()
+
+    status_str = (
+        "[bold green]PASSED[/bold green]"
+        if report.passed
+        else "[bold red]FAILED[/bold red]"
+    )
+    summary_text = (
+        f"[bold]Baseline Trace:[/bold]  {report.baseline_id}\n"
+        f"[bold]Candidate Trace:[/bold] {report.candidate_id}\n"
+        f"[bold]Comparison Status:[/bold] {status_str}\n\n"
+        f"[bold]Trajectory Divergence Index (TDI):[/bold] {report.trajectory_divergence_index:.4f}\n"
+        f"[bold]Baseline Wasted Effort (WEI):[/bold]  {report.baseline_wei:.4f}\n"
+        f"[bold]Candidate Wasted Effort (WEI):[/bold] {report.candidate_wei:.4f}\n\n"
+        f"[bold]Resource Deltas:[/bold]\n"
+        f"  • Latency Delta: {report.latency_delta_percentage:+.2f}%\n"
+        f"  • Token Delta:   {report.token_delta_percentage:+.2f}%\n"
+        f"  • Cost Delta:    {report.cost_delta_percentage:+.2f}%"
+    )
+
+    panel = Panel(
+        summary_text,
+        title="[bold cyan]AgentDiff Comparison Summary[/bold cyan]",
+        border_style="cyan",
+    )
+    console.print(panel)
+
+    if report.loops_detected:
+        loop_text = ""
+        for idx, loop in enumerate(report.loops_detected):
+            stagnant_str = " (Stagnant state changes)" if loop.get("stagnant") else ""
+            loop_text += f"[bold red]Loop #{idx + 1}:[/bold red] Repeated {loop['steps']} {loop['iterations']} times{stagnant_str}\n"
+        loop_panel = Panel(
+            loop_text.strip(),
+            title="[bold red]Warnings: Loops Detected[/bold red]",
+            border_style="red",
+        )
+        console.print(loop_panel)
+
+    console.print(render_diff_table(report))

agentdiff/testing/__init__.py

@@ -0,0 +1,5 @@
+from agentdiff.testing.assertions import assert_no_regressions
+
+__all__ = [
+    "assert_no_regressions",
+]

agentdiff/testing/assertions.py

@@ -0,0 +1,50 @@
+from agentdiff.models.report import DiffReport
+
+
+def assert_no_regressions(
+    report: DiffReport,
+    max_divergence: float = 0.25,
+    max_cost_increase_pct: float = 5.0,
+    allow_loops: bool = False,
+    max_wasted_effort: float = 0.10,
+):
+    """Expressive regression assertion helper for pytest suites.
+
+    Raises an AssertionError with descriptive failure outputs if any threshold is violated.
+    """
+    errors = []
+
+    # 1. Divergence Check
+    if report.trajectory_divergence_index > max_divergence:
+        errors.append(
+            f"Trajectory Divergence Index (TDI) of {report.trajectory_divergence_index:.4f} "
+            f"exceeded threshold of {max_divergence:.4f}."
+        )
+
+    # 2. Cost Check
+    if report.cost_delta_percentage > max_cost_increase_pct:
+        errors.append(
+            f"Cost increase of {report.cost_delta_percentage:+.2f}% "
+            f"exceeded threshold of {max_cost_increase_pct:+.2f}%."
+        )
+
+    # 3. Loops Check
+    if not allow_loops and report.loops_detected:
+        errors.append(
+            f"Detected {len(report.loops_detected)} loops in the candidate run, "
+            f"but allow_loops is False."
+        )
+
+    # 4. Wasted Effort Check
+    if report.candidate_wei > max_wasted_effort:
+        errors.append(
+            f"Candidate Wasted Effort Index (WEI) of {report.candidate_wei:.4f} "
+            f"exceeded threshold of {max_wasted_effort:.4f}."
+        )
+
+    if errors:
+        errors_str = "\n".join(f"  - {err}" for err in errors)
+        summary_str = report.summary()
+        raise AssertionError(
+            f"AgentDiff Regression Verification Failed:\n{errors_str}\n\n{summary_str}"
+        )