rakam-eval-sdk 0.2.2__py3-none-any.whl → 0.2.4__py3-none-any.whl

rakam_eval_sdk/cli.py

@@ -6,16 +6,18 @@ import uuid
 from datetime import datetime
 from pathlib import Path
 from pprint import pprint
-from typing import Any, Dict, Optional, Sequence
+from typing import Any, Dict, Optional
 
 import typer
 from dotenv import load_dotenv
 from rich.console import Console
 from rich.panel import Panel
 from rich.pretty import Pretty
+from typer import secho
 
 from rakam_eval_sdk.client import DeepEvalClient
 from rakam_eval_sdk.decorators import eval_run
+from rakam_eval_sdk.schema import MetricDiff, TestCaseComparison
 from rakam_eval_sdk.utils.decorator_utils import (
     find_decorated_functions,
     load_module_from_path,
@@ -35,55 +37,88 @@ metrics_app = typer.Typer(help="Metrics utilities")
 app.add_typer(metrics_app, name="metrics")
 
 
+def extract_metric_names(config: Any) -> list[tuple[str, str | None]]:
+    """
+    Returns [(type, name)] from EvalConfig / SchemaEvalConfig
+    """
+    if not hasattr(config, "metrics"):
+        return []
+
+    results: list[tuple[str, str | None]] = []
+
+    for metric in config.metrics or []:
+        metric_type = getattr(metric, "type", None)
+        metric_name = getattr(metric, "name", None)
+        if metric_type:
+            results.append((metric_type, metric_name))
+
+    return results
+
+
 @metrics_app.command("list")
-def list_metrics(
-    limit: int = typer.Option(
-        20,
-        "--limit",
-        help="Number of testcases to inspect for metrics",
+def metrics(
+    directory: Path = typer.Argument(
+        Path("./eval"),
+        exists=True,
+        file_okay=False,
+        dir_okay=True,
+        help="Directory to scan (default: ./eval)",
     ),
-):
+    recursive: bool = typer.Option(
+        False,
+        "-r",
+        "--recursive",
+        help="Recursively search for Python files",
+    ),
+) -> None:
     """
-    List unique metric names found in evaluation testcases.
+    List all metric types used by loaded eval configs.
     """
-    client = DeepEvalClient()
+    files = directory.rglob("*.py") if recursive else directory.glob("*.py")
+    TARGET_DECORATOR = eval_run.__name__
 
-    testcases = client.list_evaluation_testcases(
-        limit=limit,
-        offset=0,
-        raise_exception=True,
-    )
+    all_metrics: set[tuple[str, str | None]] = set()
+    found_any = False
 
-    if not testcases:
-        typer.echo("No evaluation testcases found.")
-        return
+    for file in sorted(files):
+        functions = find_decorated_functions(file, TARGET_DECORATOR)
+        if not functions:
+            continue
 
-    metric_names: set[str] = set()
+        typer.echo(f"\n📄 {file}")
 
-    def collect_metrics(entries: Sequence[Dict] | None):
-        if not entries:
-            return
-        for entry in entries:
-            for metric in entry.get("metrics", []) or []:
-                name = metric.get("name")
-                if name:
-                    metric_names.add(name)
+        try:
+            module = load_module_from_path(file)
+        except Exception as e:
+            typer.echo(f"  ❌ Failed to import module: {e}")
+            continue
 
-    for tc in testcases:
+        for fn_name in functions:
+            try:
+                func = getattr(module, fn_name)
+                result = func()
 
-        collect_metrics(tc.get("result"))
+                metrics = extract_metric_names(result)
+                if not metrics:
+                    continue
 
-    if not metric_names:
-        typer.echo("No metrics found.")
-        return
+                found_any = True
+                for metric_type, metric_name in metrics:
+                    all_metrics.add((metric_type, metric_name))
 
-    typer.echo(
-        f"📊 Found {len(metric_names)} unique metrics "
-        f"(from latest {limit} testcases)\n"
-    )
+                    if metric_name:
+                        typer.echo(f"  • {metric_type} (alias: {metric_name})")
+                    else:
+                        typer.echo(f"  • {metric_type}")
+
+            except Exception as e:
+                typer.echo(f"  ❌ Failed to inspect {fn_name}: {e}")
+
+    if not found_any:
+        typer.echo("\nNo metrics found.")
+        raise typer.Exit(code=0)
 
-    for name in sorted(metric_names):
-        typer.echo(f"- {name}")
+    typer.echo(f"\n✅ {len(all_metrics)} unique metrics found")
 
 
 @list_app.command("evals")
@@ -130,19 +165,22 @@ def list_runs(
     """
     client = DeepEvalClient()
 
-    runs = client.list_evaluation_testcases(
+    response = client.list_evaluation_testcases(
         limit=limit,
         offset=offset,
         raise_exception=True,
     )
 
-    if not runs:
+    items = response.get("items", [])
+    total = response.get("total", 0)
+
+    if not items:
         typer.echo("No evaluation runs found.")
         return
 
-    typer.echo(f"[id] " f"{'tag':<20}" f"{'label':<20}" f"created_at")
-    # pretty CLI output
-    for run in runs:
+    typer.echo(f"[id] {'tag':<20}{'label':<20}created_at")
+
+    for run in items:
         run_id = run.get("id")
         label = run.get("label") or "-"
         uid = run.get("tag") or "-"
@@ -156,8 +194,12 @@ def list_runs(
             except ValueError:
                 pass
 
-        typer.echo(
-            f"[{run_id}] " f"{uid:<20} " f"{label:<20} " f"{created_at}")
+        typer.echo(f"[{run_id}] {uid:<20} {label:<20} {created_at}")
+
+    shown = offset + len(items)
+    if shown < total:
+        typer.echo()
+        typer.echo(f"Showing {shown} of {total} runs. Use --limit to see more.")
 
 
 @app.command()
@@ -165,12 +207,12 @@ def show(
     id: Optional[int] = typer.Option(
         None,
         "--id",
-        help="Numeric evaluation testcase ID",
+        help="Run ID",
     ),
     tag: Optional[str] = typer.Option(
         None,
         "--tag",
-        help="Evaluation testcase tag",
+        help="Run tag",
     ),
     raw: bool = typer.Option(
         False,
@@ -377,8 +419,7 @@ def _print_and_save(
         return
 
     if out.exists() and not overwrite:
-        typer.echo(
-            f"❌ File already exists: {out} (use --overwrite to replace)")
+        typer.echo(f"❌ File already exists: {out} (use --overwrite to replace)")
         raise typer.Exit(code=1)
 
     out.parent.mkdir(parents=True, exist_ok=True)
@@ -389,20 +430,124 @@ def _print_and_save(
     typer.echo(f"💾 Result saved to {out}")
 
 
+def pct_change(a: float | None, b: float | None) -> str | None:
+    if a is None or b is None or a == 0:
+        return None
+    return f"{((b - a) / a) * 100:+.2f}%"
+
+
+def metric_direction(delta: float | None) -> str:
+    if delta is None:
+        return "unchanged"
+    if delta > 0:
+        return "improved"
+    if delta < 0:
+        return "regressed"
+    return "unchanged"
+
+
+def print_metric_diff(diff: MetricDiff):
+    secho(f"\nMetric: {diff.metric}", bold=True)
+
+    if diff.status == "added":
+        secho(f"+ score: {diff.score_b}", fg="green")
+        secho(f"+ threshold: {diff.threshold_b}", fg="green")
+        secho(f"+ success: {diff.success_b}", fg="green")
+        return
+
+    if diff.status == "removed":
+        secho(f"- score: {diff.score_a}", fg="red")
+        secho(f"- threshold: {diff.threshold_a}", fg="red")
+        secho(f"- success: {diff.success_a}", fg="red")
+        return
+
+    # unchanged / changed
+    if diff.score_a != diff.score_b:
+        direction = metric_direction(diff.delta)
+        color = "green" if direction == "improved" else "red"
+        pct = pct_change(diff.score_a, diff.score_b)
+
+        secho(f"- score: {diff.score_a}", fg="red")
+        secho(
+            f"+ score: {diff.score_b}" + (f"   ({pct})" if pct else ""),
+            fg=color,
+        )
+    else:
+        secho(f"  score: {diff.score_a}", dim=True)
+
+    if diff.threshold_a != diff.threshold_b:
+        secho(f"- threshold: {diff.threshold_a}", fg="red")
+        secho(f"+ threshold: {diff.threshold_b}", fg="green")
+    else:
+        secho(f"  threshold: {diff.threshold_a}", dim=True)
+
+    if diff.success_a != diff.success_b:
+        secho(f"- success: {diff.success_a}", fg="red")
+        secho(f"+ success: {diff.success_b}", fg="green")
+    else:
+        secho(f"  success: {diff.success_a}", dim=True)
+
+
+def summarize(metrics: Any) -> Dict[str, int]:
+    summary = {
+        "improved": 0,
+        "regressed": 0,
+        "unchanged": 0,
+        "added": 0,
+        "removed": 0,
+    }
+
+    for m in metrics:
+        if m.status in ("added", "removed"):
+            summary[m.status] += 1
+        else:
+            direction = metric_direction(m.delta)
+            summary[direction] += 1
+
+    return summary
+
+
+def pretty_print_comparison(resp: Any, summary_only: bool = False):
+    if not summary_only:
+        for metric in resp.metrics:
+            print_metric_diff(metric)
+        return
+
+    print_summary(resp.metrics)
+
+
+def print_summary(metrics: Any):
+    summary = summarize(metrics)
+
+    secho("\nSummary:", bold=True)
+    secho(f"  ↑ Improved: {summary['improved']}", fg="green")
+    secho(f"  ↓ Regressed: {summary['regressed']}", fg="red")
+    secho(f"  ± Unchanged: {summary['unchanged']}", dim=True)
+    secho(f"  + Added: {summary['added']}", fg="green")
+    secho(f"  - Removed: {summary['removed']}", fg="red")
+
+
 @app.command()
 def compare(
-    testcase_a_id: int = typer.Argument(
-        ...,
-        help="ID of the first testcase",
+    tag: list[str] = typer.Option(
+        [],
+        "--tag",
+        help="Label identifying a reference testcase",
     ),
-    testcase_b_id: int = typer.Argument(
-        ...,
-        help="ID of the second testcase",
+    run: list[int] = typer.Option(
+        [],
+        "--run",
+        help="Run ID identifying an evaluation testcase",
     ),
     pretty: bool = typer.Option(
         True,
         "--pretty/--raw",
-        help="Pretty-print the response",
+        help="Pretty diff output (default) or raw JSON",
+    ),
+    summary: bool = typer.Option(
+        False,
+        "--summary",
+        help="Show summary only (no per-metric diff)",
     ),
     raise_exception: bool = typer.Option(
         False,
@@ -422,29 +567,85 @@ def compare(
     ),
 ) -> None:
     """
-    Compare two DeepEval evaluation testcases.
+    Compare two evaluation testcases (runs and/or labels).
+
+    Output:
+      - Unified diff-style view by default
+      - Raw JSON with --raw
+      - Summary of improvements / regressions
     """
+
+    targets: list[tuple[str, str | int]] = []
+
+    for r in run:
+        targets.append(("run", r))
+
+    for t in tag:
+        targets.append(("label", t))
+
+    if len(targets) != 2:
+        typer.secho(
+            "❌ Provide exactly two targets using --run and/or --tag",
+            fg="red",
+            bold=True,
+        )
+        raise typer.Exit(code=1)
+
+    (type_a, value_a), (type_b, value_b) = targets
+
+    typer.secho(
+        f"🔍 Comparing {type_a} '{value_a}' ↔ {type_b} '{value_b}'",
+        bold=True,
+    )
+
     client = DeepEvalClient()
 
-    typer.echo(f"🔍 Comparing testcases {testcase_a_id} ↔ {testcase_b_id}")
+    kwargs = {"raise_exception": raise_exception}
+
+    if type_a == "run":
+        kwargs["testcase_a_id"] = value_a
+    else:
+        kwargs["testcase_a_tag"] = value_a
+
+    if type_b == "run":
+        kwargs["testcase_b_id"] = value_b
+    else:
+        kwargs["testcase_b_tag"] = value_b
 
     try:
-        resp = client.compare_testcases(
-            testcase_a_id=testcase_a_id,
-            testcase_b_id=testcase_b_id,
-            raise_exception=raise_exception,
-        )
+        resp = client.compare_testcases(**kwargs)
     except Exception as e:
-        typer.echo(f"❌ Request failed: {e}")
+        typer.secho(f"❌ Request failed: {e}", fg="red")
         raise typer.Exit(code=1)
 
     if not resp:
-        typer.echo("⚠️ No response received")
+        typer.secho("⚠️ No response received", fg="yellow")
         raise typer.Exit(code=1)
-    _print_and_save(resp, pretty, out, overwrite)
 
+    if out:
+        if out.exists() and not overwrite:
+            typer.secho(
+                f"❌ File already exists: {out} (use --overwrite)",
+                fg="red",
+            )
+            raise typer.Exit(code=1)
 
-@app.command()
+        out.write_text(json.dumps(resp, indent=2))
+        typer.secho(f"💾 Saved raw output to {out}", fg="green")
+
+    if not pretty:
+        typer.echo(json.dumps(resp, indent=2))
+        return
+
+    comparison = TestCaseComparison(**resp)
+
+    pretty_print_comparison(
+        comparison,
+        summary_only=summary,
+    )
+
+
+@app.command(hidden=True)
 def compare_label_latest(
     label_a: str = typer.Argument(
         ...,
@@ -500,7 +701,7 @@ def compare_label_latest(
     _print_and_save(resp, pretty, out, overwrite)
 
 
-@app.command()
+@app.command(hidden=True)
 def compare_last(
     label: str = typer.Argument(
         ...,
@@ -551,44 +752,62 @@ def compare_last(
     _print_and_save(resp, pretty, out, overwrite)
 
 
-@list_app.command("tag")
-def update_run_tag(
-    run_id: int = typer.Argument(..., help="Evaluation run ID"),
+@app.command("tag")
+def tag_command(
+    run_id: Optional[int] = typer.Option(
+        None,
+        "--run",
+        help="Evaluation run ID",
+    ),
     tag: Optional[str] = typer.Option(
         None,
         "--tag",
-        "-t",
-        help="Tag to add or update",
+        help="Tag to assign to the run",
     ),
-    remove: bool = typer.Option(
-        False,
-        "--remove",
-        help="Remove tag from the run",
+    delete: Optional[str] = typer.Option(
+        None,
+        "--delete",
+        help="Delete a tag",
     ),
 ):
     """
-    Add, update, or remove a tag from an evaluation run.
+    Assign a tag to a run or delete a tag.
     """
-    if not tag and not remove:
-        typer.echo("❌ You must provide --tag or --remove")
-        raise typer.Exit(code=1)
 
-    if tag and remove:
-        typer.echo("❌ Use either --tag or --remove, not both")
-        raise typer.Exit(code=1)
+    # --- validation ---
+    if delete:
+        if run_id or tag:
+            typer.echo("❌ --delete cannot be used with --run or --tag")
+            raise typer.Exit(code=1)
+    else:
+        if not run_id or not tag:
+            typer.echo("❌ Use --run and --tag together, or --delete")
+            raise typer.Exit(code=1)
 
     client = DeepEvalClient()
 
+    # --- delete mode ---
+    if delete:
+        result = client.update_evaluation_testcase_tag(
+            testcase_id=run_id,
+            tag=delete,
+            raise_exception=True,
+        )
+        typer.echo("🗑️ Tag deleted successfully")
+        typer.echo(f"Tag: {delete}")
+        return
+
+    # --- assign/update mode ---
     result = client.update_evaluation_testcase_tag(
         testcase_id=run_id,
-        tag=None if remove else tag,
+        tag=tag,
         raise_exception=True,
     )
 
-    action = "removed" if remove else "updated"
-    typer.echo(f"✅ Tag {action} successfully")
+    typer.echo("✅ Tag assigned successfully")
     typer.echo(f"Run ID: {run_id}")
-    typer.echo(f"Tag: {result.get('tag') or '-'}")
+    typer.echo(f"Tag: {result.get('tag')}")
+
 
 def main() -> None:
     app()

rakam_eval_sdk/client.py

@@ -1,7 +1,6 @@
-from typing import Optional, Dict
 import os
 import random
-from typing import Any, Dict, List, Optional, Union, cast, overload
+from typing import Any, Dict, List, Literal, Optional, Union, cast, overload
 
 import requests
 
@@ -13,7 +12,6 @@ from .schema import (
     SchemaMetricConfig,
     TextInputItem,
 )
-from typing import Optional, Literal, cast
 
 HTTPMethod = Literal["GET", "POST", "PATCH", "PUT", "DELETE"]
 
@@ -42,8 +40,7 @@ class DeepEvalClient:
         )
         self.base_url = raw_url.rstrip("/")
         self.api_token = (
-            api_token or settings_token or os.getenv(
-                "EVALFRAMEWORK_API_KEY", "")
+            api_token or settings_token or os.getenv("EVALFRAMEWORK_API_KEY", "")
         )
         self.timeout = timeout
 
@@ -52,10 +49,10 @@ class DeepEvalClient:
         method: HTTPMethod,
         endpoint: str,
         *,
-        json: dict | None = None,
-        params: dict | None = None,
+        json: Dict | None = None,
+        params: Dict | None = None,
         raise_exception: bool = False,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         url = f"{self.base_url}{endpoint}"
 
         headers = {
@@ -92,14 +89,17 @@ class DeepEvalClient:
                 "raw": resp.text,
             }
 
-    def _get(self, endpoint: str, params: dict, **kw):
-        return self._request("GET", endpoint, params=params, **kw)
+    def _get(self, endpoint: str, params: Dict, *args, **kw):
+        return self._request("GET", endpoint, params=params, *args, **kw)
+
+    def _post(self, endpoint: str, payload: Dict, *args, **kw):
+        return self._request("POST", endpoint, json=payload, *args, **kw)
 
-    def _post(self, endpoint: str, payload: dict, **kw):
-        return self._request("POST", endpoint, json=payload, **kw)
+    def _patch(self, endpoint: str, payload: Dict, *args, **kw):
+        return self._request("PATCH", endpoint, json=payload, *args, **kw)
 
-    def _patch(self, endpoint: str, payload: dict, **kw):
-        return self._request("PATCH", endpoint, json=payload, **kw)
+    def _delete(self, endpoint: str, payload: Dict, *args, **kw):
+        return self._request("DELETE", endpoint, json=payload, *args, **kw)
 
     def update_evaluation_testcase_tag(
         self,
@@ -114,8 +114,14 @@ class DeepEvalClient:
         - tag="smoke" → add / update tag
         - tag=None → remove tag
         """
+        if testcase_id is None:
+            return self._delete(
+                endpoint=f"/deepeval/tag/{tag}",
+                payload={},
+                raise_exception=raise_exception,
+            )
         return self._patch(
-            f"/evaluation-testcases/{testcase_id}/tag",
+            endpoint=f"/deepeval/{testcase_id}/tag",
             payload={"tag": tag},
             raise_exception=raise_exception,
         )
@@ -126,13 +132,14 @@ class DeepEvalClient:
         limit: int = 10,
         offset: int = 0,
         raise_exception: bool = False,
-    ) -> Optional[List[Dict]]:
+    ) -> Optional[Dict]:
         """
         List evaluation testcases for the current API token only.
         Sorted by created_at DESC (newest first).
+        Returns items + pagination metadata.
         """
         return self._get(
-            "/eval-framework/deepeval/evaluation-testcases/token",
+            endpoint="/eval-framework/deepeval/evaluation-testcases/token",
             params={
                 "limit": limit,
                 "offset": offset,
@@ -150,7 +157,7 @@ class DeepEvalClient:
         Fetch a single evaluation testcase by numeric ID.
         """
         return self._get(
-            f"/eval-framework/deepeval/id/{testcase_id}",
+            endpoint=f"/eval-framework/deepeval/id/{testcase_id}",
             params={},
             raise_exception=raise_exception,
         )
@@ -165,7 +172,7 @@ class DeepEvalClient:
         Fetch a single evaluation testcase by tag.
         """
         return self._get(
-            f"/eval-framework/deepeval/tag/{tag}",
+            endpoint=f"/eval-framework/deepeval/tag/{tag}",
             params={},
             raise_exception=raise_exception,
         )
@@ -190,19 +197,39 @@ class DeepEvalClient:
     def compare_testcases(
         self,
         *,
-        testcase_a_id: int,
-        testcase_b_id: int,
+        testcase_a_id: int | None = None,
+        testcase_a_tag: str | None = None,
+        testcase_b_id: int | None = None,
+        testcase_b_tag: str | None = None,
         raise_exception: bool = False,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         """
-        Compare two evaluation testcases.
+        Compare two evaluation testcases using IDs or tags.
+        Exactly one identifier (id or tag) must be provided per testcase.
         """
+
+        def validate(id_, tag, name: str):
+            if bool(id_) == bool(tag):
+                raise ValueError(f"Provide exactly one of {name}_id or {name}_tag")
+
+        validate(testcase_a_id, testcase_a_tag, "testcase_a")
+        validate(testcase_b_id, testcase_b_tag, "testcase_b")
+
+        params: dict[str, int | str] = {}
+
+        if testcase_a_id is not None:
+            params["testcase_a_id"] = testcase_a_id
+        else:
+            params["testcase_a_tag"] = testcase_a_tag  # type: ignore
+
+        if testcase_b_id is not None:
+            params["testcase_b_id"] = testcase_b_id
+        else:
+            params["testcase_b_tag"] = testcase_b_tag  # type: ignore
+
         return self._get(
-            "/eval-framework/deepeval/evaluation-testcases/compare",
-            params={
-                "testcase_a_id": testcase_a_id,
-                "testcase_b_id": testcase_b_id,
-            },
+            endpoint="/eval-framework/deepeval/evaluation-testcases/compare",
+            params=params,
             raise_exception=raise_exception,
         )
 
@@ -212,12 +239,12 @@ class DeepEvalClient:
         label_a: str,
         label_b: str,
         raise_exception: bool = False,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         """
         Compare the latest evaluation testcases for two labels.
         """
         return self._get(
-            "/eval-framework/deepeval/evaluation-testcases/compare-latest",
+            endpoint="/eval-framework/deepeval/evaluation-testcases/compare-latest",
             params={
                 "label_a": label_a,
                 "label_b": label_b,
@@ -230,12 +257,12 @@ class DeepEvalClient:
         *,
         label: str,
         raise_exception: bool = False,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         """
         Compare the last two evaluation testcases for a given label.
         """
         return self._get(
-            "/eval-framework/deepeval/evaluation-testcases/compare-last-two",
+            endpoint="/eval-framework/deepeval/evaluation-testcases/compare-last-two",
             params={
                 "label": label,
             },
@@ -248,7 +275,7 @@ class DeepEvalClient:
         config: EvalConfig,
         *,
         raise_exception: bool = False,
-    ) -> Optional[dict]: ...
+    ) -> Optional[Dict]: ...
 
     @overload
     def text_eval(
@@ -259,7 +286,7 @@ class DeepEvalClient:
         component: str = "unknown",
         label: str | None = None,
         raise_exception: bool = False,
-    ) -> Optional[dict]: ...
+    ) -> Optional[Dict]: ...
 
     def text_eval(
         self,
@@ -270,7 +297,7 @@ class DeepEvalClient:
         component: str = "unknown",
         label: str | None = None,
         raise_exception: bool = False,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         if config is None:
             config = EvalConfig(
                 data=data,
@@ -280,7 +307,9 @@ class DeepEvalClient:
             )
 
         return self._post(
-            "/deepeval/text-eval", config.model_dump(), raise_exception
+            endpoint="/deepeval/text-eval",
+            payload=config.model_dump(),
+            raise_exception=raise_exception,
         )
 
     def text_eval_background(
@@ -290,12 +319,16 @@ class DeepEvalClient:
         raise_exception: bool = False,
         component: str = "unknown",
         label: Union[str, None] = None,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         """Run background text evaluation (async job)."""
         payload = EvalConfig.model_construct(
             data=data, metrics=metrics, component=component, version=label
         ).model_dump()
-        return self._post("/deepeval/text-eval/background", payload, raise_exception)
+        return self._post(
+            endpoint="/deepeval/text-eval/background",
+            payload=payload,
+            raise_exception=raise_exception,
+        )
 
     @overload
     def schema_eval(
@@ -306,7 +339,7 @@ class DeepEvalClient:
         component: str = "unknown",
         label: str | None = None,
         raise_exception: bool = False,
-    ) -> Optional[dict]: ...
+    ) -> Optional[Dict]: ...
 
     @overload
     def schema_eval(
@@ -314,7 +347,7 @@ class DeepEvalClient:
         config: SchemaEvalConfig,
         *,
         raise_exception: bool = False,
-    ) -> Optional[dict]: ...
+    ) -> Optional[Dict]: ...
 
     def schema_eval(
         self,
@@ -325,7 +358,7 @@ class DeepEvalClient:
         component: str = "unknown",
         label: str | None = None,
         raise_exception: bool = False,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         if config is None:
             if data is None or metrics is None:
                 raise ValueError(
@@ -340,9 +373,9 @@ class DeepEvalClient:
             )
 
         return self._post(
-            "/deepeval/schema-eval",
-            config.model_dump(),
-            raise_exception,
+            endpoint="/deepeval/schema-eval",
+            payload=config.model_dump(),
+            raise_exception=raise_exception,
         )
 
     def schema_eval_background(
@@ -352,13 +385,15 @@ class DeepEvalClient:
         raise_exception: bool = False,
         component: str = "unknown",
         label: Union[str, None] = None,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         """Run background schema evaluation (async job)."""
         payload = SchemaEvalConfig.model_construct(
             data=data, metrics=metrics, component=component, version=label
         ).model_dump()
         return self._post(
-            "/deepeval/schema-eval/background", payload, raise_exception
+            endpoint="/deepeval/schema-eval/background",
+            payload=payload,
+            raise_exception=raise_exception,
         )
 
     def maybe_text_eval(
@@ -369,7 +404,7 @@ class DeepEvalClient:
         raise_exception: bool = False,
         component: str = "unknown",
         label: Union[str, None] = None,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         """Randomly run text_eval based on a probability between 0 and 1."""
         self._validate_chance(chance)
         return (
@@ -392,7 +427,7 @@ class DeepEvalClient:
         raise_exception: bool = False,
         component: str = "unknown",
         label: Union[str, None] = None,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         """Randomly run text_eval_background based on a probability between 0 and 1."""
         self._validate_chance(chance)
         return (
@@ -411,7 +446,7 @@ class DeepEvalClient:
         raise_exception: bool = False,
         component: str = "unknown",
         label: Union[str, None] = None,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         """Randomly run schema_eval based on a probability between 0 and 1."""
         self._validate_chance(chance)
         return (
@@ -434,7 +469,7 @@ class DeepEvalClient:
         raise_exception: bool = False,
         component: str = "unknown",
         label: Union[str, None] = None,
-    ) -> Optional[dict]:
+    ) -> Optional[Dict]:
         """Randomly run text_eval_background based on a probability between 0 and 1."""
         self._validate_chance(chance)
         return (

rakam_eval_sdk/schema.py

@@ -129,3 +129,24 @@ class SchemaEvalConfig(BaseModel):
     label: Union[str, None] = None
     data: List[SchemaInputItem]
     metrics: List[SchemaMetricConfig] = Field(default_factory=list)
+
+
+class MetricDiff(BaseModel):
+    metric: str
+    score_a: Optional[float]
+    score_b: Optional[float]
+    delta: Optional[float]
+
+    success_a: Optional[bool]
+    success_b: Optional[bool]
+
+    threshold_a: Optional[float]
+    threshold_b: Optional[float]
+
+    status: str  # "unchanged" | "changed" | "added" | "removed"
+
+
+class TestCaseComparison(BaseModel):
+    testcase_a_id: int
+    testcase_b_id: int
+    metrics: List[MetricDiff]

{rakam_eval_sdk-0.2.2.dist-info → rakam_eval_sdk-0.2.4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: rakam-eval-sdk
-Version: 0.2.2
+Version: 0.2.4
 Summary: Evaluation Framework SDK
 Author: Mohamed Bachar Touil
 License: MIT

rakam_eval_sdk-0.2.4.dist-info/RECORD

@@ -0,0 +1,10 @@
+rakam_eval_sdk/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+rakam_eval_sdk/cli.py,sha256=IvC45LyALCCoxZ7ZNifVCbPs6UveGtUCfjr1q9e1A_Y,21658
+rakam_eval_sdk/client.py,sha256=7spK7P3h5uiA3Q3erOHOwf04N1_RT1pysWyt_E-cz0Q,14666
+rakam_eval_sdk/decorators.py,sha256=_9VFQmoYWd6cqnNryZJWEwYHQRxY7vIOam4z45zBk3c,1794
+rakam_eval_sdk/schema.py,sha256=YfZgHtPh41xFYdQTsbS0OSvNdI8yIv5zoh76v21hemo,3971
+rakam_eval_sdk/utils/decorator_utils.py,sha256=g0TjXtG9o4hwhUAFP8GJsXAkjhZhzeseTAg-YBFjj2g,1763
+rakam_eval_sdk-0.2.4.dist-info/WHEEL,sha256=eh7sammvW2TypMMMGKgsM83HyA_3qQ5Lgg3ynoecH3M,79
+rakam_eval_sdk-0.2.4.dist-info/entry_points.txt,sha256=tNhwmM_UGELb3h0zOfgCrtTheUkP-k8jGv0rTOfRSps,56
+rakam_eval_sdk-0.2.4.dist-info/METADATA,sha256=XOTmtgO7NiG8NuLn4B_yJujUJONSJCvwGQ9ak3zIhv4,6019
+rakam_eval_sdk-0.2.4.dist-info/RECORD,,

rakam_eval_sdk-0.2.2.dist-info/RECORD

@@ -1,10 +0,0 @@
-rakam_eval_sdk/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-rakam_eval_sdk/cli.py,sha256=LQBaTJGpXGRMNPkkqw_NzK4d1dbF8p4A4_8SN_XALgI,15500
-rakam_eval_sdk/client.py,sha256=e15d7ycm8G7oJuX_Ie9cB39TK33taMjFvVwc_4kLMww,13255
-rakam_eval_sdk/decorators.py,sha256=_9VFQmoYWd6cqnNryZJWEwYHQRxY7vIOam4z45zBk3c,1794
-rakam_eval_sdk/schema.py,sha256=woXu5GT1wuP_d7G11idopwpGmk0nV1SrxGShgWk-RA8,3530
-rakam_eval_sdk/utils/decorator_utils.py,sha256=g0TjXtG9o4hwhUAFP8GJsXAkjhZhzeseTAg-YBFjj2g,1763
-rakam_eval_sdk-0.2.2.dist-info/WHEEL,sha256=eh7sammvW2TypMMMGKgsM83HyA_3qQ5Lgg3ynoecH3M,79
-rakam_eval_sdk-0.2.2.dist-info/entry_points.txt,sha256=tNhwmM_UGELb3h0zOfgCrtTheUkP-k8jGv0rTOfRSps,56
-rakam_eval_sdk-0.2.2.dist-info/METADATA,sha256=2KBfQDA4cIq634ZFZ0z--FtOtty_PG_T8Pgh8a9CLJk,6019
-rakam_eval_sdk-0.2.2.dist-info/RECORD,,