hte-cli 0.2.17__py3-none-any.whl → 0.2.19__py3-none-any.whl

hte_cli/cli.py

@@ -274,6 +274,7 @@ def session_join(ctx, session_id: str, force_setup: bool):
         "task": {
             "instructions": session_info.get("instructions", ""),
             "metadata": session_info.get("metadata", {}),
+            "target": session_info.get("target", ""),
             "scorer_type": session_info.get("scorer_type"),
             "intermediate_scoring": session_info.get("intermediate_scoring", False),
         },

hte_cli/runner.py

@@ -271,12 +271,22 @@ class TaskRunner:
         file_dest_base = _get_file_dest_from_compose(compose_path)
 
         files_to_mount: dict[str, str] = {}
-        excluded_files = {"compose.yaml", "README.md", "instructions.txt"}
-        for file_path in task_dir.iterdir():
-            if file_path.is_file() and file_path.name not in excluded_files:
-                dest_path = f"{file_dest_base}/{file_path.name}"
-                files_to_mount[dest_path] = str(file_path)
-                logger.info(f"Will mount file: {file_path.name} -> {dest_path}")
+        excluded_files = {"compose.yaml", "README.md", "instructions.txt", "logs"}
+
+        # Walk directory tree recursively to handle nested file structures
+        # (e.g., cybench tasks with blog/, nginx/, rcache/ subdirectories)
+        for file_path in task_dir.rglob("*"):
+            if not file_path.is_file():
+                continue
+            # Get relative path from task_dir
+            rel_path = file_path.relative_to(task_dir)
+            # Skip excluded files (check both filename and first directory component)
+            if rel_path.name in excluded_files or rel_path.parts[0] in excluded_files:
+                continue
+            # Build destination path preserving directory structure
+            dest_path = f"{file_dest_base}/{rel_path}"
+            files_to_mount[dest_path] = str(file_path)
+            logger.info(f"Will mount file: {rel_path} -> {dest_path}")
 
         # Create the Inspect task
         inspect_task = self.create_inspect_task(
@@ -364,16 +374,31 @@ class TaskRunner:
         task_id = assignment["task_id"]
         task_data = assignment.get("task", {})
         instructions = task_data.get("instructions", "")
-        # Target can be at task level, in metadata, or in dataset_task_metadata.flag
-        target = task_data.get("target", "") or task_data.get("metadata", {}).get("target", "")
-        # Fallback to dataset_task_metadata.flag for CTF benchmarks (nyuctf, cybench, etc.)
+        # Target comes from backend (preferred) or can be extracted from metadata
+        target = task_data.get("target", "")
         if not target or target == "?":
-            dataset_meta = task_data.get("dataset_task_metadata", {})
-            target = dataset_meta.get("flag", "") or dataset_meta.get("answer", "")
+            # Fallback to metadata fields for various benchmarks
+            meta = task_data.get("metadata", {})
+            target = (
+                meta.get("target", "")
+                or meta.get("solution_flag", "")  # intercode-ctf
+                or meta.get("flag", "")  # cybench, nyuctf
+                or meta.get("answer", "")  # generic
+            )
 
-        # Extract scoring configuration from backend
-        scorer_type = task_data["scorer_type"]
-        intermediate_scoring = task_data["intermediate_scoring"]
+        # Extract scoring configuration from backend (required fields)
+        scorer_type = task_data.get("scorer_type")
+        if not scorer_type:
+            raise ValueError(
+                f"Backend did not return scorer_type for task {task_id}. "
+                "This is a backend configuration error."
+            )
+        intermediate_scoring = task_data.get("intermediate_scoring")
+        if intermediate_scoring is None:
+            raise ValueError(
+                f"Backend did not return intermediate_scoring for task {task_id}. "
+                "This is a backend configuration error."
+            )
 
         return self.run(
             task_id=task_id,

hte_cli/scorers.py

@@ -137,16 +137,25 @@ def get_scorer(scorer_type: ScorerType | str, target: str) -> Scorer | None:
     Get the appropriate scorer based on type.
 
     Matches *_human.py implementations for human/AI parity.
+    Raises ValueError for invalid configurations.
     """
     # Handle string input
     if isinstance(scorer_type, str):
         try:
             scorer_type = ScorerType(scorer_type)
         except ValueError:
-            return None
+            raise ValueError(
+                f"Unknown scorer type: {scorer_type}. "
+                f"Valid types: {[s.value for s in ScorerType]}"
+            )
 
     if scorer_type == ScorerType.FLAG_INCLUDES:
-        return includes() if target else None
+        if not target:
+            raise ValueError(
+                "FLAG_INCLUDES scorer requires a target flag but none was provided. "
+                "Backend should return 'target' in session info."
+            )
+        return includes()
     elif scorer_type == ScorerType.LLM_BASH:
         # LLM-based scoring happens server-side, no client scorer
         return None

{hte_cli-0.2.17.dist-info → hte_cli-0.2.19.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hte-cli
-Version: 0.2.17
+Version: 0.2.19
 Summary: Human Time-to-Completion Evaluation CLI
 Project-URL: Homepage, https://github.com/sean-peters-au/lyptus-mono
 Author: Lyptus Research
@@ -25,6 +25,7 @@ Requires-Dist: pyyaml>=6.0
 Requires-Dist: rich>=13.0
 Provides-Extra: dev
 Requires-Dist: pexpect>=4.8; extra == 'dev'
+Requires-Dist: pytest-asyncio>=0.23; extra == 'dev'
 Requires-Dist: pytest>=7.0; extra == 'dev'
 Requires-Dist: requests>=2.28; extra == 'dev'
 Description-Content-Type: text/markdown

{hte_cli-0.2.17.dist-info → hte_cli-0.2.19.dist-info}/RECORD

@@ -1,15 +1,15 @@
 hte_cli/__init__.py,sha256=fDGXp-r8bIoLtlQnn5xJ_CpwMhonvk9bGjZQsjA2mDI,914
 hte_cli/__main__.py,sha256=63n0gNGfskidWDU0aAIF2N8lylVCLYKVIkrN9QiORoo,107
 hte_cli/api_client.py,sha256=m42kfFZS72Nu_VuDwxRsLNy4ziCcvgk7KNWBh9gwqy0,9257
-hte_cli/cli.py,sha256=lyCuhnSiViTKpWYAeVQ1C4D5idg4FRh_q39PXat7HXg,42261
+hte_cli/cli.py,sha256=cJ4g9UgBXHfmcNe4mu9imL8DSKkYzVDp8sR785z8h8M,42315
 hte_cli/config.py,sha256=42Xv__YMSeRLs2zhGukJkIXFKtnBtYCHnONfViGyt2g,3387
 hte_cli/errors.py,sha256=1J5PpxcUKBu6XjigMMCPOq4Zc12tnv8LhAsiaVFWLQM,2762
 hte_cli/events.py,sha256=Zn-mroqaLHNzdT4DFf8st1Qclglshihdc09dBfCN070,5522
 hte_cli/image_utils.py,sha256=TLwJdswUQrSD2bQcAXW03R8j8WG2pbHzd12TWcE7zy4,6418
-hte_cli/runner.py,sha256=DhC8FMjHwfLR193iP4thLDRZrNssYA9KH1WYKU2JKeg,13535
-hte_cli/scorers.py,sha256=sFoPJePRt-K191-Ga4cVmrldruJclYXTOLkU_C9nCDI,6025
+hte_cli/runner.py,sha256=SWl9FF4X3e9eBbZyL0ujhmmSL5OK8J6st-Ty0jD5AWM,14550
+hte_cli/scorers.py,sha256=NZWMlS2h2Hczm-bldH35wRhL3RYzGhQgCCp3rP9zhJo,6414
 hte_cli/version_check.py,sha256=WVZyGy2XfAghQYdd2N9-0Qfg-7pgp9gt4761-PnmacI,1708
-hte_cli-0.2.17.dist-info/METADATA,sha256=tVT3em2qTUXNMZeWKRk1vHp2fuO5SV3kl43dzPyCqEE,3768
-hte_cli-0.2.17.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-hte_cli-0.2.17.dist-info/entry_points.txt,sha256=XbyEEi1H14DFAt0Kdl22e_IRVEGzimSzYSh5HlhKlFA,41
-hte_cli-0.2.17.dist-info/RECORD,,
+hte_cli-0.2.19.dist-info/METADATA,sha256=-JBgXEkarLFsDv82zZ36XVsEoOgMiIQR5UFnEmJ8ssI,3820
+hte_cli-0.2.19.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+hte_cli-0.2.19.dist-info/entry_points.txt,sha256=XbyEEi1H14DFAt0Kdl22e_IRVEGzimSzYSh5HlhKlFA,41
+hte_cli-0.2.19.dist-info/RECORD,,