inspect-swe 0.2.7__tar.gz → 0.2.9__tar.gz

{inspect_swe-0.2.7 → inspect_swe-0.2.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: inspect_swe
-Version: 0.2.7
+Version: 0.2.9
 Summary: Software engineering agents for Inspect AI.
 Project-URL: Documentation, https://meridianlabs-ai.github.io/inspect_swe/
 Project-URL: Source Code, https://github.com/meridianlabs-ai/inspect_swe
@@ -10,7 +10,7 @@ License: MIT License
 License-File: LICENSE
 Requires-Python: >=3.10
 Requires-Dist: httpx
-Requires-Dist: inspect-ai>=0.3.126
+Requires-Dist: inspect-ai>=0.3.128
 Requires-Dist: nest-asyncio
 Requires-Dist: platformdirs
 Requires-Dist: pydantic>=2.11.4

{inspect_swe-0.2.7 → inspect_swe-0.2.9}/pyproject.toml

@@ -12,7 +12,7 @@ requires-python = ">=3.10"
 license = { text = "MIT License" }
 dependencies = [
     "httpx",
-    "inspect_ai>=0.3.126",
+    "inspect_ai>=0.3.128",
     "nest_asyncio",
     "platformdirs",
     "pydantic>=2.11.4",

{inspect_swe-0.2.7 → inspect_swe-0.2.9}/src/inspect_swe/_claude_code/claude_code.py

@@ -16,6 +16,8 @@ from inspect_ai.tool import MCPServerConfig
 from inspect_ai.util import sandbox as sandbox_env
 from pydantic_core import to_json
 
+from inspect_swe._util.trace import trace
+
 from .._util._async import is_callable_coroutine
 from .install.install import ensure_claude_code_installed
 
@@ -34,10 +36,11 @@ def claude_code(
     attempts: int | AgentAttempts = 1,
     model: str | None = None,
     small_model: str | None = None,
+    cwd: str | None = None,
     env: dict[str, str] | None = None,
-    version: Literal["auto", "sandbox", "stable", "latest"] | str = "auto",
     user: str | None = None,
     sandbox: str | None = None,
+    version: Literal["auto", "sandbox", "stable", "latest"] | str = "auto",
 ) -> Agent:
     """Claude Code agent.
 
@@ -60,15 +63,16 @@ def claude_code(
         attempts: Configure agent to make multiple attempts.
         model: Model name to use for Opus and Sonnet calls (defaults to main model for task).
         small_model: Model to use for Haiku calls (defaults to main model for task).
+        cwd: Working directory to run claude code within.
         env: Environment variables to set for claude code.
+        user: User to execute claude code with.
+        sandbox: Optional sandbox environment name.
         version: Version of claude code to use. One of:
             - "auto": Use any available version of claude code in the sandbox, otherwise download the current stable version.
             - "sandbox": Use the version of claude code in the sandbox (raises `RuntimeError` if claude is not available in the sandbox)
             - "stable": Download and use the current stable version of claude code.
             - "latest": Download and use the very latest version of claude code.
             - "x.x.x": Download and use a specific version of claude code.
-        user: User to execute claude code with.
-        sandbox: Optional sandbox environment name.
     """
     # resolve models
     model = f"inspect/{model}" if model is not None else "inspect"
@@ -91,6 +95,8 @@ def claude_code(
             cmd = [
                 "--print",  # run without interactions
                 "--dangerously-skip-permissions",
+                "--debug",
+                "--verbose",
                 "--model",
                 model,
             ]
@@ -127,7 +133,8 @@ def claude_code(
             # resolve sandbox
             sbox = sandbox_env(sandbox)
 
-            # execute the agent
+            # execute the agent (track debug output)
+            debug_output: list[str] = []
             agent_prompt = prompt
             attempt_count = 0
             while True:
@@ -140,6 +147,7 @@ def claude_code(
                 # run agent
                 result = await sbox.exec(
                     cmd=agent_cmd,
+                    cwd=cwd,
                     env={
                         "ANTHROPIC_BASE_URL": f"http://localhost:{bridge.port}",
                         "ANTHROPIC_API_KEY": "sk-ant-api03-DOq5tyLPrk9M4hPE",
@@ -156,6 +164,10 @@ def claude_code(
                     user=user,
                 )
 
+                # track debug output
+                debug_output.append(result.stdout)
+                debug_output.append(result.stderr)
+
                 # raise for error
                 if not result.success:
                     f"Error executing claude code agent: {result.stdout}\n{result.stderr}"
@@ -185,6 +197,10 @@ def claude_code(
                     else:
                         agent_prompt = attempts.incorrect_message
 
+            # trace debug info
+            debug_output.insert(0, "Claude Code Debug Output:")
+            trace("\n".join(debug_output))
+
         return bridge.state
 
     # return agent with specified name and descritpion

{inspect_swe-0.2.7 → inspect_swe-0.2.9}/src/inspect_swe/_claude_code/install/download.py

@@ -1,12 +1,11 @@
 import re
-from typing import Literal
+from typing import Callable, Literal
 
 from pydantic import BaseModel
 
 from ..._util.checksum import verify_checksum
 from ..._util.download import download_file, download_text_file
 from ..._util.sandbox import SandboxPlatform
-from ..._util.trace import trace
 from .cache import (
     read_cached_claude_code_binary,
     write_cached_claude_code_binary,
@@ -14,8 +13,13 @@ from .cache import (
 
 
 async def download_claude_code_async(
-    version: Literal["stable", "latest"] | str, platform: SandboxPlatform
+    version: Literal["stable", "latest"] | str,
+    platform: SandboxPlatform,
+    logger: Callable[[str], None] | None = None,
 ) -> bytes:
+    # resovle logger
+    logger = logger or print
+
     # determine version and checksum
     gcs_bucket = await _claude_code_gcs_bucket()
     version = await _claude_code_version(gcs_bucket, version)
@@ -35,9 +39,9 @@ async def download_claude_code_async(
         write_cached_claude_code_binary(binary_data, version, platform)
 
         # trace
-        trace(f"Downloaded claude code binary: {version} ({platform})")
+        logger(f"Downloaded claude code binary: {version} ({platform})")
     else:
-        trace(f"Used claude code binary from cache: {version} ({platform})")
+        logger(f"Used claude code binary from cache: {version} ({platform})")
 
     # return data
     return binary_data

{inspect_swe-0.2.7 → inspect_swe-0.2.9}/src/inspect_swe/_claude_code/install/install.py

@@ -50,7 +50,9 @@ async def ensure_claude_code_installed(
 
         # download the binary
         if claude_binary_bytes is None:
-            claude_binary_bytes = await download_claude_code_async(version, platform)
+            claude_binary_bytes = await download_claude_code_async(
+                version, platform, trace
+            )
 
         # write it into the container and return it
         claude_binary = f"/opt/claude-{version}-{platform}"

inspect_swe-0.2.9/src/inspect_swe/_util/trace.py

@@ -0,0 +1,10 @@
+from logging import getLogger
+
+from inspect_ai.util import trace_message
+
+logger = getLogger(__file__)
+
+
+def trace(message: str) -> None:
+    logger.setLevel("TRACE")
+    trace_message(logger, category="Inspect SWE", message=message)

{inspect_swe-0.2.7 → inspect_swe-0.2.9}/src/inspect_swe/_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.2.7'
-__version_tuple__ = version_tuple = (0, 2, 7)
+__version__ = version = '0.2.9'
+__version_tuple__ = version_tuple = (0, 2, 9)
 
 __commit_id__ = commit_id = None

inspect_swe-0.2.7/src/inspect_swe/_util/trace.py

@@ -1,7 +0,0 @@
-from logging import getLogger
-
-logger = getLogger(__file__)
-
-
-def trace(message: str) -> None:
-    logger.info(f"[Inspect SWE] {message}")