feat: Reduce resources for test containers

.gitea/workflows/ci.yml

@@ -7,47 +7,53 @@ on:
     branches: ["main"]
 
 jobs:
-  test:
+  unit-tests:
     runs-on: ubuntu-latest
 
     steps:
       - uses: actions/checkout@v4
 
-      - uses: actions/setup-python@v5
+      - uses: astral-sh/setup-uv@v5
         with:
-          python-version: "3.12"
+          enable-cache: true
 
-      - name: Install dependencies
-        run: |
-          python -m venv .venv
-          .venv/bin/pip install -e ".[dev]" -q
+      - name: Run unit tests
+        run: uv run --extra dev pytest tests/unit/ -v --tb=short
 
-      - name: Run tests
-        run: .venv/bin/pytest tests/ -v
+  integration-tests:
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: astral-sh/setup-uv@v5
+        with:
+          enable-cache: true
+
+      - name: Pull container image
+        run: docker pull python:3.11-slim
+
+      - name: Run integration tests
+        run: uv run --extra dev pytest tests/integration/ -v --tb=short
 
   publish:
     runs-on: ubuntu-latest
-    needs: test
+    needs: [unit-tests, integration-tests]
     if: github.event_name == 'push' && github.ref == 'refs/heads/main'
 
     steps:
       - uses: actions/checkout@v4
 
-      - uses: actions/setup-python@v5
+      - uses: astral-sh/setup-uv@v5
         with:
-          python-version: "3.12"
+          enable-cache: true
 
       - name: Build package
-        run: |
-          pip install build -q
-          python -m build
+        run: uv build
 
       - name: Publish to Gitea package registry
         env:
-          TWINE_USERNAME: ${{ github.repository_owner }}
-          TWINE_PASSWORD: ${{ secrets.GITEA_TOKEN }}
-        run: |
-          pip install twine -q
-          twine upload \
-            --repository-url ${{ gitea.server_url }}/api/packages/${{ github.repository_owner }}/pypi \
-            dist/*
+          UV_PUBLISH_URL: ${{ gitea.server_url }}/api/packages/${{ github.repository_owner }}/pypi
+          UV_PUBLISH_USERNAME: ${{ github.repository_owner }}
+          UV_PUBLISH_PASSWORD: ${{ secrets.GITEA_TOKEN }}
+        run: uv publish

docker_agent_sandbox/tools/read_file.py

@@ -7,6 +7,8 @@ from typing import TYPE_CHECKING
 from langchain_core.tools import BaseTool, tool
 from loguru import logger
 
+from docker_agent_sandbox.tools._utils import _MAX_OUTPUT_CHARS
+
 if TYPE_CHECKING:
     from docker_agent_sandbox.sandbox import DockerSandbox
 
@@ -15,38 +17,69 @@ def make_read_file_tool(sandbox: "DockerSandbox") -> BaseTool:
     """Return a read_file tool bound to *sandbox*."""
 
     @tool
-    def read_file(path: str, offset: int = 0, length: int = 5000) -> str:
+    def read_file(path: str, start_line: int = 1, end_line: int | None = None) -> str:
         """
-        Read a file at *path*.
+        Read a file at *path*, returning its contents as text.
 
-        *path* can be absolute (``/tmp/re-agent/result.csv``) or relative to the
+        *path* can be absolute (/tmp/re-agent/result.csv) or relative to the
         working directory.
 
-        *offset* is the number of bytes to skip from the start of the file.
-        *length* is the maximum number of bytes to return.  If the file is
-        longer than ``offset + length``, the output is trimmed and a summary
-        line is appended showing how many bytes were omitted.
+        *start_line* is the 1-based line number to start reading from (default: 1).
 
-        Returns the (possibly trimmed) file contents as text, or an error message.
+        *end_line* is the last line to include, inclusive (default: read as many
+        lines as the MAX_CHARS cap allows). Use start_line/end_line to page
+        through large files in chunks.
+
+        At most 5,000 characters are returned per call. If the requested range
+        exceeds this, output is truncated and a summary line is appended showing
+        how many lines were omitted.
+
+        Returns the (possibly truncated) file contents, or an error message.
         """
         logger.debug(
-            "Reading file inside sandbox: {} offset={} length={}", path, offset, length
+            "Reading file inside sandbox: {} start_line={} end_line={}",
+            path,
+            start_line,
+            end_line,
         )
         try:
             data = sandbox.read_file(path)
         except (FileNotFoundError, IsADirectoryError, RuntimeError) as exc:
             return f"[ERROR reading {path!r}] {exc}"
 
-        total = len(data)
-        chunk = data[offset : offset + length]
-        text = chunk.decode("utf-8", errors="replace")
+        lines = data.decode("utf-8", errors="replace").splitlines(keepends=True)
+        total_lines = len(lines)
 
-        suffix = ""
-        if offset + length < total:
-            remaining = total - (offset + length)
-            suffix = f"\n[... {remaining} more bytes not shown (total {total} bytes). Use offset/length to read further.]"
-        elif offset > 0 or total > length:
-            suffix = f"\n[File total: {total} bytes, showing {len(chunk)} bytes from offset {offset}.]"
-        return text + suffix
+        # Clamp to valid range (1-based, inclusive)
+        start_idx = max(0, start_line - 1)
+        end_idx = total_lines if end_line is None else min(end_line, total_lines)
+        selected = lines[start_idx:end_idx]
+
+        # Enforce character cap
+        text = ""
+        last_included_line = start_idx  # track how far we got
+        for i, line in enumerate(selected):
+            if len(text) + len(line) > _MAX_OUTPUT_CHARS:
+                break
+            text += line
+            last_included_line = start_idx + i + 1  # 1-based
+
+        # Build informative suffix
+        suffix_parts = []
+        if last_included_line < end_idx:
+            omitted = end_idx - last_included_line
+            suffix_parts.append(
+                f"[... {omitted} more lines not shown (char cap reached). "
+                f"Call again with start_line={last_included_line + 1}.]"
+            )
+        elif end_idx < total_lines:
+            suffix_parts.append(
+                f"[Showing lines {start_line}–{end_idx} of {total_lines} total.]"
+            )
+
+        if suffix_parts:
+            text += "\n" + " ".join(suffix_parts)
+
+        return text
 
     return read_file

tests/integration/test_sandbox.py

@@ -46,12 +46,87 @@ def test_exec_returns_error_when_container_not_running():
     assert "not running" in out.lower()
 
 
+def test_exec_instant_command(sandbox: DockerSandbox):
+    code, out = sandbox.exec("true")
+    assert code == 0
+    assert out == ""
+
+
+def test_exec_instant_nonzero(sandbox: DockerSandbox):
+    code, _ = sandbox.exec("false")
+    assert code == 1
+
+
+def test_exec_delayed_command_within_timeout(sandbox: DockerSandbox):
+    code, out = sandbox.exec("sleep 1 && echo done", timeout=10)
+    assert code == 0
+    assert "done" in out
+
+
 def test_exec_timeout(sandbox: DockerSandbox):
     code, out = sandbox.exec("sleep 60", timeout=2)
     assert code == 124
     assert "timed out" in out
 
 
+def test_exec_timeout_longer_than_sleep(sandbox: DockerSandbox):
+    # Command finishes before timeout — must not raise or return 124.
+    code, out = sandbox.exec("sleep 1 && echo ok", timeout=10)
+    assert code == 0
+    assert "ok" in out
+
+
+def test_exec_and_chain_both_succeed(sandbox: DockerSandbox):
+    code, out = sandbox.exec("echo first && echo second")
+    assert code == 0
+    assert "first" in out
+    assert "second" in out
+
+
+def test_exec_and_chain_short_circuits_on_failure(sandbox: DockerSandbox):
+    code, out = sandbox.exec("false && echo should_not_print")
+    assert code != 0
+    assert "should_not_print" not in out
+
+
+def test_exec_pipe(sandbox: DockerSandbox):
+    code, out = sandbox.exec("echo hello world | tr ' ' '_'")
+    assert code == 0
+    assert "hello_world" in out
+
+
+def test_exec_pipe_exit_code_is_last_command(sandbox: DockerSandbox):
+    # grep finds no match → exit 1, even though echo succeeded
+    code, _ = sandbox.exec("echo hello | grep nomatch")
+    assert code == 1
+
+
+def test_exec_stdout_redirect_to_file(sandbox: DockerSandbox, workdir: str):
+    code, out = sandbox.exec(f"echo redirected > {workdir}/out.txt && cat {workdir}/out.txt")
+    assert code == 0
+    assert "redirected" in out
+
+
+def test_exec_stderr_redirect_to_stdout(sandbox: DockerSandbox):
+    code, out = sandbox.exec("echo err_msg >&2 2>&1")
+    assert code == 0
+    assert "err_msg" in out
+
+
+def test_exec_subshell(sandbox: DockerSandbox):
+    code, out = sandbox.exec("result=$(echo inner) && echo $result")
+    assert code == 0
+    assert "inner" in out
+
+
+def test_exec_multiline_via_semicolons(sandbox: DockerSandbox):
+    code, out = sandbox.exec("echo a; echo b; echo c")
+    assert code == 0
+    assert "a" in out
+    assert "b" in out
+    assert "c" in out
+
+
 def test_exec_working_dir_respected():
     """When working_dir is set, exec uses it as cwd."""
     sb = DockerSandbox(
@@ -59,6 +134,8 @@ def test_exec_working_dir_respected():
         image="python:3.11-slim",
         command="sleep infinity",
         working_dir="/tmp",
+        cpu_limit=1,
+        memory_limit="256m",
     )
     sb.start()
     try:
@@ -174,6 +251,8 @@ def test_context_manager_stops_container():
         container_name="test-ctx-manager",
         image="python:3.11-slim",
         command="sleep infinity",
+        cpu_limit=1,
+        memory_limit="256m",
     )
     sb.start()
     with sb:

tests/integration/test_tools.py

@@ -7,7 +7,6 @@ exactly as they would be when called by an LLM agent.
 
 from __future__ import annotations
 
-import pytest
 from langchain_core.tools import BaseTool
 
 from docker_agent_sandbox import (
@@ -26,7 +25,6 @@ from docker_agent_sandbox import (
     make_write_file_tool,
 )
 
-
 # ---------------------------------------------------------------------------
 # bash
 # ---------------------------------------------------------------------------
@@ -113,7 +111,9 @@ def test_read_file_pagination(sandbox: DockerSandbox, workdir: str):
     content = "\n".join(f"line{i}" for i in range(1, 11)) + "\n"
     sandbox.write_file(f"{workdir}/paged.txt", content)
     tool = make_read_file_tool(sandbox)
-    result = tool.invoke({"path": f"{workdir}/paged.txt", "start_line": 3, "end_line": 5})
+    result = tool.invoke(
+        {"path": f"{workdir}/paged.txt", "start_line": 3, "end_line": 5}
+    )
     assert "line3" in result
     assert "line5" in result
     assert "line1" not in result
@@ -124,7 +124,9 @@ def test_read_file_shows_total_line_count(sandbox: DockerSandbox, workdir: str):
     content = "\n".join(f"line{i}" for i in range(1, 21)) + "\n"
     sandbox.write_file(f"{workdir}/info.txt", content)
     tool = make_read_file_tool(sandbox)
-    result = tool.invoke({"path": f"{workdir}/info.txt", "start_line": 1, "end_line": 5})
+    result = tool.invoke(
+        {"path": f"{workdir}/info.txt", "start_line": 1, "end_line": 5}
+    )
     # There are 20 lines but we only requested 1-5, suffix should mention totals.
     assert "20" in result
 
@@ -187,7 +189,9 @@ def test_edit_file_delete_block(sandbox: DockerSandbox, workdir: str):
 
 def test_edit_file_missing_file_returns_error(sandbox: DockerSandbox, workdir: str):
     tool = make_edit_file_tool(sandbox)
-    result = tool.invoke({"path": f"{workdir}/ghost.txt", "old_str": "x", "new_str": "y"})
+    result = tool.invoke(
+        {"path": f"{workdir}/ghost.txt", "old_str": "x", "new_str": "y"}
+    )
     assert result.startswith("[ERROR")
 
 
@@ -195,7 +199,9 @@ def test_edit_file_multiline_replace(sandbox: DockerSandbox, workdir: str):
     path = f"{workdir}/multi.txt"
     sandbox.write_file(path, "line1\nline2\nline3\n")
     tool = make_edit_file_tool(sandbox)
-    result = tool.invoke({"path": path, "old_str": "line1\nline2\n", "new_str": "replaced\n"})
+    result = tool.invoke(
+        {"path": path, "old_str": "line1\nline2\n", "new_str": "replaced\n"}
+    )
     assert result.startswith("[OK]")
     assert sandbox.read_file(path) == b"replaced\nline3\n"
 
@@ -438,7 +444,9 @@ def test_grep_recursive(sandbox: DockerSandbox, workdir: str):
     sandbox.write_file(f"{workdir}/d/a.txt", "find_me\n")
     sandbox.write_file(f"{workdir}/d/b.txt", "not here\n")
     tool = make_grep_tool(sandbox)
-    result = tool.invoke({"pattern": "find_me", "path": f"{workdir}/d", "recursive": True})
+    result = tool.invoke(
+        {"pattern": "find_me", "path": f"{workdir}/d", "recursive": True}
+    )
     assert "find_me" in result
     assert "a.txt" in result