@pushpalsdev/cli 1.0.18 → 1.0.19

package/runtime/sandbox/apps/workerpals/src/backends/shared/executor_base.py

@@ -0,0 +1,746 @@
+"""
+Shared infrastructure for PushPals executor scripts.
+
+Both ``miniswe_executor.py`` and ``openhands_executor.py`` (and any future
+executors) import from here instead of duplicating config loading, LLM
+resolution, result emission, payload decoding, git helpers, etc.
+"""
+
+from __future__ import annotations
+
+import base64
+import json
+import os
+import re
+import subprocess
+import sys
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Callable, Dict, List, Mapping, Optional, Sequence, Set, Tuple
+
+try:
+    import tomllib
+except Exception:  # pragma: no cover - python <3.11 fallback
+    tomllib = None  # type: ignore[assignment]
+
+
+# ─── Constants ───────────────────────────────────────────────────────────────
+
+RESULT_PREFIX = "__PUSHPALS_OH_RESULT__ "
+
+KNOWN_LITELLM_PROVIDER_PREFIXES: Set[str] = {
+    "openai",
+    "azure",
+    "ollama",
+    "openrouter",
+    "anthropic",
+    "google",
+    "gemini",
+    "vertex_ai",
+    "bedrock",
+    "cohere",
+    "groq",
+    "mistral",
+    "huggingface",
+    "replicate",
+    "deepseek",
+    "xai",
+    "together_ai",
+    "fireworks_ai",
+}
+
+DEFAULT_TOOLCALL_RETRY_MAX = 1
+
+# Superset of signals from both executors indicating the model failed to
+# emit tool calls / tool actions.
+NO_TOOL_CALL_SIGNALS: Tuple[str, ...] = (
+    "no tool calls found",
+    "no tool call found",
+    "no function calls found",
+    "no function call found",
+    "tool_calls",
+    "function_call",
+    "did not call any tools",
+    "didn't call any tools",
+    "tool use required",
+    "must use tools",
+    "no actions found",
+    "no action found",
+    "no tool messages",
+)
+
+# ─── Core helpers ────────────────────────────────────────────────────────────
+
+def emit(result: Dict[str, Any]) -> None:
+    """Write a structured result line that the TS host parses."""
+    sys.stdout.write(f"{RESULT_PREFIX}{json.dumps(result, ensure_ascii=True)}\n")
+    sys.stdout.flush()
+
+
+def executor_log(message: str) -> None:
+    line = message if message.endswith("\n") else f"{message}\n"
+    sys.stdout.write(line)
+    sys.stdout.flush()
+
+
+def _debug_enabled() -> bool:
+    return os.environ.get("WORKERPALS_DEBUG", "").strip().lower() in {"1", "true", "yes"}
+
+
+class Logger:
+    """Simple levelled logger for executor scripts.
+
+    Usage::
+
+        log = Logger("[MiniSweExecutor]")
+        log.info("Starting execution")
+        log.debug("Instruction: ...")   # only when WORKERPALS_DEBUG=1
+    """
+
+    def __init__(self, prefix: str) -> None:
+        self.prefix = prefix
+
+    def info(self, message: str) -> None:
+        executor_log(f"{self.prefix} {message}")
+
+    def debug(self, message: str) -> None:
+        if _debug_enabled():
+            executor_log(f"{self.prefix} {message}")
+
+
+def fail(summary: str, stderr: Optional[str] = None, exit_code: int = 1) -> int:
+    """Emit a failure result and return the exit code."""
+    emit({"ok": False, "summary": summary, "stderr": stderr or "", "exitCode": exit_code})
+    return exit_code
+
+
+def decode_payload(raw: str) -> Dict[str, Any]:
+    decoded = base64.b64decode(raw).decode("utf-8")
+    payload = json.loads(decoded)
+    if not isinstance(payload, dict):
+        raise ValueError("payload must be a JSON object")
+    return payload
+
+
+def resolve_repo_within_assigned_root(repo: str) -> Tuple[Optional[str], Optional[str]]:
+    raw_repo = str(repo or "").strip()
+    if not raw_repo:
+        return None, "Invalid payload: missing 'repo'"
+
+    try:
+        repo_path = Path(raw_repo).resolve()
+    except Exception as exc:
+        return None, f"Invalid payload repo path: {exc}"
+
+    if not repo_path.exists() or not repo_path.is_dir():
+        return None, f"Invalid payload repo path: not a directory ({repo_path})"
+
+    assigned_raw = (os.environ.get("PUSHPALS_ASSIGNED_REPO_ROOT") or "").strip()
+    if assigned_raw:
+        try:
+            assigned_root = Path(assigned_raw).resolve()
+        except Exception as exc:
+            return None, f"Invalid assigned repo root: {exc}"
+        if repo_path != assigned_root and assigned_root not in repo_path.parents:
+            return (
+                None,
+                "Refusing repo path outside assigned root: "
+                f"repo={repo_path} assigned_root={assigned_root}",
+            )
+
+    return str(repo_path), None
+
+
+def to_int(value: Any, default: int) -> int:
+    try:
+        return int(value)
+    except Exception:
+        return default
+
+
+def to_float(value: Any, default: float) -> float:
+    try:
+        return float(value)
+    except Exception:
+        return default
+
+
+def to_single_line(value: Any, max_chars: int = 240) -> str:
+    text = str(value or "").replace("\r", " ").replace("\n", " ").strip()
+    if not text:
+        return ""
+    if len(text) <= max_chars:
+        return text
+    return text[: max(1, max_chars - 3)] + "..."
+
+
+def is_no_tool_calls_error(exc: Exception) -> bool:
+    lowered = str(exc).lower()
+    return any(sig in lowered for sig in NO_TOOL_CALL_SIGNALS)
+
+
+# ─── Config loading (TOML + env) ────────────────────────────────────────────
+
+_CONFIG_CACHE: Optional[Dict[str, Any]] = None
+_MISSING = object()
+
+
+def _deep_merge(base: Dict[str, Any], override: Dict[str, Any]) -> Dict[str, Any]:
+    out = dict(base)
+    for key, value in override.items():
+        existing = out.get(key)
+        if isinstance(existing, dict) and isinstance(value, dict):
+            out[key] = _deep_merge(existing, value)
+        else:
+            out[key] = value
+    return out
+
+
+def repo_root_for_runtime_config() -> Path:
+    explicit = (os.environ.get("PUSHPALS_REPO_PATH") or "").strip()
+    if explicit:
+        return Path(explicit)
+    return Path(__file__).resolve().parents[3]
+
+
+def _parse_toml_file(path: Path) -> Dict[str, Any]:
+    if not path.exists() or not tomllib:
+        return {}
+    try:
+        parsed = tomllib.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return {}
+    return parsed if isinstance(parsed, dict) else {}
+
+
+def runtime_config() -> Dict[str, Any]:
+    global _CONFIG_CACHE
+    if _CONFIG_CACHE is not None:
+        return _CONFIG_CACHE
+    repo_root = repo_root_for_runtime_config()
+    legacy_config_dir = repo_root / "config"
+    config_dir = repo_root / "configs"
+    if not (config_dir / "default.toml").exists():
+        if (legacy_config_dir / "default.toml").exists():
+            config_dir = legacy_config_dir
+    default_cfg = _parse_toml_file(config_dir / "default.toml")
+    profile = (
+        (os.environ.get("PUSHPALS_PROFILE") or "").strip()
+        or str(default_cfg.get("profile") or "").strip()
+        or "dev"
+    )
+    profile_cfg = _parse_toml_file(config_dir / f"{profile}.toml")
+    local_cfg = _parse_toml_file(config_dir / "local.toml")
+    if (
+        not local_cfg
+        and config_dir != legacy_config_dir
+        and (legacy_config_dir / "local.toml").exists()
+    ):
+        local_cfg = _parse_toml_file(legacy_config_dir / "local.toml")
+    _CONFIG_CACHE = _deep_merge(_deep_merge(default_cfg, profile_cfg), local_cfg)
+    return _CONFIG_CACHE
+
+
+class SettingsResolver:
+    """Thin config interface over env + runtime TOML config.
+
+    This isolates source precedence (env vs TOML paths) from backend logic so
+    call sites consume stable typed accessors.
+    """
+
+    def __init__(
+        self,
+        *,
+        env: Optional[Mapping[str, str]] = None,
+        config_loader: Optional[Callable[[], Dict[str, Any]]] = None,
+    ) -> None:
+        self._env: Mapping[str, str] = env if env is not None else os.environ
+        self._config_loader: Callable[[], Dict[str, Any]] = config_loader or runtime_config
+
+    def _config_value(self, path: str, default: Any = _MISSING) -> Any:
+        node: Any = self._config_loader()
+        for part in path.split("."):
+            if not isinstance(node, dict) or part not in node:
+                return default
+            node = node[part]
+        return node
+
+    def _first_env(self, names: Sequence[str]) -> Any:
+        for name in names:
+            raw = self._env.get(name)
+            if raw is None:
+                continue
+            text = str(raw).strip()
+            if text:
+                return text
+        return _MISSING
+
+    def _first_config(self, paths: Sequence[str]) -> Any:
+        for path in paths:
+            value = self._config_value(path, _MISSING)
+            if value is _MISSING:
+                continue
+            if isinstance(value, str):
+                trimmed = value.strip()
+                if trimmed:
+                    return trimmed
+                continue
+            return value
+        return _MISSING
+
+    def get_str(
+        self,
+        *,
+        env_names: Sequence[str] = (),
+        config_paths: Sequence[str] = (),
+        default: str = "",
+    ) -> str:
+        env_value = self._first_env(env_names)
+        if env_value is not _MISSING:
+            return str(env_value)
+        cfg_value = self._first_config(config_paths)
+        if cfg_value is _MISSING:
+            return default
+        return str(cfg_value).strip() or default
+
+    def get_int(
+        self,
+        *,
+        env_names: Sequence[str] = (),
+        config_paths: Sequence[str] = (),
+        default: int,
+    ) -> int:
+        env_value = self._first_env(env_names)
+        if env_value is not _MISSING:
+            return to_int(env_value, default)
+        cfg_value = self._first_config(config_paths)
+        if cfg_value is _MISSING:
+            return default
+        return to_int(cfg_value, default)
+
+    def get_float(
+        self,
+        *,
+        env_names: Sequence[str] = (),
+        config_paths: Sequence[str] = (),
+        default: float,
+    ) -> float:
+        env_value = self._first_env(env_names)
+        if env_value is not _MISSING:
+            return to_float(env_value, default)
+        cfg_value = self._first_config(config_paths)
+        if cfg_value is _MISSING:
+            return default
+        return to_float(cfg_value, default)
+
+    def get_bool(
+        self,
+        *,
+        env_names: Sequence[str] = (),
+        config_paths: Sequence[str] = (),
+        default: bool = False,
+    ) -> bool:
+        env_value = self._first_env(env_names)
+        if env_value is not _MISSING:
+            lowered = str(env_value).strip().lower()
+            if lowered in {"1", "true", "yes", "on"}:
+                return True
+            if lowered in {"0", "false", "no", "off"}:
+                return False
+            return default
+
+        cfg_value = self._first_config(config_paths)
+        if cfg_value is _MISSING:
+            return default
+        if isinstance(cfg_value, bool):
+            return cfg_value
+        if isinstance(cfg_value, (int, float)):
+            return bool(cfg_value)
+        if isinstance(cfg_value, str):
+            lowered = cfg_value.strip().lower()
+            if lowered in {"1", "true", "yes", "on"}:
+                return True
+            if lowered in {"0", "false", "no", "off"}:
+                return False
+        return default
+
+
+def build_settings_resolver(
+    *,
+    env: Optional[Mapping[str, str]] = None,
+    config: Optional[Dict[str, Any]] = None,
+) -> SettingsResolver:
+    if config is None:
+        return SettingsResolver(env=env)
+    return SettingsResolver(env=env, config_loader=lambda: config)
+
+
+def config_get(path: str, default: Any = None) -> Any:
+    return build_settings_resolver()._config_value(path, default)
+
+
+def setting_str(name: str, config_path: str, default: str = "") -> str:
+    return build_settings_resolver().get_str(
+        env_names=(name,),
+        config_paths=(config_path,),
+        default=default,
+    )
+
+
+def setting_int(name: str, config_path: str, default: int) -> int:
+    return build_settings_resolver().get_int(
+        env_names=(name,),
+        config_paths=(config_path,),
+        default=default,
+    )
+
+
+def setting_float(name: str, config_path: str, default: float) -> float:
+    return build_settings_resolver().get_float(
+        env_names=(name,),
+        config_paths=(config_path,),
+        default=default,
+    )
+
+
+def setting_bool(name: str, config_path: str, default: bool = False) -> bool:
+    return build_settings_resolver().get_bool(
+        env_names=(name,),
+        config_paths=(config_path,),
+        default=default,
+    )
+
+
+def is_truthy_env(name: str, default: bool = False, config_path: str = "") -> bool:
+    if config_path:
+        return setting_bool(name, config_path, default)
+    return build_settings_resolver().get_bool(env_names=(name,), default=default)
+
+
+# ─── LLM config resolution ──────────────────────────────────────────────────
+
+def _normalize_base_url(raw: str) -> str:
+    base = raw.strip()
+    if not base:
+        return ""
+    base = base.rstrip("/")
+    if base.endswith("/api/chat"):
+        base = base[: -len("/api/chat")]
+    if base.endswith("/chat/completions"):
+        base = base[: -len("/chat/completions")]
+    return base
+
+
+def _model_is_provider_qualified(model: str) -> bool:
+    if "/" not in model:
+        return False
+    provider = model.split("/", 1)[0].strip().lower()
+    return provider in KNOWN_LITELLM_PROVIDER_PREFIXES
+
+
+def infer_litellm_provider(base_url: str) -> str:
+    backend = setting_str("WORKERPALS_LLM_BACKEND", "workerpals.llm.backend", "").lower()
+    if backend in {"ollama", "ollama_chat"}:
+        return "ollama"
+    if backend in {"lmstudio", "openai", "openai_compatible"}:
+        return "openai"
+    lowered = base_url.lower()
+    if "11434" in lowered:
+        return "ollama"
+    return "openai"
+
+
+def _normalize_litellm_model(model: str, provider: str) -> str:
+    normalized = model.strip()
+    if not normalized:
+        return normalized
+    if _model_is_provider_qualified(normalized):
+        return normalized
+    if not provider:
+        return normalized
+    return f"{provider}/{normalized}"
+
+
+def _normalize_base_url_for_provider(base_url: str, provider: str) -> str:
+    normalized = _normalize_base_url(base_url)
+    if not normalized:
+        return normalized
+    if provider != "openai":
+        return normalized
+    if re.match(r"^https?://[^/]+$", normalized, flags=re.I):
+        return f"{normalized}/v1"
+    return normalized
+
+
+def running_in_container() -> bool:
+    return os.path.exists("/.dockerenv") or os.path.exists("/run/.containerenv")
+
+
+def rewrite_localhost_for_container(base_url: str) -> str:
+    import urllib.parse
+
+    normalized = base_url.strip()
+    if not normalized:
+        return normalized
+    try:
+        parsed = urllib.parse.urlparse(normalized)
+    except Exception:
+        return normalized
+    host = (parsed.hostname or "").lower()
+    if host not in {"localhost", "127.0.0.1", "::1"}:
+        return normalized
+    user_info = ""
+    if parsed.username:
+        user_info = parsed.username
+        if parsed.password:
+            user_info += f":{parsed.password}"
+        user_info += "@"
+    netloc = f"{user_info}host.docker.internal"
+    if parsed.port:
+        netloc += f":{parsed.port}"
+    rewritten = urllib.parse.urlunparse(
+        (parsed.scheme, netloc, parsed.path, parsed.params, parsed.query, parsed.fragment)
+    )
+    return rewritten or normalized
+
+
+def looks_local_base_url(base_url: str) -> bool:
+    if not base_url:
+        return False
+    lowered = base_url.lower()
+    return "localhost" in lowered or "127.0.0.1" in lowered or "host.docker.internal" in lowered
+
+
+def resolve_llm_config(
+    default_model: str = "local-model",
+    logger: Optional[Logger] = None,
+) -> Tuple[str, str, str]:
+    """Returns (model, api_key, base_url) resolved from config + env."""
+    log = logger or Logger("[Executor]")
+    raw_model = setting_str("WORKERPALS_LLM_MODEL", "workerpals.llm.model", "")
+    api_key = setting_str("WORKERPALS_LLM_API_KEY", "workerpals.llm.api_key", "")
+    raw_base_url = setting_str("WORKERPALS_LLM_ENDPOINT", "workerpals.llm.endpoint", "")
+    provider = infer_litellm_provider(raw_base_url)
+    configured_model = _normalize_litellm_model(raw_model or default_model, provider)
+    base_url = _normalize_base_url_for_provider(raw_base_url, provider)
+    if running_in_container():
+        rewritten = rewrite_localhost_for_container(base_url)
+        if rewritten != base_url:
+            log.info(f"Rewriting local LLM base URL for container networking: {base_url} -> {rewritten}")
+            base_url = rewritten
+    if not raw_model.strip():
+        log.info(f"No explicit model configured; using default model {default_model}.")
+    return configured_model, api_key, base_url
+
+
+# ─── Git helpers ─────────────────────────────────────────────────────────────
+
+def summarize_git_changes(repo: str) -> List[str]:
+    try:
+        proc = subprocess.run(
+            ["git", "status", "--porcelain"],
+            cwd=repo,
+            capture_output=True,
+            text=True,
+            timeout=20,
+            check=False,
+        )
+        if proc.returncode != 0:
+            return []
+        paths: List[str] = []
+        for raw_line in proc.stdout.splitlines():
+            line = str(raw_line or "").rstrip("\r\n")
+            if not line.strip():
+                continue
+            # Porcelain format uses two status columns + space prefix.
+            # Do not trim leading whitespace before slicing, otherwise
+            # paths like "README.md" become "EADME.md".
+            if len(line) < 4:
+                continue
+            path = line[3:].strip()
+            if " -> " in path:
+                path = path.split(" -> ", 1)[1]
+            if path:
+                paths.append(path)
+        return paths
+    except Exception:
+        return []
+
+
+def log_git_status(repo: str, logger: Optional[Logger] = None) -> None:
+    """Log ``git status --porcelain`` and ``git diff --stat`` for post-execution visibility."""
+    log = logger or Logger("[Executor]")
+    try:
+        status = subprocess.run(
+            ["git", "status", "--porcelain"],
+            cwd=repo, capture_output=True, text=True, timeout=10, check=False,
+        )
+        lines = [l for l in (status.stdout or "").splitlines() if l.strip()]
+        if lines:
+            log.debug("Git status after execution:")
+            for line in lines[:30]:
+                log.debug(f"  {line}")
+        else:
+            log.debug("Git status: clean (no changes)")
+    except Exception as exc:
+        log.debug(f"Git status failed: {exc}")
+
+    try:
+        diff = subprocess.run(
+            ["git", "diff", "--stat"],
+            cwd=repo, capture_output=True, text=True, timeout=10, check=False,
+        )
+        diff_lines = [l for l in (diff.stdout or "").splitlines() if l.strip()]
+        if diff_lines:
+            log.debug("Git diff stat:")
+            for line in diff_lines[:20]:
+                log.debug(f"  {line}")
+    except Exception:
+        pass
+
+
+def log_agent_messages(messages: list, logger: Optional[Logger] = None, max_chars: int = 200) -> None:
+    """Log a summary of agent message history (works with miniswe's message format).
+
+    Only emits output when WORKERPALS_DEBUG=1.
+    """
+    if not _debug_enabled():
+        return
+    log = logger or Logger("[Executor]")
+    step = 0
+    for msg in messages:
+        if not isinstance(msg, dict):
+            continue
+        role = str(msg.get("role") or "").strip()
+        if not role:
+            continue
+
+        step += 1
+        content = str(msg.get("content") or "").strip()
+
+        # Tool calls (assistant requesting a tool)
+        tool_calls = msg.get("tool_calls") or []
+        if tool_calls and isinstance(tool_calls, list):
+            for tc in tool_calls:
+                if isinstance(tc, dict):
+                    fn = tc.get("function") or {}
+                    name = fn.get("name") or "unknown"
+                    args_raw = str(fn.get("arguments") or "")[:80]
+                    log.debug(f"Step {step} (tool_call): {name}({args_raw})")
+            continue
+
+        # Tool response
+        if role == "tool":
+            name = str(msg.get("name") or msg.get("tool_call_id") or "tool")
+            excerpt = to_single_line(content, max_chars)
+            log.debug(f"Step {step} (tool_result/{name}): {excerpt}")
+            continue
+
+        # Assistant or user text
+        excerpt = to_single_line(content, max_chars)
+        if excerpt:
+            log.debug(f"Step {step} ({role}): {excerpt}")
+
+
+# ─── Payload parsing ────────────────────────────────────────────────────────
+
+@dataclass
+class TaskExecutePayload:
+    """Validated fields extracted from the base64 job payload."""
+    kind: str
+    params: Dict[str, Any]
+    repo: str
+    instruction: str
+    supplemental_guidance: List[str] = field(default_factory=list)
+    payload: Dict[str, Any] = field(default_factory=dict)
+
+
+def _is_non_actionable_planner_guidance(text: str) -> bool:
+    lower = str(text or "").strip().lower()
+    if not lower:
+        return True
+    blocked_markers = (
+        "no worker instruction needed",
+        "no additional instruction needed",
+        "purely documentation update",
+        "already updated",
+        "nothing to do",
+    )
+    return any(marker in lower for marker in blocked_markers)
+
+
+def parse_task_execute_payload(
+    argv: List[str],
+    *,
+    accepted_kinds: Tuple[str, ...] = ("task.execute",),
+    logger: Optional[Logger] = None,
+) -> TaskExecutePayload:
+    """Decode argv[1], validate required fields, return structured payload.
+
+    Raises ``SystemExit`` via ``fail()`` on validation errors so callers
+    don't need to handle them.
+    """
+    log = logger or Logger("[Executor]")
+    if len(argv) < 2:
+        raise SystemExit(fail("Missing base64 job payload", exit_code=2))
+
+    try:
+        payload = decode_payload(argv[1])
+    except Exception as exc:
+        raise SystemExit(fail(f"Failed to decode job payload: {exc}", exit_code=2))
+
+    kind = payload.get("kind")
+    params = payload.get("params", {})
+    repo_raw = payload.get("repo")
+
+    if not isinstance(kind, str) or not kind:
+        raise SystemExit(fail("Invalid payload: missing 'kind'", exit_code=2))
+    if not isinstance(params, dict):
+        raise SystemExit(fail("Invalid payload: 'params' must be an object", exit_code=2))
+    if not isinstance(repo_raw, str) or not repo_raw:
+        raise SystemExit(fail("Invalid payload: missing 'repo'", exit_code=2))
+    repo, repo_error = resolve_repo_within_assigned_root(repo_raw)
+    if repo_error or not repo:
+        raise SystemExit(fail(repo_error or "Invalid payload repo path", exit_code=2))
+
+    if kind not in accepted_kinds:
+        kinds_str = ", ".join(accepted_kinds)
+        raise SystemExit(
+            fail(f"Unsupported job kind '{kind}'. Accepted: {kinds_str}.", exit_code=2)
+        )
+
+    instruction = str(params.get("instruction") or "").strip()
+    if not instruction:
+        raise SystemExit(fail("task.execute requires 'instruction'", exit_code=2))
+
+    planner_instruction = str(params.get("plannerWorkerInstruction") or "").strip()
+    quality_revision_hint = str(params.get("qualityRevisionHint") or "").strip()
+
+    supplemental_guidance: List[str] = []
+    if planner_instruction and planner_instruction != instruction:
+        if _is_non_actionable_planner_guidance(planner_instruction):
+            log.info(
+                "Planner guidance was provided but ignored due to "
+                "non-actionable placeholder content."
+            )
+        else:
+            log.info(
+                "Planner guidance was provided, but preserving original "
+                "user instruction as canonical task input."
+            )
+            supplemental_guidance.append(planner_instruction)
+    if quality_revision_hint:
+        log.info(
+            "Quality revision guidance provided for this attempt; "
+            "preserving canonical user instruction and applying additive guidance."
+        )
+        supplemental_guidance.append(quality_revision_hint)
+
+    return TaskExecutePayload(
+        kind=kind,
+        params=params,
+        repo=repo,
+        instruction=instruction,
+        supplemental_guidance=supplemental_guidance,
+        payload=payload,
+    )