codex-autorunner 1.1.0__py3-none-any.whl → 1.2.0__py3-none-any.whl

codex_autorunner/core/pma_context.py

@@ -0,0 +1,496 @@
+from __future__ import annotations
+
+import asyncio
+import json
+from pathlib import Path
+from typing import Any, Optional
+
+from ..bootstrap import ensure_pma_docs
+from ..tickets.files import list_ticket_paths, safe_relpath, ticket_is_done
+from ..tickets.outbox import parse_dispatch, resolve_outbox_paths
+from .config import load_hub_config, load_repo_config
+from .flows.models import FlowRunStatus
+from .flows.store import FlowStore
+from .hub import HubSupervisor
+from .state_roots import resolve_hub_templates_root
+
+PMA_MAX_REPOS = 25
+PMA_MAX_MESSAGES = 10
+PMA_MAX_TEXT = 800
+PMA_MAX_TEMPLATE_REPOS = 25
+PMA_MAX_TEMPLATE_FIELD_CHARS = 120
+
+# Defaults used when hub config is not available (should be rare).
+PMA_DOCS_MAX_CHARS = 12_000
+PMA_ACTIVE_CONTEXT_MAX_LINES = 200
+PMA_CONTEXT_LOG_TAIL_LINES = 120
+
+
+def _tail_lines(text: str, max_lines: int) -> str:
+    if max_lines <= 0:
+        return ""
+    lines = (text or "").splitlines()
+    if len(lines) <= max_lines:
+        return "\n".join(lines)
+    return "\n".join(lines[-max_lines:])
+
+
+def load_pma_workspace_docs(hub_root: Path) -> dict[str, Any]:
+    """Load hub-level PMA workspace docs for prompt injection.
+
+    These docs act as durable memory and working context for PMA.
+    """
+    try:
+        ensure_pma_docs(hub_root)
+    except Exception:
+        pass
+
+    docs_max_chars = PMA_DOCS_MAX_CHARS
+    active_context_max_lines = PMA_ACTIVE_CONTEXT_MAX_LINES
+    context_log_tail_lines = PMA_CONTEXT_LOG_TAIL_LINES
+    try:
+        hub_config = load_hub_config(hub_root)
+        pma_cfg = getattr(hub_config, "pma", None)
+        if pma_cfg is not None:
+            docs_max_chars = int(getattr(pma_cfg, "docs_max_chars", docs_max_chars))
+            active_context_max_lines = int(
+                getattr(pma_cfg, "active_context_max_lines", active_context_max_lines)
+            )
+            context_log_tail_lines = int(
+                getattr(pma_cfg, "context_log_tail_lines", context_log_tail_lines)
+            )
+    except Exception:
+        pass
+
+    pma_dir = hub_root / ".codex-autorunner" / "pma"
+    agents_path = pma_dir / "AGENTS.md"
+    active_context_path = pma_dir / "active_context.md"
+    context_log_path = pma_dir / "context_log.md"
+
+    def _read(path: Path) -> str:
+        try:
+            return path.read_text(encoding="utf-8")
+        except Exception:
+            return ""
+
+    agents = _truncate(_read(agents_path), docs_max_chars)
+    active_context = _read(active_context_path)
+    active_context_lines = len((active_context or "").splitlines())
+    active_context = _truncate(active_context, docs_max_chars)
+    context_log_tail = _tail_lines(_read(context_log_path), context_log_tail_lines)
+    context_log_tail = _truncate(context_log_tail, docs_max_chars)
+
+    return {
+        "agents": agents,
+        "active_context": active_context,
+        "active_context_line_count": active_context_lines,
+        "active_context_max_lines": active_context_max_lines,
+        "context_log_tail": context_log_tail,
+    }
+
+
+def _truncate(text: Optional[str], limit: int) -> str:
+    raw = text or ""
+    if len(raw) <= limit:
+        return raw
+    return raw[: max(0, limit - 3)] + "..."
+
+
+def _trim_extra(extra: Any, limit: int) -> Any:
+    if extra is None:
+        return None
+    if isinstance(extra, str):
+        return _truncate(extra, limit)
+    try:
+        raw = json.dumps(extra, ensure_ascii=True, sort_keys=True, default=str)
+    except Exception:
+        raw = str(extra)
+    if len(raw) <= limit:
+        return extra
+    return {
+        "_omitted": True,
+        "note": "extra omitted due to size",
+        "preview": _truncate(raw, limit),
+    }
+
+
+def _load_template_scan_summary(
+    hub_root: Optional[Path],
+    *,
+    max_field_chars: int = PMA_MAX_TEMPLATE_FIELD_CHARS,
+) -> Optional[dict[str, Any]]:
+    if hub_root is None:
+        return None
+    try:
+        scans_root = resolve_hub_templates_root(hub_root) / "scans"
+        if not scans_root.exists():
+            return None
+        candidates = [
+            entry
+            for entry in scans_root.iterdir()
+            if entry.is_file() and entry.suffix == ".json"
+        ]
+        if not candidates:
+            return None
+        newest = max(candidates, key=lambda entry: entry.stat().st_mtime)
+        payload = json.loads(newest.read_text(encoding="utf-8"))
+        if not isinstance(payload, dict):
+            return None
+        return {
+            "repo_id": _truncate(str(payload.get("repo_id", "")), max_field_chars),
+            "decision": _truncate(str(payload.get("decision", "")), max_field_chars),
+            "severity": _truncate(str(payload.get("severity", "")), max_field_chars),
+            "scanned_at": _truncate(
+                str(payload.get("scanned_at", "")), max_field_chars
+            ),
+        }
+    except Exception:
+        return None
+
+
+def _build_templates_snapshot(
+    supervisor: HubSupervisor,
+    *,
+    hub_root: Optional[Path] = None,
+    max_repos: int = PMA_MAX_TEMPLATE_REPOS,
+    max_field_chars: int = PMA_MAX_TEMPLATE_FIELD_CHARS,
+) -> dict[str, Any]:
+    hub_config = getattr(supervisor, "hub_config", None)
+    templates_cfg = getattr(hub_config, "templates", None)
+    if templates_cfg is None:
+        return {"enabled": False, "repos": []}
+    repos = []
+    for repo in templates_cfg.repos[: max(0, max_repos)]:
+        repos.append(
+            {
+                "id": _truncate(repo.id, max_field_chars),
+                "default_ref": _truncate(repo.default_ref, max_field_chars),
+                "trusted": bool(repo.trusted),
+            }
+        )
+    payload: dict[str, Any] = {
+        "enabled": bool(templates_cfg.enabled),
+        "repos": repos,
+    }
+    scan_summary = _load_template_scan_summary(
+        hub_root, max_field_chars=max_field_chars
+    )
+    if scan_summary:
+        payload["last_scan"] = scan_summary
+    return payload
+
+
+def load_pma_prompt(hub_root: Path) -> str:
+    path = hub_root / ".codex-autorunner" / "pma" / "prompt.md"
+    try:
+        ensure_pma_docs(hub_root)
+    except Exception:
+        pass
+    try:
+        return path.read_text(encoding="utf-8")
+    except Exception:
+        return ""
+
+
+def format_pma_prompt(
+    base_prompt: str,
+    snapshot: dict[str, Any],
+    message: str,
+    hub_root: Optional[Path] = None,
+) -> str:
+    snapshot_text = json.dumps(snapshot, sort_keys=True)
+
+    pma_docs: Optional[dict[str, Any]] = None
+    if hub_root is not None:
+        try:
+            pma_docs = load_pma_workspace_docs(hub_root)
+        except Exception:
+            pma_docs = None
+
+    prompt = f"{base_prompt}\n\n"
+    prompt += (
+        "Ops guide: `.codex-autorunner/pma/ABOUT_CAR.md`.\n"
+        "Durable guidance: `.codex-autorunner/pma/AGENTS.md`.\n"
+        "Working context: `.codex-autorunner/pma/active_context.md`.\n"
+        "History: `.codex-autorunner/pma/context_log.md`.\n"
+        "To send a file to the user, write it to `.codex-autorunner/pma/outbox/`.\n"
+        "User uploaded files are in `.codex-autorunner/pma/inbox/`.\n\n"
+    )
+
+    if pma_docs:
+        max_lines = pma_docs.get("active_context_max_lines")
+        line_count = pma_docs.get("active_context_line_count")
+        prompt += (
+            "<pma_workspace_docs>\n"
+            "<AGENTS_MD>\n"
+            f"{pma_docs.get('agents', '')}\n"
+            "</AGENTS_MD>\n"
+            "<ACTIVE_CONTEXT_MD>\n"
+            f"{pma_docs.get('active_context', '')}\n"
+            "</ACTIVE_CONTEXT_MD>\n"
+            f"<ACTIVE_CONTEXT_BUDGET lines='{max_lines}' current_lines='{line_count}' />\n"
+            "<CONTEXT_LOG_TAIL_MD>\n"
+            f"{pma_docs.get('context_log_tail', '')}\n"
+            "</CONTEXT_LOG_TAIL_MD>\n"
+            "</pma_workspace_docs>\n\n"
+        )
+
+    prompt += (
+        "<hub_snapshot>\n"
+        f"{snapshot_text}\n"
+        "</hub_snapshot>\n\n"
+        "<user_message>\n"
+        f"{message}\n"
+        "</user_message>\n"
+    )
+    return prompt
+
+
+def _get_ticket_flow_summary(repo_path: Path) -> Optional[dict[str, Any]]:
+    db_path = repo_path / ".codex-autorunner" / "flows.db"
+    if not db_path.exists():
+        return None
+    try:
+        config = load_repo_config(repo_path)
+        with FlowStore(db_path, durable=config.durable_writes) as store:
+            runs = store.list_flow_runs(flow_type="ticket_flow")
+            if not runs:
+                return None
+            latest = runs[0]
+
+            ticket_dir = repo_path / ".codex-autorunner" / "tickets"
+            total = 0
+            done = 0
+            for path in list_ticket_paths(ticket_dir):
+                total += 1
+                try:
+                    if ticket_is_done(path):
+                        done += 1
+                except Exception:
+                    continue
+
+            if total == 0:
+                return None
+
+            state = latest.state if isinstance(latest.state, dict) else {}
+            engine = state.get("ticket_engine") if isinstance(state, dict) else {}
+            engine = engine if isinstance(engine, dict) else {}
+            current_step = engine.get("total_turns")
+
+            return {
+                "status": latest.status.value,
+                "done_count": done,
+                "total_count": total,
+                "current_step": current_step,
+            }
+    except Exception:
+        return None
+
+
+def _latest_dispatch(
+    repo_root: Path, run_id: str, input_data: dict, *, max_text_chars: int
+) -> Optional[dict[str, Any]]:
+    try:
+        workspace_root = Path(input_data.get("workspace_root") or repo_root)
+        runs_dir = Path(input_data.get("runs_dir") or ".codex-autorunner/runs")
+        outbox_paths = resolve_outbox_paths(
+            workspace_root=workspace_root, runs_dir=runs_dir, run_id=run_id
+        )
+        history_dir = outbox_paths.dispatch_history_dir
+        if not history_dir.exists() or not history_dir.is_dir():
+            return None
+        seq_dirs: list[Path] = []
+        for child in history_dir.iterdir():
+            if not child.is_dir():
+                continue
+            name = child.name
+            if len(name) == 4 and name.isdigit():
+                seq_dirs.append(child)
+        if not seq_dirs:
+            return None
+        latest_dir = sorted(seq_dirs, key=lambda p: p.name)[-1]
+        seq = int(latest_dir.name)
+        dispatch_path = latest_dir / "DISPATCH.md"
+        dispatch, errors = parse_dispatch(dispatch_path)
+        if errors or dispatch is None:
+            return {
+                "seq": seq,
+                "dir": safe_relpath(latest_dir, repo_root),
+                "dispatch": None,
+                "errors": errors,
+                "files": [],
+            }
+        files: list[str] = []
+        for child in sorted(latest_dir.iterdir(), key=lambda p: p.name):
+            if child.name.startswith("."):
+                continue
+            if child.name == "DISPATCH.md":
+                continue
+            if child.is_file():
+                files.append(child.name)
+        dispatch_dict = {
+            "mode": dispatch.mode,
+            "title": _truncate(dispatch.title, max_text_chars),
+            "body": _truncate(dispatch.body, max_text_chars),
+            "extra": _trim_extra(dispatch.extra, max_text_chars),
+            "is_handoff": dispatch.is_handoff,
+        }
+        return {
+            "seq": seq,
+            "dir": safe_relpath(latest_dir, repo_root),
+            "dispatch": dispatch_dict,
+            "errors": [],
+            "files": files,
+        }
+    except Exception:
+        return None
+
+
+def _gather_inbox(
+    supervisor: HubSupervisor, *, max_text_chars: int
+) -> list[dict[str, Any]]:
+    messages: list[dict[str, Any]] = []
+    try:
+        snapshots = supervisor.list_repos()
+    except Exception:
+        return []
+    for snap in snapshots:
+        if not (snap.initialized and snap.exists_on_disk):
+            continue
+        repo_root = snap.path
+        db_path = repo_root / ".codex-autorunner" / "flows.db"
+        if not db_path.exists():
+            continue
+        try:
+            config = load_repo_config(repo_root)
+            with FlowStore(db_path, durable=config.durable_writes) as store:
+                paused = store.list_flow_runs(
+                    flow_type="ticket_flow", status=FlowRunStatus.PAUSED
+                )
+        except Exception:
+            continue
+        if not paused:
+            continue
+        for record in paused:
+            latest = _latest_dispatch(
+                repo_root,
+                str(record.id),
+                dict(record.input_data or {}),
+                max_text_chars=max_text_chars,
+            )
+            if not latest or not latest.get("dispatch"):
+                continue
+            messages.append(
+                {
+                    "repo_id": snap.id,
+                    "repo_display_name": snap.display_name,
+                    "run_id": record.id,
+                    "run_created_at": record.created_at,
+                    "seq": latest["seq"],
+                    "dispatch": latest["dispatch"],
+                    "files": latest.get("files") or [],
+                    "open_url": f"/repos/{snap.id}/?tab=inbox&run_id={record.id}",
+                }
+            )
+    messages.sort(key=lambda m: (m.get("run_created_at") or ""), reverse=True)
+    return messages
+
+
+def _gather_lifecycle_events(
+    supervisor: HubSupervisor, limit: int = 20
+) -> list[dict[str, Any]]:
+    events = supervisor.lifecycle_store.get_unprocessed(limit=limit)
+    result: list[dict[str, Any]] = []
+    for event in events[:limit]:
+        result.append(
+            {
+                "event_type": event.event_type.value,
+                "repo_id": event.repo_id,
+                "run_id": event.run_id,
+                "timestamp": event.timestamp,
+                "data": event.data,
+            }
+        )
+    return result
+
+
+async def build_hub_snapshot(
+    supervisor: Optional[HubSupervisor],
+    hub_root: Optional[Path] = None,
+) -> dict[str, Any]:
+    if supervisor is None:
+        return {
+            "repos": [],
+            "inbox": [],
+            "templates": {"enabled": False, "repos": []},
+            "lifecycle_events": [],
+        }
+
+    snapshots = await asyncio.to_thread(supervisor.list_repos)
+    snapshots = sorted(snapshots, key=lambda snap: snap.id)
+    pma_config = supervisor.hub_config.pma if supervisor else None
+    max_repos = (
+        pma_config.max_repos
+        if pma_config and pma_config.max_repos > 0
+        else PMA_MAX_REPOS
+    )
+    max_messages = (
+        pma_config.max_messages
+        if pma_config and pma_config.max_messages > 0
+        else PMA_MAX_MESSAGES
+    )
+    max_text_chars = (
+        pma_config.max_text_chars
+        if pma_config and pma_config.max_text_chars > 0
+        else PMA_MAX_TEXT
+    )
+    repos: list[dict[str, Any]] = []
+    for snap in snapshots[:max_repos]:
+        summary: dict[str, Any] = {
+            "id": snap.id,
+            "display_name": snap.display_name,
+            "status": snap.status.value,
+            "last_run_id": snap.last_run_id,
+            "last_run_started_at": snap.last_run_started_at,
+            "last_run_finished_at": snap.last_run_finished_at,
+            "last_exit_code": snap.last_exit_code,
+            "ticket_flow": None,
+        }
+        if snap.initialized and snap.exists_on_disk:
+            summary["ticket_flow"] = _get_ticket_flow_summary(snap.path)
+        repos.append(summary)
+
+    inbox = await asyncio.to_thread(
+        _gather_inbox, supervisor, max_text_chars=max_text_chars
+    )
+    inbox = inbox[:max_messages]
+
+    lifecycle_events = await asyncio.to_thread(
+        _gather_lifecycle_events, supervisor, limit=20
+    )
+
+    templates = _build_templates_snapshot(supervisor, hub_root=hub_root)
+
+    pma_files: dict[str, list[str]] = {"inbox": [], "outbox": []}
+    if hub_root:
+        try:
+            pma_dir = hub_root / ".codex-autorunner" / "pma"
+            for box in ["inbox", "outbox"]:
+                box_dir = pma_dir / box
+                if box_dir.exists():
+                    files = [
+                        f.name
+                        for f in box_dir.iterdir()
+                        if f.is_file() and not f.name.startswith(".")
+                    ]
+                    pma_files[box] = sorted(files)
+        except Exception:
+            pass
+
+    return {
+        "repos": repos,
+        "inbox": inbox,
+        "templates": templates,
+        "pma_files": pma_files,
+        "lifecycle_events": lifecycle_events,
+    }