codex-autorunner 0.1.2__py3-none-any.whl → 1.1.0__py3-none-any.whl

codex_autorunner/spec_ingest.py

@@ -1,812 +0,0 @@
-import asyncio
-import contextlib
-import difflib
-import logging
-import re
-import threading
-from contextlib import asynccontextmanager, contextmanager
-from dataclasses import dataclass, field
-from pathlib import Path
-from typing import Any, Dict, MutableMapping, Optional
-
-import httpx
-
-from .agents.opencode.runtime import (
-    PERMISSION_ALLOW,
-    OpenCodeTurnOutput,
-    build_turn_id,
-    collect_opencode_output,
-    extract_session_id,
-    opencode_missing_env,
-    parse_message_response,
-    split_model_id,
-)
-from .agents.opencode.supervisor import OpenCodeSupervisor
-from .core.app_server_events import AppServerEventBuffer
-from .core.app_server_prompts import (
-    build_spec_ingest_prompt as build_app_server_spec_ingest_prompt,
-)
-from .core.app_server_threads import (
-    AppServerThreadRegistry,
-    default_app_server_threads_path,
-)
-from .core.docs import validate_todo_markdown
-from .core.engine import Engine
-from .core.locks import FileLock, FileLockBusy, FileLockError
-from .core.patch_utils import (
-    PatchError,
-    apply_patch_file,
-    ensure_patch_targets_allowed,
-    normalize_patch_text,
-    preview_patch,
-)
-from .core.utils import atomic_write
-from .integrations.app_server.client import CodexAppServerError
-from .integrations.app_server.supervisor import WorkspaceAppServerSupervisor
-
-logger = logging.getLogger("codex_autorunner.spec_ingest")
-
-SPEC_INGEST_TIMEOUT_SECONDS = 240
-SPEC_INGEST_INTERRUPT_GRACE_SECONDS = 10
-SPEC_INGEST_PATCH_NAME = "spec-ingest.patch"
-
-
-class SpecIngestError(Exception):
-    """Raised when ingesting a SPEC fails."""
-
-
-@dataclass
-class ActiveSpecIngestTurn:
-    thread_id: str
-    turn_id: str
-    client: Any
-    interrupted: bool = False
-    interrupt_sent: bool = False
-    interrupt_event: asyncio.Event = field(default_factory=asyncio.Event)
-
-
-def ensure_can_overwrite(engine: Engine, force: bool) -> None:
-    if force:
-        return
-    for key in ("todo", "progress", "opinions"):
-        existing = engine.docs.read_doc(key).strip()
-        if existing:
-            raise SpecIngestError(
-                "TODO/PROGRESS/OPINIONS already contain content; rerun with --force to overwrite"
-            )
-
-
-def clear_work_docs(engine: Engine) -> Dict[str, str]:
-    defaults = {
-        "todo": "# TODO\n\n",
-        "progress": "# Progress\n\n",
-        "opinions": "# Opinions\n\n",
-    }
-    for key, content in defaults.items():
-        atomic_write(engine.config.doc_path(key), content)
-    # Read back to reflect actual on-disk content.
-    return {k: engine.docs.read_doc(k) for k in defaults.keys()}
-
-
-class SpecIngestService:
-    def __init__(
-        self,
-        engine: Engine,
-        *,
-        app_server_supervisor: Optional[WorkspaceAppServerSupervisor] = None,
-        app_server_threads: Optional[AppServerThreadRegistry] = None,
-        app_server_events: Optional[AppServerEventBuffer] = None,
-        opencode_supervisor: Optional[OpenCodeSupervisor] = None,
-        env: Optional[MutableMapping[str, str]] = None,
-    ) -> None:
-        self.engine = engine
-        self._env = env
-        self._app_server_supervisor = app_server_supervisor
-        self._app_server_threads = app_server_threads or AppServerThreadRegistry(
-            default_app_server_threads_path(self.engine.repo_root)
-        )
-        self._app_server_events = app_server_events
-        self._opencode_supervisor = opencode_supervisor
-        self.patch_path = (
-            self.engine.repo_root / ".codex-autorunner" / SPEC_INGEST_PATCH_NAME
-        )
-        self.last_agent_message: Optional[str] = None
-        self._lock: Optional[asyncio.Lock] = None
-        self._lock_path = (
-            self.engine.repo_root / ".codex-autorunner" / "locks" / "spec_ingest.lock"
-        )
-        self._thread_lock = threading.Lock()
-        self._active_turn: Optional[ActiveSpecIngestTurn] = None
-        self._active_turn_lock = threading.Lock()
-        self._pending_interrupt = False
-
-    def _ensure_lock(self) -> asyncio.Lock:
-        if self._lock is None:
-            try:
-                self._lock = asyncio.Lock()
-            except RuntimeError:
-                asyncio.set_event_loop(asyncio.new_event_loop())
-                self._lock = asyncio.Lock()
-        return self._lock
-
-    def _ingest_busy(self) -> bool:
-        lock = self._ensure_lock()
-        if lock.locked():
-            return True
-        file_lock = FileLock(self._lock_path)
-        try:
-            file_lock.acquire(blocking=False)
-        except FileLockBusy:
-            return True
-        except FileLockError:
-            return True
-        finally:
-            file_lock.release()
-        return False
-
-    @asynccontextmanager
-    async def ingest_lock(self):
-        if not self._thread_lock.acquire(blocking=False):
-            raise SpecIngestError("Spec ingest is already running")
-        lock = self._ensure_lock()
-        if lock.locked():
-            self._thread_lock.release()
-            raise SpecIngestError("Spec ingest is already running")
-        await lock.acquire()
-        file_lock = FileLock(self._lock_path)
-        try:
-            try:
-                file_lock.acquire(blocking=False)
-            except FileLockBusy as exc:
-                raise SpecIngestError("Spec ingest is already running") from exc
-            except FileLockError as exc:
-                raise SpecIngestError(str(exc)) from exc
-            yield
-        finally:
-            file_lock.release()
-            lock.release()
-            self._thread_lock.release()
-            with self._active_turn_lock:
-                self._pending_interrupt = False
-
-    @contextmanager
-    def _patch_lock(self):
-        if not self._thread_lock.acquire(blocking=False):
-            raise SpecIngestError("Spec ingest is already running")
-        lock = self._ensure_lock()
-        if lock.locked():
-            self._thread_lock.release()
-            raise SpecIngestError("Spec ingest is already running")
-        file_lock = FileLock(self._lock_path)
-        try:
-            file_lock.acquire(blocking=False)
-        except FileLockBusy as exc:
-            self._thread_lock.release()
-            raise SpecIngestError("Spec ingest is already running") from exc
-        except FileLockError as exc:
-            self._thread_lock.release()
-            raise SpecIngestError(str(exc)) from exc
-        try:
-            yield
-        finally:
-            file_lock.release()
-            self._thread_lock.release()
-
-    def _ensure_app_server(self) -> WorkspaceAppServerSupervisor:
-        if self._app_server_supervisor is None:
-            raise SpecIngestError("App-server backend is not configured")
-        return self._app_server_supervisor
-
-    def _ensure_opencode(self) -> OpenCodeSupervisor:
-        if self._opencode_supervisor is None:
-            raise SpecIngestError("OpenCode backend is not configured")
-        return self._opencode_supervisor
-
-    def _get_active_turn(self) -> Optional[ActiveSpecIngestTurn]:
-        with self._active_turn_lock:
-            return self._active_turn
-
-    def _clear_active_turn(self, turn_id: str) -> None:
-        with self._active_turn_lock:
-            if self._active_turn and self._active_turn.turn_id == turn_id:
-                self._active_turn = None
-
-    def _register_active_turn(
-        self, client: Any, turn_id: str, thread_id: str
-    ) -> ActiveSpecIngestTurn:
-        interrupt_event = asyncio.Event()
-        active = ActiveSpecIngestTurn(
-            thread_id=thread_id,
-            turn_id=turn_id,
-            client=client,
-            interrupted=False,
-            interrupt_sent=False,
-            interrupt_event=interrupt_event,
-        )
-        with self._active_turn_lock:
-            self._active_turn = active
-            if self._pending_interrupt:
-                self._pending_interrupt = False
-                active.interrupted = True
-                interrupt_event.set()
-        return active
-
-    async def _interrupt_turn(self, active: ActiveSpecIngestTurn) -> None:
-        if active.interrupt_sent:
-            return
-        active.interrupt_sent = True
-        try:
-            if not hasattr(active.client, "turn_interrupt"):
-                return
-            await asyncio.wait_for(
-                active.client.turn_interrupt(
-                    active.turn_id, thread_id=active.thread_id
-                ),
-                timeout=SPEC_INGEST_INTERRUPT_GRACE_SECONDS,
-            )
-        except asyncio.TimeoutError:
-            pass
-        except CodexAppServerError:
-            pass
-
-    async def _abort_opencode(
-        self, active: ActiveSpecIngestTurn, thread_id: str
-    ) -> None:
-        if active.interrupt_sent:
-            return
-        active.interrupt_sent = True
-        try:
-            if not hasattr(active.client, "abort"):
-                return
-            await asyncio.wait_for(
-                active.client.abort(thread_id),
-                timeout=SPEC_INGEST_INTERRUPT_GRACE_SECONDS,
-            )
-        except asyncio.TimeoutError:
-            pass
-        except (OSError, RuntimeError, httpx.HTTPError) as exc:
-            logger.debug("Failed to abort spec ingest turn: %s", exc)
-
-    async def interrupt(self) -> Dict[str, str]:
-        active = self._get_active_turn()
-        if active is None:
-            pending = self._ingest_busy()
-            with self._active_turn_lock:
-                self._pending_interrupt = pending
-            return self._assemble_response(
-                {},
-                status="interrupted",
-                agent_message="Spec ingest interrupted",
-            )
-        active.interrupted = True
-        active.interrupt_event.set()
-        await self._interrupt_turn(active)
-        return self._assemble_response(
-            {},
-            status="interrupted",
-            agent_message="Spec ingest interrupted",
-        )
-
-    def _allowed_targets(self) -> Dict[str, str]:
-        config = self.engine.config
-        rel = {}
-        for key in ("todo", "progress", "opinions"):
-            rel[key] = str(config.doc_path(key).relative_to(self.engine.repo_root))
-        return rel
-
-    def _spec_path(self, spec_path: Optional[Path]) -> Path:
-        target = spec_path or self.engine.config.doc_path("spec")
-        if not target.exists():
-            raise SpecIngestError(f"SPEC not found at {target}")
-        text = target.read_text(encoding="utf-8")
-        if not text.strip():
-            raise SpecIngestError(f"SPEC at {target} is empty")
-        return target
-
-    def _assemble_response(
-        self,
-        docs: Dict[str, str],
-        *,
-        patch: Optional[str] = None,
-        agent_message: Optional[str] = None,
-        status: str = "ok",
-    ) -> Dict[str, str]:
-        return {
-            "status": status,
-            "todo": docs.get("todo", self.engine.docs.read_doc("todo")),
-            "progress": docs.get("progress", self.engine.docs.read_doc("progress")),
-            "opinions": docs.get("opinions", self.engine.docs.read_doc("opinions")),
-            "spec": self.engine.docs.read_doc("spec"),
-            "summary": self.engine.docs.read_doc("summary"),
-            "patch": patch or "",
-            "agent_message": agent_message or "",
-        }
-
-    def pending_patch(self) -> Optional[Dict[str, str]]:
-        with self._patch_lock():
-            if not self.patch_path.exists():
-                return None
-            patch_text_raw = self.patch_path.read_text(encoding="utf-8")
-            targets = self._allowed_targets()
-            try:
-                patch_text, raw_targets = normalize_patch_text(patch_text_raw)
-                ensure_patch_targets_allowed(raw_targets, targets.values())
-                preview = preview_patch(self.engine.repo_root, patch_text, raw_targets)
-            except PatchError:
-                return None
-            docs = {
-                key: preview.get(path, self.engine.docs.read_doc(key))
-                for key, path in targets.items()
-            }
-            return self._assemble_response(
-                docs, patch=patch_text, agent_message=self.last_agent_message
-            )
-
-    def apply_patch(self) -> Dict[str, str]:
-        with self._patch_lock():
-            if not self.patch_path.exists():
-                raise SpecIngestError("No pending spec ingest patch")
-            patch_text_raw = self.patch_path.read_text(encoding="utf-8")
-            targets = self._allowed_targets()
-            try:
-                patch_text, raw_targets = normalize_patch_text(patch_text_raw)
-                ensure_patch_targets_allowed(raw_targets, targets.values())
-                self.patch_path.write_text(patch_text, encoding="utf-8")
-                apply_patch_file(self.engine.repo_root, self.patch_path, raw_targets)
-            except PatchError as exc:
-                raise SpecIngestError(str(exc)) from exc
-            self.patch_path.unlink(missing_ok=True)
-            return self._assemble_response(
-                {
-                    key: self.engine.docs.read_doc(key)
-                    for key in ("todo", "progress", "opinions")
-                }
-            )
-
-    def discard_patch(self) -> Dict[str, str]:
-        with self._patch_lock():
-            if self.patch_path.exists():
-                self.patch_path.unlink(missing_ok=True)
-            return self._assemble_response(
-                {
-                    key: self.engine.docs.read_doc(key)
-                    for key in ("todo", "progress", "opinions")
-                }
-            )
-
-    def _build_patch(self, rel_path: str, before: str, after: str) -> str:
-        diff = difflib.unified_diff(
-            before.splitlines(),
-            after.splitlines(),
-            fromfile=f"a/{rel_path}",
-            tofile=f"b/{rel_path}",
-            lineterm="",
-        )
-        return "\n".join(diff)
-
-    def _restore_docs(self, backups: Dict[str, str]) -> None:
-        config = self.engine.config
-        for key, content in backups.items():
-            path = config.doc_path(key)
-            try:
-                current = path.read_text(encoding="utf-8")
-            except OSError:
-                current = ""
-            if current != content:
-                atomic_write(path, content)
-
-    async def _execute_app_server(
-        self,
-        *,
-        force: bool,
-        spec_path: Optional[Path],
-        message: Optional[str],
-        model: Optional[str] = None,
-        reasoning: Optional[str] = None,
-    ) -> Dict[str, str]:
-        if not force:
-            ensure_can_overwrite(self.engine, force=False)
-        spec_target = self._spec_path(spec_path)
-        prompt = build_app_server_spec_ingest_prompt(
-            self.engine.config,
-            message=message or "Ingest SPEC into TODO/PROGRESS/OPINIONS.",
-            spec_path=spec_target,
-        )
-
-        # Backup docs
-        backups = {}
-        for key in ("todo", "progress", "opinions"):
-            backups[key] = self.engine.docs.read_doc(key)
-
-        supervisor = self._ensure_app_server()
-        client = await supervisor.get_client(self.engine.repo_root)
-        key = "spec_ingest"
-        thread_id = self._app_server_threads.get_thread_id(key)
-        if thread_id:
-            try:
-                result = await client.thread_resume(thread_id)
-                resumed = result.get("id")
-                if isinstance(resumed, str) and resumed:
-                    thread_id = resumed
-                    self._app_server_threads.set_thread_id(key, thread_id)
-            except CodexAppServerError:
-                self._app_server_threads.reset_thread(key)
-                thread_id = None
-        if not thread_id:
-            thread = await client.thread_start(str(self.engine.repo_root))
-            thread_id = thread.get("id")
-            if not isinstance(thread_id, str) or not thread_id:
-                raise SpecIngestError("App-server did not return a thread id")
-            self._app_server_threads.set_thread_id(key, thread_id)
-
-        turn_kwargs: dict[str, Any] = {}
-        if model:
-            turn_kwargs["model"] = model
-        if reasoning:
-            turn_kwargs["effort"] = reasoning
-        handle = await client.turn_start(
-            thread_id,
-            prompt,
-            approval_policy="never",
-            sandbox_policy="dangerFullAccess",  # Allowed for doc edits per user request
-            **turn_kwargs,
-        )
-        active = self._register_active_turn(client, handle.turn_id, handle.thread_id)
-        if self._app_server_events is not None:
-            try:
-                await self._app_server_events.register_turn(
-                    handle.thread_id, handle.turn_id
-                )
-            except (KeyError, TypeError, RuntimeError) as exc:
-                logger.debug("Failed to register turn: %s", exc)
-
-        turn_task = asyncio.create_task(handle.wait(timeout=None))
-        timeout_task = asyncio.create_task(asyncio.sleep(SPEC_INGEST_TIMEOUT_SECONDS))
-        interrupt_task = asyncio.create_task(active.interrupt_event.wait())
-
-        try:
-            tasks = {turn_task, timeout_task, interrupt_task}
-            done, _pending = await asyncio.wait(
-                tasks, return_when=asyncio.FIRST_COMPLETED
-            )
-            if timeout_task in done:
-                turn_task.add_done_callback(lambda task: task.exception())
-                raise SpecIngestError("Spec ingest agent timed out")
-            if interrupt_task in done:
-                active.interrupted = True
-                await self._interrupt_turn(active)
-                done, _pending = await asyncio.wait(
-                    {turn_task}, timeout=SPEC_INGEST_INTERRUPT_GRACE_SECONDS
-                )
-                if not done:
-                    turn_task.add_done_callback(lambda task: task.exception())
-                    return self._assemble_response(
-                        {},
-                        status="interrupted",
-                        agent_message="Spec ingest interrupted",
-                    )
-            result = await turn_task
-        finally:
-            self._clear_active_turn(handle.turn_id)
-            timeout_task.cancel()
-            with contextlib.suppress(asyncio.CancelledError):
-                await timeout_task
-            interrupt_task.cancel()
-            with contextlib.suppress(asyncio.CancelledError):
-                await interrupt_task
-
-        if active.interrupted:
-            # Restore docs if interrupted
-            self._restore_docs(backups)
-            return self._assemble_response(
-                {},
-                status="interrupted",
-                agent_message="Spec ingest interrupted",
-            )
-
-        if result.errors:
-            # Restore docs on error
-            self._restore_docs(backups)
-            raise SpecIngestError(result.errors[-1])
-
-        output = "\n".join(result.agent_messages).strip()
-        agent_message = SpecIngestPatchParser.parse_agent_message(output)
-
-        # Compute patch from file changes
-        patches = []
-        docs_preview = {}
-        targets = self._allowed_targets()
-
-        for key in ("todo", "progress", "opinions"):
-            path = self.engine.config.doc_path(key)
-            try:
-                after = path.read_text(encoding="utf-8")
-            except OSError:
-                after = ""
-            before = backups.get(key, "")
-            docs_preview[key] = after
-
-            if after == before:
-                continue
-
-            rel_path = targets[key]
-            patch = self._build_patch(rel_path, before, after)
-            if patch.strip():
-                patches.append(patch)
-
-        todo_errors = validate_todo_markdown(docs_preview.get("todo", ""))
-        if todo_errors:
-            # Restore docs before failing.
-            self._restore_docs(backups)
-            raise SpecIngestError("Invalid TODO format: " + "; ".join(todo_errors))
-
-        # Always restore docs to state before ingest (user must apply patch)
-        self._restore_docs(backups)
-
-        patch_text = "\n".join(patches)
-        if not patch_text.strip():
-            raise SpecIngestError(
-                "App-server did not make any changes to TODO/PROGRESS/OPINIONS"
-            )
-
-        self.patch_path.write_text(patch_text, encoding="utf-8")
-        self.last_agent_message = agent_message
-
-        return self._assemble_response(
-            docs_preview, patch=patch_text, agent_message=agent_message
-        )
-
-    async def _execute_opencode(
-        self,
-        *,
-        force: bool,
-        spec_path: Optional[Path],
-        message: Optional[str],
-        model: Optional[str],
-        reasoning: Optional[str],
-    ) -> Dict[str, str]:
-        if not force:
-            ensure_can_overwrite(self.engine, force=False)
-        spec_target = self._spec_path(spec_path)
-        prompt = build_app_server_spec_ingest_prompt(
-            self.engine.config,
-            message=message or "Ingest SPEC into TODO/PROGRESS/OPINIONS.",
-            spec_path=spec_target,
-        )
-        backups = {
-            key: self.engine.docs.read_doc(key)
-            for key in ("todo", "progress", "opinions")
-        }
-        supervisor = self._ensure_opencode()
-        client = await supervisor.get_client(self.engine.repo_root)
-        key = "spec_ingest.opencode"
-        thread_id = self._app_server_threads.get_thread_id(key)
-        if thread_id:
-            try:
-                await client.get_session(thread_id)
-            except (OSError, KeyError, ValueError, httpx.HTTPError) as exc:
-                logger.debug("OpenCode session not found, resetting thread: %s", exc)
-                self._app_server_threads.reset_thread(key)
-                thread_id = None
-        if not thread_id:
-            session = await client.create_session(directory=str(self.engine.repo_root))
-            thread_id = extract_session_id(session, allow_fallback_id=True)
-            if not isinstance(thread_id, str) or not thread_id:
-                raise SpecIngestError("OpenCode did not return a session id")
-            self._app_server_threads.set_thread_id(key, thread_id)
-
-        model_payload = split_model_id(model)
-        missing_env = await opencode_missing_env(
-            client, str(self.engine.repo_root), model_payload, env=self._env
-        )
-        if missing_env:
-            provider_id = model_payload.get("providerID") if model_payload else None
-            missing_label = ", ".join(missing_env)
-            raise SpecIngestError(
-                "OpenCode provider "
-                f"{provider_id or 'selected'} requires env vars: {missing_label}"
-            )
-        opencode_turn_started = False
-        await supervisor.mark_turn_started(self.engine.repo_root)
-        opencode_turn_started = True
-        turn_id = build_turn_id(thread_id)
-        active = self._register_active_turn(client, turn_id, thread_id)
-        permission_policy = PERMISSION_ALLOW
-        ready_event = asyncio.Event()
-        output_task = asyncio.create_task(
-            collect_opencode_output(
-                client,
-                session_id=thread_id,
-                workspace_path=str(self.engine.repo_root),
-                permission_policy=permission_policy,
-                question_policy="auto_first_option",
-                should_stop=active.interrupt_event.is_set,
-                ready_event=ready_event,
-            )
-        )
-        with contextlib.suppress(asyncio.TimeoutError):
-            await asyncio.wait_for(ready_event.wait(), timeout=2.0)
-        prompt_task = asyncio.create_task(
-            client.prompt_async(
-                thread_id,
-                message=prompt,
-                model=model_payload,
-                variant=reasoning,
-            )
-        )
-        timeout_task = asyncio.create_task(asyncio.sleep(SPEC_INGEST_TIMEOUT_SECONDS))
-        interrupt_task = asyncio.create_task(active.interrupt_event.wait())
-        try:
-            prompt_response = None
-            try:
-                prompt_response = await prompt_task
-            except Exception as exc:
-                active.interrupt_event.set()
-                output_task.cancel()
-                with contextlib.suppress(asyncio.CancelledError):
-                    await output_task
-                raise SpecIngestError(f"OpenCode prompt failed: {exc}") from exc
-            tasks = {output_task, timeout_task, interrupt_task}
-            done, _pending = await asyncio.wait(
-                tasks, return_when=asyncio.FIRST_COMPLETED
-            )
-            if timeout_task in done:
-                output_task.add_done_callback(lambda task: task.exception())
-                raise SpecIngestError("Spec ingest agent timed out")
-            if interrupt_task in done:
-                active.interrupted = True
-                await self._abort_opencode(active, thread_id)
-                done, _pending = await asyncio.wait(
-                    {output_task}, timeout=SPEC_INGEST_INTERRUPT_GRACE_SECONDS
-                )
-                if not done:
-                    output_task.add_done_callback(lambda task: task.exception())
-            output_result = await output_task
-            if output_result.text or output_result.error:
-                pass
-            elif prompt_response is not None:
-                fallback = parse_message_response(prompt_response)
-                if fallback.text:
-                    output_result = OpenCodeTurnOutput(
-                        text=fallback.text, error=fallback.error
-                    )
-        finally:
-            self._clear_active_turn(turn_id)
-            timeout_task.cancel()
-            with contextlib.suppress(asyncio.CancelledError):
-                await timeout_task
-            interrupt_task.cancel()
-            with contextlib.suppress(asyncio.CancelledError):
-                await interrupt_task
-            if opencode_turn_started:
-                await supervisor.mark_turn_finished(self.engine.repo_root)
-
-        if active.interrupted:
-            self._restore_docs(backups)
-            return self._assemble_response(
-                {},
-                status="interrupted",
-                agent_message="Spec ingest interrupted",
-            )
-
-        if output_result.error:
-            raise SpecIngestError(output_result.error)
-        agent_message = SpecIngestPatchParser.parse_agent_message(output_result.text)
-        patches = []
-        docs_preview = {}
-        targets = self._allowed_targets()
-
-        for key in ("todo", "progress", "opinions"):
-            path = self.engine.config.doc_path(key)
-            try:
-                after = path.read_text(encoding="utf-8")
-            except OSError:
-                after = ""
-            before = backups.get(key, "")
-            docs_preview[key] = after
-
-            if after == before:
-                continue
-
-            rel_path = targets[key]
-            patch = self._build_patch(rel_path, before, after)
-            if patch.strip():
-                patches.append(patch)
-
-        todo_errors = validate_todo_markdown(docs_preview.get("todo", ""))
-        if todo_errors:
-            self._restore_docs(backups)
-            raise SpecIngestError("Invalid TODO format: " + "; ".join(todo_errors))
-
-        self._restore_docs(backups)
-
-        patch_text = "\n".join(patches)
-        if not patch_text.strip():
-            raise SpecIngestError(
-                "OpenCode did not make any changes to TODO/PROGRESS/OPINIONS"
-            )
-
-        self.patch_path.write_text(patch_text, encoding="utf-8")
-        self.last_agent_message = agent_message
-
-        return self._assemble_response(
-            docs_preview, patch=patch_text, agent_message=agent_message
-        )
-
-    async def execute(
-        self,
-        *,
-        force: bool,
-        spec_path: Optional[Path] = None,
-        message: Optional[str] = None,
-        agent: Optional[str] = None,
-        model: Optional[str] = None,
-        reasoning: Optional[str] = None,
-    ) -> Dict[str, str]:
-        async with self.ingest_lock():
-            if (agent or "").strip().lower() == "opencode":
-                return await self._execute_opencode(
-                    force=force,
-                    spec_path=spec_path,
-                    message=message,
-                    model=model,
-                    reasoning=reasoning,
-                )
-            return await self._execute_app_server(
-                force=force,
-                spec_path=spec_path,
-                message=message,
-                model=model,
-                reasoning=reasoning,
-            )
-
-
-class SpecIngestPatchParser:
-    @staticmethod
-    def parse_agent_message(text: str) -> str:
-        clean = (text or "").strip()
-        if not clean:
-            return "Updated docs via spec ingest."
-        for line in clean.splitlines():
-            if line.lower().startswith("agent:"):
-                return line[len("agent:") :].strip() or "Updated docs via spec ingest."
-        return clean.splitlines()[0].strip()
-
-    @staticmethod
-    def strip_code_fences(text: str) -> str:
-        # Kept for backward compatibility if needed, but likely unused in new flow
-        lines = text.strip().splitlines()
-        if (
-            len(lines) >= 2
-            and lines[0].startswith("```")
-            and lines[-1].startswith("```")
-        ):
-            return "\n".join(lines[1:-1]).strip()
-        return text.strip()
-
-    @classmethod
-    def split_patch(cls, output: str) -> tuple[str, str]:
-        # Kept for backward compatibility if needed, but likely unused in new flow
-        if not output:
-            return "", ""
-        match = re.search(
-            r"<PATCH>(.*?)</PATCH>", output, flags=re.IGNORECASE | re.DOTALL
-        )
-        if match:
-            patch_text = cls.strip_code_fences(match.group(1))
-            before = output[: match.start()].strip()
-            after = output[match.end() :].strip()
-            message_text = "\n".join(part for part in [before, after] if part)
-            return message_text, patch_text
-        lines = output.splitlines()
-        start_idx = None
-        for idx, line in enumerate(lines):
-            if line.startswith("--- ") or line.startswith("*** Begin Patch"):
-                start_idx = idx
-                break
-        if start_idx is None:
-            return output.strip(), ""
-        message_text = "\n".join(lines[:start_idx]).strip()
-        patch_text = "\n".join(lines[start_idx:]).strip()
-        patch_text = cls.strip_code_fences(patch_text)
-        return message_text, patch_text