codex-autorunner 1.1.0__py3-none-any.whl → 1.2.1__py3-none-any.whl

codex_autorunner/flows/review/service.py

@@ -14,7 +14,6 @@ from ...agents.opencode.run_prompt import OpenCodeRunConfig, run_opencode_prompt
 from ...agents.opencode.supervisor import OpenCodeSupervisor
 from ...agents.registry import has_capability, validate_agent_id
 from ...core.config import RepoConfig
-from ...core.engine import Engine
 from ...core.locks import (
     FileLock,
     FileLockBusy,
@@ -22,6 +21,7 @@ from ...core.locks import (
     process_alive,
     read_lock_info,
 )
+from ...core.runtime import RuntimeContext
 from ...core.state import now_iso
 from ...core.utils import atomic_write, read_json
 
@@ -252,7 +252,7 @@ Stop digging deeper when:
 """
 REVIEW_PROMPT_SPEC_PROGRESS = """# Autorunner Final Review (Spec + Progress Focus)
 
-You are coordinating a multi-agent review immediately after an autorunner loop completes. The goal is to validate work against SPEC.md and PROGRESS.md, not to suggest generic code style fixes.
+You are coordinating a multi-agent review immediately after an autorunner loop completes. The goal is to validate work against spec.md, not to suggest generic code style fixes.
 
 ## Required Scratchpad + Output
 
@@ -269,16 +269,16 @@ You are coordinating a multi-agent review immediately after an autorunner loop c
 
 ## Source of Truth + Focus
 
-* Read AUTORUNNER_CONTEXT.md first. It contains the exit reason, SPEC.md, PROGRESS.md, optional TODO/SUMMARY, and the last-run artifacts (output/diff/plan excerpts).
-* Treat SPEC.md + PROGRESS.md as contracts: extract explicit requirements, promised validation steps, and open questions.
+* Read AUTORUNNER_CONTEXT.md first. It contains the exit reason, spec.md, and last-run artifacts (output/diff/plan excerpts).
+* Treat spec.md as the contract: extract explicit requirements, promised validation steps, and open questions.
 * Derive must-hold invariants directly from SPEC (not generic guesses).
 * Buckets should be anchored to spec sections mapped to implementation areas, plus any recently changed/critical modules from AUTORUNNER_CONTEXT.md.
 
 ## North Star
 
-* Spec compliance and claim verification over style.
-* High-signal risks and regressions that would violate SPEC/PROGRESS commitments.
-* Trackable output that can be turned into TODO items.
+* Spec compliance over style.
+* High-signal risks and regressions that would violate spec commitments.
+* Trackable output that can be turned into tickets.
 
 ## Phase 0: Setup (Coordinator)
 
@@ -288,8 +288,7 @@ Prepare scratchpad files under {{scratchpad_dir}} (see list above).
 
 1) Read AUTORUNNER_CONTEXT.md and summarize:
    * Project shape and runtime assumptions
-   * SPEC requirements + invariants
-   * PROGRESS claims + validation evidence
+   * Spec requirements + invariants
    * Open questions/gaps
 2) Define buckets in BUCKETS.md:
    * Buckets by review dimension + code areas (spec sections → code paths)
@@ -299,7 +298,6 @@ Prepare scratchpad files under {{scratchpad_dir}} (see list above).
 
 Launch subagents by review dimension:
 * Spec compliance agent: requirement → evidence mapping.
-* Progress verification agent: PROGRESS claims → repo/tests evidence.
 * Risk & regression agent: likely failure modes introduced by recent changes.
 * Optional: Test adequacy agent if tests are in-scope.
 
@@ -320,8 +318,8 @@ Create the final report at {{final_output_path}} with this structure:
 ## Spec-to-Implementation Matrix
 | Spec item | Status (met/partial/missing) | Evidence | Notes |
 
-## Progress Claim Verification
-- Claim: ... (cite PROGRESS.md section)
+## Spec Verification
+- Spec requirement: ... (cite spec section)
   - Verified by: ...
   - Not verified because: ...
 
@@ -393,19 +391,19 @@ def _default_state() -> dict[str, Any]:
 class ReviewService:
     def __init__(
         self,
-        engine: Engine,
+        ctx: RuntimeContext,
         *,
         opencode_supervisor: Optional[OpenCodeSupervisor] = None,
         app_server_supervisor: Optional[Any] = None,
         logger: Optional[logging.Logger] = None,
     ) -> None:
-        self.engine = engine
+        self.ctx = ctx
         self._opencode_supervisor = opencode_supervisor
         self._app_server_supervisor = app_server_supervisor
         self._logger = logger or logging.getLogger("codex_autorunner.review")
-        self._state_path = _workflow_root(engine.repo_root) / "state.json"
+        self._state_path = _workflow_root(self.ctx.repo_root) / "state.json"
         self._lock_path = (
-            engine.repo_root / ".codex-autorunner" / "locks" / "review.lock"
+            self.ctx.repo_root / ".codex-autorunner" / "locks" / "review.lock"
         )
         self._thread: Optional[threading.Thread] = None
         self._thread_lock = threading.Lock()
@@ -413,9 +411,9 @@ class ReviewService:
         self._lock_handle: Optional[FileLock] = None
 
     def _repo_config(self) -> RepoConfig:
-        if not isinstance(self.engine.config, RepoConfig):
+        if not isinstance(self.ctx.config, RepoConfig):
             raise ReviewError("Review requires a repo workspace config")
-        return self.engine.config
+        return self.ctx.config
 
     def status(self) -> dict[str, Any]:
         state = self._load_state()
@@ -438,7 +436,7 @@ class ReviewService:
                 raise ReviewBusyError("Review already running", status_code=409)
             if self._thread and self._thread.is_alive():
                 raise ReviewBusyError("Review already running", status_code=409)
-            busy_reason = self.engine.repo_busy_reason()
+            busy_reason = self.ctx.repo_busy_reason()
             if busy_reason:
                 raise ReviewConflictError(
                     f"Cannot start review: {busy_reason}", status_code=409
@@ -475,7 +473,7 @@ class ReviewService:
             state = self.status()
             if state.get("status") in ("running", "stopping"):
                 raise ReviewBusyError("Review already running", status_code=409)
-            busy_reason = self.engine.repo_busy_reason()
+            busy_reason = self.ctx.repo_busy_reason()
             if busy_reason and not ignore_repo_busy:
                 raise ReviewConflictError(
                     f"Cannot start review: {busy_reason}", status_code=409
@@ -680,7 +678,7 @@ class ReviewService:
             )
 
         run_id = state["id"]
-        runs_dir = _workflow_root(self.engine.repo_root) / "runs"
+        runs_dir = _workflow_root(self.ctx.repo_root) / "runs"
         run_dir = runs_dir / run_id
         run_dir.mkdir(parents=True, exist_ok=True)
 
@@ -735,7 +733,7 @@ class ReviewService:
         if agent_id == "codex":
             if self._app_server_supervisor is None:
                 raise ReviewError("Codex backend is not configured")
-            client = await self._app_server_supervisor.get_client(self.engine.repo_root)
+            client = await self._app_server_supervisor.get_client(self.ctx.repo_root)
             thread_id = uuid.uuid4().hex
             review_kwargs: dict[str, Any] = {}
             if state.get("model"):
@@ -746,7 +744,7 @@ class ReviewService:
                 thread_id=thread_id,
                 target={"type": "custom", "instructions": prompt},
                 delivery="inline",
-                cwd=str(self.engine.repo_root),
+                cwd=str(self.ctx.repo_root),
                 **review_kwargs,
             )
 
@@ -802,7 +800,7 @@ class ReviewService:
             subagent_model = review_cfg.get("subagent_model")
             if subagent_agent_id:
                 await self._opencode_supervisor.ensure_subagent_config(
-                    workspace_root=self.engine.repo_root,
+                    workspace_root=self.ctx.repo_root,
                     agent_id=subagent_agent_id,
                     model=subagent_model,
                 )
@@ -812,7 +810,7 @@ class ReviewService:
                 model=state["model"],
                 reasoning=state.get("reasoning"),
                 prompt=prompt,
-                workspace_root=str(self.engine.repo_root),
+                workspace_root=str(self.ctx.repo_root),
                 timeout_seconds=timeout_seconds,
                 interrupt_grace_seconds=REVIEW_INTERRUPT_GRACE_SECONDS,
                 permission_policy="allow",

codex_autorunner/flows/ticket_flow/definition.py

@@ -6,6 +6,7 @@ from typing import Any, Dict, Optional
 from ...core.flows.definition import EmitEventFn, FlowDefinition, StepOutcome
 from ...core.flows.models import FlowEventType, FlowRunRecord
 from ...core.utils import find_repo_root
+from ...manifest import ManifestError, load_manifest
 from ...tickets import (
     DEFAULT_MAX_TOTAL_TURNS,
     AgentPool,
@@ -35,18 +36,36 @@ def build_ticket_flow_definition(*, agent_pool: AgentPool) -> FlowDefinition:
         engine_state = dict(engine_state) if isinstance(engine_state, dict) else {}
 
         repo_root = find_repo_root()
-        workspace_root = Path(input_data.get("workspace_root") or repo_root)
+        raw_workspace = input_data.get("workspace_root") or repo_root
+        workspace_root = Path(raw_workspace)
+        if not workspace_root.is_absolute():
+            workspace_root = (Path(repo_root) / workspace_root).resolve()
+        else:
+            workspace_root = workspace_root.resolve()
+
         ticket_dir = Path(input_data.get("ticket_dir") or ".codex-autorunner/tickets")
+        if not ticket_dir.is_absolute():
+            ticket_dir = (workspace_root / ticket_dir).resolve()
+
         runs_dir = Path(input_data.get("runs_dir") or ".codex-autorunner/runs")
+        if not runs_dir.is_absolute():
+            runs_dir = (workspace_root / runs_dir).resolve()
         max_total_turns = int(
             input_data.get("max_total_turns") or DEFAULT_MAX_TOTAL_TURNS
         )
         max_lint_retries = int(input_data.get("max_lint_retries") or 3)
         max_commit_retries = int(input_data.get("max_commit_retries") or 2)
+        max_network_retries = int(input_data.get("max_network_retries") or 5)
         auto_commit = bool(
             input_data.get("auto_commit") if "auto_commit" in input_data else True
         )
+        include_previous_ticket_context = bool(
+            input_data.get("include_previous_ticket_context")
+            if "include_previous_ticket_context" in input_data
+            else False
+        )
 
+        repo_id = _resolve_ticket_flow_repo_id(workspace_root)
         runner = TicketRunner(
             workspace_root=workspace_root,
             run_id=str(record.id),
@@ -56,9 +75,12 @@ def build_ticket_flow_definition(*, agent_pool: AgentPool) -> FlowDefinition:
                 max_total_turns=max_total_turns,
                 max_lint_retries=max_lint_retries,
                 max_commit_retries=max_commit_retries,
+                max_network_retries=max_network_retries,
                 auto_commit=auto_commit,
+                include_previous_ticket_context=include_previous_ticket_context,
             ),
             agent_pool=agent_pool,
+            repo_id=repo_id,
         )
 
         if emit_event is not None:
@@ -91,8 +113,28 @@ def build_ticket_flow_definition(*, agent_pool: AgentPool) -> FlowDefinition:
                 "max_total_turns": {"type": "integer"},
                 "max_lint_retries": {"type": "integer"},
                 "max_commit_retries": {"type": "integer"},
+                "max_network_retries": {"type": "integer"},
                 "auto_commit": {"type": "boolean"},
+                "include_previous_ticket_context": {"type": "boolean"},
             },
         },
         steps={"ticket_turn": _ticket_turn_step},
     )
+
+
+def _resolve_ticket_flow_repo_id(workspace_root: Path) -> str:
+    current = workspace_root
+    for _ in range(5):
+        manifest_path = current / ".codex-autorunner" / "manifest.yml"
+        if manifest_path.exists():
+            try:
+                manifest = load_manifest(manifest_path, current)
+            except ManifestError:
+                return ""
+            entry = manifest.get_by_path(current, workspace_root)
+            return entry.id if entry else ""
+        parent = current.parent
+        if parent == current:
+            break
+        current = parent
+    return ""

codex_autorunner/integrations/agents/__init__.py

@@ -1,3 +1,4 @@
+from .backend_orchestrator import build_backend_orchestrator
 from .codex_adapter import CodexAdapterOrchestrator
 from .codex_backend import CodexAppServerBackend
 from .opencode_adapter import OpenCodeAdapterOrchestrator
@@ -14,4 +15,5 @@ __all__ = [
     "OpenCodeBackend",
     "build_agent_backend_factory",
     "build_app_server_supervisor_factory",
+    "build_backend_orchestrator",
 ]

codex_autorunner/integrations/agents/backend_orchestrator.py

@@ -20,6 +20,9 @@ from ...core.app_server_threads import (
 )
 from ...core.config import RepoConfig
 from ...core.ports.agent_backend import AgentBackend
+from ...core.ports.backend_orchestrator import (
+    BackendOrchestrator as BackendOrchestratorProtocol,
+)
 from ...core.ports.run_event import RunEvent
 from ...core.state import RunnerState
 from .codex_backend import CodexAppServerBackend
@@ -278,7 +281,22 @@ class BackendOrchestrator:
         return factory_fn(event_prefix, notification_handler)
 
 
+def build_backend_orchestrator(
+    repo_root: Path, config: RepoConfig
+) -> BackendOrchestratorProtocol:
+    """
+    Build a BackendOrchestrator for protocol-agnostic backend management.
+    """
+    return BackendOrchestrator(
+        repo_root=repo_root,
+        config=config,
+        notification_handler=None,
+        logger=logging.getLogger("codex_autorunner.backend"),
+    )
+
+
 __all__ = [
     "BackendOrchestrator",
     "BackendContext",
+    "build_backend_orchestrator",
 ]

codex_autorunner/integrations/agents/codex_backend.py

@@ -1,9 +1,11 @@
 import asyncio
+import hashlib
 import logging
 from pathlib import Path
 from typing import Any, AsyncGenerator, Awaitable, Callable, Dict, Optional, Union
 
 from ...core.circuit_breaker import CircuitBreaker
+from ...core.logging_utils import log_event
 from ...core.ports.agent_backend import AgentBackend, AgentEvent, now_iso
 from ...core.ports.run_event import (
     ApprovalRequested,
@@ -172,10 +174,14 @@ class CodexAppServerBackend(AgentBackend):
         if not self._thread_id:
             await self.start_session(target={}, context={})
 
-        _logger.info(
-            "Running turn on thread %s with message: %s",
-            self._thread_id or "unknown",
-            message[:100],
+        message_hash = hashlib.sha256(message.encode()).hexdigest()[:16]
+        log_event(
+            self._logger,
+            logging.INFO,
+            "agent.turn_started",
+            thread_id=self._thread_id,
+            message_length=len(message),
+            message_hash=message_hash,
         )
 
         turn_kwargs: Dict[str, Any] = {}
@@ -220,10 +226,15 @@ class CodexAppServerBackend(AgentBackend):
         else:
             actual_session_id = self._thread_id
 
-        _logger.info(
-            "Running turn events on thread %s with message: %s",
-            actual_session_id or "unknown",
-            message[:100],
+        message_hash = hashlib.sha256(message.encode()).hexdigest()[:16]
+        log_event(
+            self._logger,
+            logging.INFO,
+            "agent.turn_events_started",
+            thread_id=actual_session_id,
+            turn_id=self._turn_id,
+            message_length=len(message),
+            message_hash=message_hash,
         )
 
         yield Started(

codex_autorunner/integrations/agents/runner.py

@@ -9,6 +9,7 @@ from ...core.ports.run_event import (
     OutputDelta,
     RunEvent,
     Started,
+    now_iso,
 )
 
 _logger = logging.getLogger(__name__)
@@ -36,7 +37,7 @@ async def run_turn_with_backend(
             session_id = await backend.start_session(target={}, context={})
 
         if event_callback:
-            event_callback(Started(timestamp=timestamp(), session_id=session_id))
+            event_callback(Started(timestamp=now_iso(), session_id=session_id))
 
         if log_handler:
             log_handler(message)
@@ -64,7 +65,7 @@ async def run_turn_with_backend(
     except Exception as exc:
         _logger.error("Turn execution failed: %s", exc)
         if event_callback:
-            event_callback(Failed(timestamp=timestamp(), error_message=str(exc)))
+            event_callback(Failed(timestamp=now_iso(), error_message=str(exc)))
         return 1
 
 
@@ -83,9 +84,3 @@ async def stream_turn_events(
             yield event
     else:
         yield AgentEvent.stream_delta(content="", delta_type="noop")
-
-
-def timestamp() -> str:
-    from datetime import datetime, timezone
-
-    return datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")

codex_autorunner/integrations/agents/wiring.py

@@ -192,6 +192,7 @@ class AgentBackendFactory:
             max_handles=self._config.app_server.max_handles,
             idle_ttl_seconds=self._config.app_server.idle_ttl_seconds,
             session_stall_timeout_seconds=self._config.opencode.session_stall_timeout_seconds,
+            max_text_chars=self._config.opencode.max_text_chars,
             base_env=None,
             subagent_models=subagent_models,
         )
@@ -269,3 +270,10 @@ def build_app_server_supervisor_factory(
         )
 
     return factory
+
+
+__all__ = [
+    "AgentBackendFactory",
+    "build_agent_backend_factory",
+    "build_app_server_supervisor_factory",
+]

codex_autorunner/integrations/telegram/adapter.py

@@ -1370,7 +1370,7 @@ class TelegramBotClient:
         return bool(result) if isinstance(result, bool) else False
 
     async def download_file(
-        self, file_path: str, max_size_bytes: int = 50 * 1024 * 1024
+        self, file_path: str, max_size_bytes: int = 100 * 1024 * 1024
     ) -> bytes:
         safe_path = file_path.lstrip("/")
         url = f"{self._file_base_url}/{safe_path}"

codex_autorunner/integrations/telegram/config.py

@@ -44,7 +44,7 @@ DEFAULT_APP_SERVER_TURN_TIMEOUT_SECONDS = 28800
 DEFAULT_APPROVAL_TIMEOUT_SECONDS = 300.0
 DEFAULT_MEDIA_MAX_IMAGE_BYTES = 10 * 1024 * 1024
 DEFAULT_MEDIA_MAX_VOICE_BYTES = 10 * 1024 * 1024
-DEFAULT_MEDIA_MAX_FILE_BYTES = 10 * 1024 * 1024
+DEFAULT_MEDIA_MAX_FILE_BYTES = 100 * 1024 * 1024
 DEFAULT_MEDIA_IMAGE_PROMPT = (
     "The user sent an image with no caption. Use it to continue the "
     "conversation; if no clear task, describe the image and ask what they want."