remote-coder 0.5.1__tar.gz → 0.5.3__tar.gz

{remote_coder-0.5.1 → remote_coder-0.5.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: remote-coder
-Version: 0.5.1
+Version: 0.5.3
 Summary: Telegram-based remote AI coding automation server
 Author: Remote AI Coder contributors
 License:                                  Apache License
@@ -236,7 +236,7 @@ Run Claude Code, Codex, or Gemini on your local development machine by sending a
 - Claude, Codex, and Gemini runners behind the same job flow.
 - Reply-linked jobs continue the same AI CLI session, so follow-ups build on prior context.
 - Local admin UI for project setup, advanced settings, logs, and conversation memory.
-- Read-only `plan:` and `ask:` modes when you want analysis without commits. PLAN mode asks open decisions through inline buttons first, then finalizes the plan from your answers.
+- Read-only `plan:`, `ask:`, and `research:` modes when you want analysis without commits. PLAN mode asks open decisions through inline buttons first, then finalizes the plan from your answers; RESEARCH mode asks the selected AI CLI to use internet search when useful.
 
 ## Quick Start
 
@@ -292,7 +292,8 @@ Each project uses its own bot. The webhook path contains the first 16 hex charac
 |---|---|
 | `/start`, `/help` | Open the menu or command help |
 | `/model` | View or change the chat's default model |
-| `/status [job_id]` | Inspect recent or specific jobs |
+| `/status [job_id]` | Inspect recent or specific jobs; running jobs show the latest captured AI output when available |
+| `/log [job_id]` | Pick a recent job or show captured AI stdout for a specific job |
 | `/branch [name]` | Show or switch the bound project's local branch |
 | `/pull` | Fetch remotes and pull the current branch |
 | `/rebase [branch]` | Rebase and fast-forward a completed branch into `main` or `master` |
@@ -311,6 +312,7 @@ Natural-language examples:
 Fix the login validation bug with model: codex
 plan: outline the migration before changing code
 /ask what test command does this repo use?
+/research compare current Telegram webhook security guidance
 수정: 방금 작업에서 README 문구만 더 간결하게 바꿔줘
 ```
 
@@ -321,6 +323,7 @@ Day-to-day setup happens in the local admin UI. Files live under `REMOTE_CODER_H
 - `projects.json` stores project records, bot tokens, allowlists, root paths, and default models.
 - `advanced_settings.json` stores global behavior such as timeouts, sandbox mode, language, worktree retention, and memory limits.
 - `worktrees/<project>/` contains managed job worktrees and logs.
+- On server startup, queued Jobs stored in SQLite are rerun; Jobs that were running when the server stopped are marked failed with `server_restart` for `/status` review.
 
 Useful overrides: `REMOTE_CODER_HOME`, `PROJECTS_CONFIG_PATH`, `CONVERSATION_DB_PATH`, and `JOB_DB_PATH`.
 

{remote_coder-0.5.1 → remote_coder-0.5.3}/README.md

@@ -15,7 +15,7 @@ Run Claude Code, Codex, or Gemini on your local development machine by sending a
 - Claude, Codex, and Gemini runners behind the same job flow.
 - Reply-linked jobs continue the same AI CLI session, so follow-ups build on prior context.
 - Local admin UI for project setup, advanced settings, logs, and conversation memory.
-- Read-only `plan:` and `ask:` modes when you want analysis without commits. PLAN mode asks open decisions through inline buttons first, then finalizes the plan from your answers.
+- Read-only `plan:`, `ask:`, and `research:` modes when you want analysis without commits. PLAN mode asks open decisions through inline buttons first, then finalizes the plan from your answers; RESEARCH mode asks the selected AI CLI to use internet search when useful.
 
 ## Quick Start
 
@@ -71,7 +71,8 @@ Each project uses its own bot. The webhook path contains the first 16 hex charac
 |---|---|
 | `/start`, `/help` | Open the menu or command help |
 | `/model` | View or change the chat's default model |
-| `/status [job_id]` | Inspect recent or specific jobs |
+| `/status [job_id]` | Inspect recent or specific jobs; running jobs show the latest captured AI output when available |
+| `/log [job_id]` | Pick a recent job or show captured AI stdout for a specific job |
 | `/branch [name]` | Show or switch the bound project's local branch |
 | `/pull` | Fetch remotes and pull the current branch |
 | `/rebase [branch]` | Rebase and fast-forward a completed branch into `main` or `master` |
@@ -90,6 +91,7 @@ Natural-language examples:
 Fix the login validation bug with model: codex
 plan: outline the migration before changing code
 /ask what test command does this repo use?
+/research compare current Telegram webhook security guidance
 수정: 방금 작업에서 README 문구만 더 간결하게 바꿔줘
 ```
 
@@ -100,6 +102,7 @@ Day-to-day setup happens in the local admin UI. Files live under `REMOTE_CODER_H
 - `projects.json` stores project records, bot tokens, allowlists, root paths, and default models.
 - `advanced_settings.json` stores global behavior such as timeouts, sandbox mode, language, worktree retention, and memory limits.
 - `worktrees/<project>/` contains managed job worktrees and logs.
+- On server startup, queued Jobs stored in SQLite are rerun; Jobs that were running when the server stopped are marked failed with `server_restart` for `/status` review.
 
 Useful overrides: `REMOTE_CODER_HOME`, `PROJECTS_CONFIG_PATH`, `CONVERSATION_DB_PATH`, and `JOB_DB_PATH`.
 

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/__init__.py

@@ -1,3 +1,3 @@
 """Remote AI Coder application package."""
 
-__version__ = "0.5.1"
+__version__ = "0.5.3"

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/admin/router.py

@@ -378,9 +378,10 @@ def create_admin_router(
         report = check_prerequisites()
         installed = [cli.name for cli in report.ai_clis if cli.installed]
         _adminlog.info(
-            "prerequisites queried ngrok_ok=%s ai_clis=%s",
+            "prerequisites queried ngrok_ok=%s ai_clis=%s github_cli=%s",
             report.ngrok_ok,
             ",".join(installed) or "-",
+            report.github_cli.installed,
         )
         return report.model_dump()
 

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/admin/static/i18n.js

@@ -146,6 +146,11 @@
       en: "None found (install at least one: claude / codex / gemini)",
       ko: "설치된 것 없음 (claude / codex / gemini 중 최소 1개 설치)",
     },
+    "setup.githubCli": { en: "GitHub CLI", ko: "GitHub CLI" },
+    "setup.githubCliNone": {
+      en: "Not found (install gh before using /pr)",
+      ko: "설치되지 않음 (/pr 사용 전 gh 설치 필요)",
+    },
     "setup.checking": { en: "Checking…", ko: "확인 중…" },
     "setup.recheck": { en: "Re-check", ko: "다시 확인" },
     "setup.cta": { en: "Add your first project", ko: "첫 프로젝트 추가" },

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/admin/templates/admin.html

@@ -248,9 +248,12 @@
       const installed = (report.ai_clis || []).filter((c) => c.installed).map((c) => c.name);
       const aiOk = installed.length > 0;
       const aiDetail = aiOk ? installed.join(", ") : i18n.t("setup.aiCliNone");
+      const ghOk = !!(report.github_cli && report.github_cli.installed);
+      const ghDetail = ghOk ? "gh" : i18n.t("setup.githubCliNone");
       $("#setup-prereq").innerHTML =
         renderPrereqRow(!!report.ngrok_ok, i18n.t("setup.ngrok"), ngrokDetail) +
-        renderPrereqRow(aiOk, i18n.t("setup.aiCli"), aiDetail);
+        renderPrereqRow(aiOk, i18n.t("setup.aiCli"), aiDetail) +
+        renderPrereqRow(ghOk, i18n.t("setup.githubCli"), ghDetail);
     }
 
     async function loadPrerequisites() {

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/ai/base.py

@@ -4,9 +4,11 @@ import re
 import subprocess
 import threading
 from abc import ABC, abstractmethod
+from collections.abc import Callable
 from dataclasses import dataclass, field
 from datetime import UTC, datetime
 from pathlib import Path
+from typing import Literal
 
 from app.jobs.schemas import JobMode
 from app.monitoring.events import EventLogger
@@ -57,6 +59,18 @@ _PLAN_DECISIONS_INSTRUCTION = (
 )
 
 
+_RESEARCH_INSTRUCTION = (
+    "You are in RESEARCH mode. Read the repository context and answer the user's research "
+    "question. Do not modify files.\n\n"
+    "Use internet search when it is useful or necessary for the question, similar to a deep "
+    "research workflow. Compare multiple perspectives or sources when possible, and clearly "
+    "separate repository-derived facts from external findings. Include citations or source "
+    "links for external claims, call out uncertainty or limitations, and finish with a direct "
+    "answer to the user's problem.\n\n"
+    "User research request:\n"
+)
+
+
 def instruction_for_runner_mode(instruction: str, mode: JobMode) -> str:
     if mode == JobMode.PLAN:
         return f"{_PLAN_DECISIONS_INSTRUCTION}{instruction}"
@@ -66,6 +80,8 @@ def instruction_for_runner_mode(instruction: str, mode: JobMode) -> str:
             "Do not modify files.\n\n"
             f"User question:\n{instruction}"
         )
+    if mode == JobMode.RESEARCH:
+        return f"{_RESEARCH_INSTRUCTION}{instruction}"
     return instruction
 
 
@@ -81,6 +97,9 @@ class RunnerInput:
     session_id: str | None = None
     resume_token: str | None = None
     native_resume_cwd_stable: bool = True
+    output_callback: Callable[[Literal["stdout", "stderr"], str], None] | None = field(
+        default=None, compare=False
+    )
 
 
 @dataclass
@@ -199,23 +218,30 @@ class BaseCliRunner(AiRunner):
                 cancelled.set()
 
             threading.Thread(target=_watch, daemon=True).start()
-        try:
-            stdout_data, stderr_data = proc.communicate(timeout=runner_input.timeout_seconds)
-        except subprocess.TimeoutExpired as exc:
-            proc.kill()
-            stdout_data, stderr_data = proc.communicate()
-            self._log.warning(
-                "timeout after %ds stdout_len=%d stderr_len=%d",
-                runner_input.timeout_seconds,
-                len(stdout_data),
-                len(stderr_data),
+        if runner_input.output_callback is None:
+            try:
+                stdout_data, stderr_data = proc.communicate(timeout=runner_input.timeout_seconds)
+            except subprocess.TimeoutExpired as exc:
+                proc.kill()
+                stdout_data, stderr_data = proc.communicate()
+                self._log.warning(
+                    "timeout after %ds stdout_len=%d stderr_len=%d",
+                    runner_input.timeout_seconds,
+                    len(stdout_data),
+                    len(stderr_data),
+                )
+                raise RunnerExecutionError(
+                    f"runner timed out after {runner_input.timeout_seconds}s",
+                    stdout=stdout_data,
+                    stderr=stderr_data,
+                    started_at=started_at,
+                ) from exc
+        else:
+            stdout_data, stderr_data = self._communicate_with_output_callback(
+                proc,
+                runner_input,
+                started_at,
             )
-            raise RunnerExecutionError(
-                f"runner timed out after {runner_input.timeout_seconds}s",
-                stdout=stdout_data,
-                stderr=stderr_data,
-                started_at=started_at,
-            ) from exc
         finished_at = datetime.now(UTC)
         if cancelled.is_set():
             raise RunnerExecutionError(
@@ -241,3 +267,61 @@ class BaseCliRunner(AiRunner):
             finished_at=finished_at,
             session_id=self._resolve_result_session_id(runner_input, session_files_before),
         )
+
+    def _communicate_with_output_callback(
+        self,
+        proc: subprocess.Popen[str],
+        runner_input: RunnerInput,
+        started_at: datetime,
+    ) -> tuple[str, str]:
+        output_callback = runner_input.output_callback
+        if output_callback is None:
+            raise ValueError("output_callback is required")
+        stdout_parts: list[str] = []
+        stderr_parts: list[str] = []
+
+        def _read_stream(
+            stream_name: Literal["stdout", "stderr"],
+            parts: list[str],
+        ) -> None:
+            stream = proc.stdout if stream_name == "stdout" else proc.stderr
+            if stream is None:
+                return
+            while True:
+                chunk = stream.readline()
+                if chunk == "":
+                    break
+                parts.append(chunk)
+                try:
+                    output_callback(stream_name, chunk)
+                except Exception:  # pylint: disable=broad-except
+                    self._log.warning("output callback failed stream=%s", stream_name)
+
+        stdout_thread = threading.Thread(target=_read_stream, args=("stdout", stdout_parts))
+        stderr_thread = threading.Thread(target=_read_stream, args=("stderr", stderr_parts))
+        stdout_thread.start()
+        stderr_thread.start()
+        try:
+            proc.wait(timeout=runner_input.timeout_seconds)
+        except subprocess.TimeoutExpired as exc:
+            proc.kill()
+            proc.wait()
+            stdout_thread.join()
+            stderr_thread.join()
+            stdout_data = "".join(stdout_parts)
+            stderr_data = "".join(stderr_parts)
+            self._log.warning(
+                "timeout after %ds stdout_len=%d stderr_len=%d",
+                runner_input.timeout_seconds,
+                len(stdout_data),
+                len(stderr_data),
+            )
+            raise RunnerExecutionError(
+                f"runner timed out after {runner_input.timeout_seconds}s",
+                stdout=stdout_data,
+                stderr=stderr_data,
+                started_at=started_at,
+            ) from exc
+        stdout_thread.join()
+        stderr_thread.join()
+        return "".join(stdout_parts), "".join(stderr_parts)

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/ai/codex.py

@@ -3,7 +3,7 @@ from __future__ import annotations
 from pathlib import Path
 
 from app.ai.base import BaseCliRunner, RunnerInput, instruction_for_runner_mode
-from app.jobs.schemas import JobMode
+from app.jobs.schemas import is_read_only_job_mode
 from app.models import CodexSandboxMode
 from app.monitoring.events import EventLogger
 
@@ -16,7 +16,7 @@ class CodexRunner(BaseCliRunner):
         self._sandbox = sandbox
 
     def _resolve_sandbox(self, runner_input: RunnerInput) -> CodexSandboxMode:
-        if runner_input.mode in (JobMode.PLAN, JobMode.ASK):
+        if is_read_only_job_mode(runner_input.mode):
             return CodexSandboxMode.READ_ONLY
         return self._sandbox
 
@@ -29,7 +29,7 @@ class CodexRunner(BaseCliRunner):
 
     def build_argv(self, runner_input: RunnerInput) -> list[str]:
         sandbox = self._resolve_sandbox(runner_input)
-        if runner_input.mode in (JobMode.PLAN, JobMode.ASK):
+        if is_read_only_job_mode(runner_input.mode):
             instruction = instruction_for_runner_mode(runner_input.instruction, runner_input.mode)
         else:
             instruction = runner_input.instruction

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/ai/gemini.py

@@ -3,7 +3,7 @@ from __future__ import annotations
 from pathlib import Path
 
 from app.ai.base import BaseCliRunner, RunnerInput, instruction_for_runner_mode
-from app.jobs.schemas import JobMode
+from app.jobs.schemas import is_read_only_job_mode
 from app.monitoring.events import EventLogger
 
 
@@ -17,7 +17,7 @@ class GeminiRunner(BaseCliRunner):
         return Path.home() / ".gemini" / "sessions"
 
     def build_argv(self, runner_input: RunnerInput) -> list[str]:
-        if runner_input.mode in (JobMode.PLAN, JobMode.ASK):
+        if is_read_only_job_mode(runner_input.mode):
             prompt = instruction_for_runner_mode(runner_input.instruction, runner_input.mode)
             argv = ["gemini", "--skip-trust", "-p", prompt]
         else:

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/cli.py

@@ -32,7 +32,7 @@ def build_parser() -> argparse.ArgumentParser:
         help="Run the server only (skip ngrok and webhook registration)",
     )
 
-    subparsers.add_parser("doctor", help="Check prerequisites (ngrok, AI CLIs)")
+    subparsers.add_parser("doctor", help="Check prerequisites (ngrok, AI CLIs, GitHub CLI)")
     return parser
 
 
@@ -128,6 +128,10 @@ def run_doctor() -> None:
             "  ⚠️ AI CLI (claude/codex/gemini) not found. Install at least one. "
             "(e.g. npm install -g @anthropic-ai/claude-code)"
         )
+    if report.github_cli.installed:
+        print("  ✅ GitHub CLI: gh")
+    else:
+        print("  ⚠️ GitHub CLI (gh) not found. Install it before using /pr.")
 
 
 if __name__ == "__main__":

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/diagnostics.py

@@ -7,6 +7,7 @@ from pydantic import BaseModel
 from app import tunnel
 
 AI_CLI_TOOLS = ("claude", "codex", "gemini")
+GITHUB_CLI_TOOL = "gh"
 
 
 class AiCliStatus(BaseModel):
@@ -18,6 +19,7 @@ class PrerequisitesReport(BaseModel):
     ngrok_ok: bool
     ngrok_detail: str
     ai_clis: list[AiCliStatus]
+    github_cli: AiCliStatus
 
 
 def check_prerequisites() -> PrerequisitesReport:
@@ -34,4 +36,13 @@ def check_prerequisites() -> PrerequisitesReport:
         AiCliStatus(name=tool, installed=shutil.which(tool) is not None)
         for tool in AI_CLI_TOOLS
     ]
-    return PrerequisitesReport(ngrok_ok=ngrok_ok, ngrok_detail=ngrok_detail, ai_clis=ai_clis)
+    github_cli = AiCliStatus(
+        name=GITHUB_CLI_TOOL,
+        installed=shutil.which(GITHUB_CLI_TOOL) is not None,
+    )
+    return PrerequisitesReport(
+        ngrok_ok=ngrok_ok,
+        ngrok_detail=ngrok_detail,
+        ai_clis=ai_clis,
+        github_cli=github_cli,
+    )

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/jobs/execution_pipeline.py

@@ -6,7 +6,7 @@ from pathlib import Path
 from app.ai.base import RunnerInput
 from app.git.commit_message import CommitMessageFormatter
 from app.jobs.plan_decisions import PlanDecisionQuestion, parse_plan_decisions
-from app.jobs.schemas import Job, JobMode
+from app.jobs.schemas import Job, JobMode, is_read_only_job_mode
 from app.monitoring.events import EventLogger
 
 _joblog = EventLogger("app.jobs.lifecycle", "job.lifecycle")
@@ -50,7 +50,7 @@ def run_job(manager, job_id: str) -> Job:
     created_worktree_for_job = False
     failed_stage: str | None = None
     remote = manager._effective_git_remote_name()
-    read_only_job = job.request.mode in (JobMode.PLAN, JobMode.ASK)
+    read_only_job = is_read_only_job_mode(job.request.mode)
     plan_decision_questions: list[PlanDecisionQuestion] | None = None
     try:
         job.mark_running()
@@ -83,6 +83,7 @@ def run_job(manager, job_id: str) -> Job:
             len(job.request.instruction),
             **manager._job_ctx(job),
         )
+        runner_log = manager._start_incremental_runner_log(job, worktree_base)
         heartbeat = manager._start_heartbeat(job)
         try:
             runner_result = runner.run(
@@ -97,10 +98,12 @@ def run_job(manager, job_id: str) -> Job:
                     session_id=job.request.session_id,
                     resume_token=job.request.resume_session_token,
                     native_resume_cwd_stable=not created_worktree_for_job,
+                    output_callback=runner_log.output_callback,
                 )
             )
         finally:
             heartbeat.set()
+            runner_log.flush()
         manager._save_runner_log(job, runner_result, worktree_base)
         _joblog.info(
             "runner exit=%d stdout_len=%d stderr_len=%d",
@@ -238,9 +241,7 @@ def run_job(manager, job_id: str) -> Job:
     finally:
         manager._cancel_events.pop(job_id, None)
         manager._cancelled_job_ids.discard(job_id)
-        read_only_succeeded = (
-            job.request.mode in (JobMode.PLAN, JobMode.ASK) and job.status.value == "succeeded"
-        )
+        read_only_succeeded = is_read_only_job_mode(job.request.mode) and job.status.value == "succeeded"
         cleanup_on_success = read_only_succeeded or not manager._effective_keep_worktree_on_success()
         _joblog.info(
             "job finalizing status=%s created_worktree=%s cleanup_on_success=%s",

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/jobs/fix_pipeline.py

@@ -79,6 +79,7 @@ def run_fix_job(manager, job_id: str) -> Job:
         runner = manager._runner_factory.create(job.request.model)
         timeout_seconds = manager._effective_job_timeout_seconds()
         fix_prompt = manager.compose_fix_source_prompt(parent_job, job.request.instruction)
+        runner_log = manager._start_incremental_runner_log(job, worktree_base)
         heartbeat = manager._start_heartbeat(job)
         try:
             runner_result = runner.run(
@@ -93,10 +94,12 @@ def run_fix_job(manager, job_id: str) -> Job:
                     session_id=job.request.session_id,
                     resume_token=job.request.resume_session_token,
                     native_resume_cwd_stable=not created_worktree_for_job,
+                    output_callback=runner_log.output_callback,
                 )
             )
         finally:
             heartbeat.set()
+            runner_log.flush()
         manager._save_runner_log(job, runner_result, worktree_base)
         if runner_result.exit_code != 0:
             raise RuntimeError(runner_result.stderr.strip() or "runner failed")

{remote_coder-0.5.1 → remote_coder-0.5.3}/app/jobs/manager.py

@@ -26,6 +26,7 @@ from app.jobs.result_writer import (
     make_output_summary,
     preserve_partial_output,
     save_runner_log,
+    start_incremental_runner_log,
     strip_links_for_stdout_summary,
 )
 from app.jobs.schemas import FixKind, Job, JobMode, JobRequest
@@ -186,6 +187,15 @@ class JobManager:
         with self._project_lock(job.request.project):
             return run_job(self, job_id)
 
+    def recover(self, job_id: str) -> Job:
+        job = self._job_store.get(job_id)
+        if job is None:
+            return run_job(self, job_id)
+        with self._project_lock(job.request.project):
+            if job.request.mode is JobMode.AGENT_FIX:
+                return self._run_fix(job_id)
+            return run_job(self, job_id)
+
     def _project_lock(self, project: str) -> threading.Lock:
         with self._project_locks_guard:
             lock = self._project_locks.get(project)
@@ -282,6 +292,9 @@ class JobManager:
     def _save_runner_log(self, job: Job, runner_result, worktree_base: Path) -> None:
         save_runner_log(job, runner_result, worktree_base)
 
+    def _start_incremental_runner_log(self, job: Job, worktree_base: Path):
+        return start_incremental_runner_log(job, worktree_base, self._job_store.update)
+
     @classmethod
     def _strip_links_for_stdout_summary(cls, text: str) -> str:
         return strip_links_for_stdout_summary(text)