agent-sin 0.1.0

package/CHANGELOG.md

@@ -0,0 +1,33 @@
+## CHANGELOG
+
+agent-sin のリリース履歴。形式は [Keep a Changelog](https://keepachangelog.com/ja/1.1.0/) ベース、バージョニングは [Semantic Versioning](https://semver.org/lang/ja/) に従う。
+
+エントリ種別:
+
+- **Added** — 新機能
+- **Changed** — 既存挙動の変更（互換維持）
+- **Deprecated** — 廃止予定（次メジャーで削除）
+- **Removed** — 削除（破壊的変更）
+- **Fixed** — バグ修正
+- **Security** — セキュリティ修正
+
+互換性ポリシーは [公式ドキュメント](https://agent.shingoirie.com/versioning) を参照。
+
+---
+
+## [0.1.0] — 2026-05-14
+
+### Added
+
+- 初回公開リリース。`agent-sin` CLI、スキル ランタイム（Python / TypeScript）、Discord bot、Builder モード、ビルトインスキル、Codex / OpenAI / Gemini / Claude Code プロバイダ統合、ローカルメモリ／インデックス／スケジューラの基盤。
+- Telegram bot 連携（`src/telegram/`）。Discord と同じ chat / build / intent ルーティングを Telegram でも利用可能に。
+- Discord 添付ファイル取り込み（テキスト/画像）と、ビルド進捗の detail モード。
+- profile memory（`soul.md` / `user.md` / `memory.md`）と日次メモの自動昇格（`daily-memory-promotion`）。
+- `profile-save` ビルトインスキル。
+- `install.sh` によるセットアップ補助。
+- 互換性ポリシー（[versioning](https://agent.shingoirie.com/versioning)）と CHANGELOG。
+
+### Changed
+
+- ビルドモード進入時の固定文言「作業を開始しました」を廃止し、detail モード以外は進捗メッセージを送らない静かな挙動に変更。
+- Discord/Telegram のビルド進捗は `AGENT_SIN_DISCORD_PROGRESS_DETAIL=1` / `AGENT_SIN_TELEGRAM_PROGRESS_DETAIL=1` または runtime 設定の `progress_detail=true` でのみ詳細表示。

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Agent-Sin contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,81 @@
+<p align="center">
+  <img src="assets/logo.png" alt="Agent-Sin logo" width="120">
+</p>
+
+<h1 align="center">Agent-Sin</h1>
+
+<p align="center">
+  A safer, more reliable<br>
+  <strong>program-driven personal AI agent.</strong>
+</p>
+
+<p align="center">
+  <a href="https://agent.shingoirie.com">Website</a> ·
+  <a href="https://agent.shingoirie.com/overview">Docs</a> ·
+  <a href="https://agent.shingoirie.com/getting-started">Quick Start</a>
+</p>
+
+---
+
+Agent-Sin freezes daily work into small **Program Skills** instead of letting an LLM reinterpret every task. Conversation is handled by an LLM at the entrance; execution is delegated to verified programs.
+
+> Design inspired by [openclaw/openclaw](https://github.com/openclaw/openclaw).
+
+## Why program-driven
+
+|  | Traditional AI agent | Agent-Sin |
+|---|---|---|
+| Execution | LLM follows steps every time | Runs pre-built programs |
+| Speed | Bound to inference latency | Fast program execution |
+| Stability | Output drifts run to run | Same input, same behavior |
+| Cost | Charged per LLM call | No LLM cost for skill runs |
+| Safety | Unexpected actions are possible | Only registered actions run |
+
+When a new capability is needed, Build Mode uses Claude Code or Codex to generate a skill. After that it runs the same way — fast, cheap, and predictable.
+
+## Highlights
+
+- **Conversation Mode and Build Mode** — daily chat stays minimum-privilege; elevated permissions kick in only while authoring or editing a skill.
+- **Multi-channel** — talk to the same agent from terminal, Discord, or Telegram.
+- **Flexible notifications** — OS, Discord, Telegram, Slack, and email.
+- **Free model mix** — pick a light model for chat and a stronger one for skill authoring.
+- **Always-on gateway** — starts at login and bundles the scheduler with the Discord / Telegram bots.
+- **Long-term memory** — agent persona, your profile, and daily context persist across sessions.
+
+## Install
+
+Requires Node.js 22+ and Python 3. Works on macOS, Linux, and Windows.
+
+**macOS / Linux**
+
+```bash
+curl -fsSL https://agent.shingoirie.com/install.sh | bash
+```
+
+**Windows (PowerShell)**
+
+```powershell
+irm https://agent.shingoirie.com/install.ps1 | iex
+```
+
+The installer sets up `agent-sin`, runs initial setup, and registers a login service. Workspace data lives in `~/.agent-sin/`.
+
+Full walkthrough: [Getting Started](https://agent.shingoirie.com/getting-started).
+
+## Documentation
+
+- [Overview](https://agent.shingoirie.com/overview) — the big picture
+- [Getting Started](https://agent.shingoirie.com/getting-started) — install to first conversation
+- [Concepts](https://agent.shingoirie.com/concepts) — design and Runtime
+- [Skill Authoring](https://agent.shingoirie.com/skill-authoring) — write your own skill
+- [Built-in Skills](https://agent.shingoirie.com/built-in-skills) — bundled skills
+- [CLI Reference](https://agent.shingoirie.com/cli) — every command
+- [Configuration](https://agent.shingoirie.com/configuration) — settings and API keys
+- [Gateway & Ops](https://agent.shingoirie.com/operations) — always-on, schedules, notifications, backup
+- [Discord](https://agent.shingoirie.com/discord) · [Telegram](https://agent.shingoirie.com/telegram) — channel integrations
+- [Versioning](https://agent.shingoirie.com/versioning) — compatibility and release policy
+- [Changelog](CHANGELOG.md) — release history
+
+## License
+
+MIT

package/builtin-skills/_shared/_models_lib.py

@@ -0,0 +1,227 @@
+"""Common helpers for model-add / model-set / model-list builtin skills.
+
+Reads/writes ~/.agent-sin/models.yaml in the format used by
+src/core/config.ts. PyYAML is used; we accept that block-level comments
+written by hand may not be preserved on save. Round-trip safety is achieved
+by re-parsing before writing.
+
+Mirrors the PROVIDER_CATALOG in src/core/config.ts. Keep them in sync.
+"""
+
+from __future__ import annotations
+
+import os
+import tempfile
+from typing import Any, Dict, List, Optional, Tuple
+
+try:
+    import yaml as _yaml  # type: ignore
+    HAS_PYYAML = True
+except Exception:
+    HAS_PYYAML = False
+
+
+ALLOWED_EFFORTS = ("low", "medium", "high", "xhigh")
+ALLOWED_ROLES = ("chat", "builder")
+
+
+# Mirrors PROVIDER_CATALOG in src/core/config.ts.
+PROVIDER_CATALOG: List[Dict[str, Any]] = [
+    {
+        "id": "codex",
+        "label": "Codex CLI",
+        "type": "cli",
+        "default_model": "gpt-5.5",
+        "needs_effort": True,
+        "default_chat_effort": "low",
+        "default_builder_effort": "xhigh",
+    },
+    {
+        "id": "claude-code",
+        "label": "Claude Code CLI",
+        "type": "cli",
+        "default_model": "opus",
+        "needs_effort": True,
+        "default_chat_effort": "medium",
+        "default_builder_effort": "xhigh",
+    },
+    {
+        "id": "openai",
+        "label": "OpenAI API",
+        "type": "api",
+        "default_model": "gpt-5.5",
+        "needs_effort": False,
+    },
+    {
+        "id": "gemini",
+        "label": "Google Gemini API",
+        "type": "api",
+        "default_model": "gemini-2.5-flash",
+        "needs_effort": False,
+    },
+    {
+        "id": "anthropic",
+        "label": "Anthropic API",
+        "type": "api",
+        "default_model": "claude-opus-4-7",
+        "needs_effort": False,
+    },
+    {
+        "id": "ollama",
+        "label": "Ollama (local)",
+        "type": "ollama",
+        "default_model": "gemma4:26b",
+        "needs_effort": False,
+    },
+]
+
+
+PROVIDER_INDEX: Dict[str, Dict[str, Any]] = {p["id"]: p for p in PROVIDER_CATALOG}
+
+
+def models_path(workspace: str) -> str:
+    return os.path.join(workspace, "models.yaml")
+
+
+def load_models(workspace: str) -> Dict[str, Any]:
+    """Returns {"roles": {"chat": id?, "builder": id?}, "models": {id: entry, ...}}."""
+    path = models_path(workspace)
+    if not os.path.exists(path):
+        return {"roles": {}, "models": {}}
+    with open(path, "r", encoding="utf-8") as f:
+        raw = f.read()
+    if not raw.strip():
+        return {"roles": {}, "models": {}}
+    if not HAS_PYYAML:
+        raise RuntimeError("PyYAML is required to read models.yaml")
+    data = _yaml.safe_load(raw) or {}
+    if not isinstance(data, dict):
+        return {"roles": {}, "models": {}}
+    roles = data.get("roles") if isinstance(data.get("roles"), dict) else {}
+    models = data.get("models") if isinstance(data.get("models"), dict) else {}
+    # Normalize legacy "login" -> "cli" in memory.
+    normalized: Dict[str, Any] = {}
+    for entry_id, entry in models.items():
+        if not isinstance(entry, dict):
+            continue
+        copy = dict(entry)
+        if copy.get("type") == "login":
+            copy["type"] = "cli"
+        normalized[str(entry_id)] = copy
+    return {"roles": dict(roles), "models": normalized}
+
+
+def save_models(workspace: str, data: Dict[str, Any]) -> str:
+    """Write models.yaml. May drop hand-written comments."""
+    if not HAS_PYYAML:
+        raise RuntimeError("PyYAML is required to write models.yaml")
+    path = models_path(workspace)
+    os.makedirs(os.path.dirname(path) or ".", exist_ok=True)
+    payload: Dict[str, Any] = {}
+    roles = data.get("roles") or {}
+    if isinstance(roles, dict) and roles:
+        payload["roles"] = {k: v for k, v in roles.items() if v}
+    payload["models"] = data.get("models") or {}
+    text = _yaml.safe_dump(
+        payload,
+        allow_unicode=True,
+        sort_keys=False,
+        default_flow_style=False,
+    )
+    # round-trip check before writing
+    parsed = _yaml.safe_load(text) or {}
+    if not isinstance(parsed.get("models"), dict) or len(parsed["models"]) != len(payload["models"]):
+        raise RuntimeError("Round-trip check failed when serializing models.yaml")
+    fd, tmp_path = tempfile.mkstemp(
+        prefix=".models.", suffix=".yaml.tmp", dir=os.path.dirname(path) or "."
+    )
+    try:
+        with os.fdopen(fd, "w", encoding="utf-8") as f:
+            f.write(text)
+        os.replace(tmp_path, path)
+    except Exception:
+        try:
+            os.unlink(tmp_path)
+        except Exception:
+            pass
+        raise
+    return path
+
+
+def derive_id(provider: str, effort: Optional[str], type_: str) -> str:
+    """Mirror src/core/config.ts deriveSetupId."""
+    if type_ == "cli" and effort:
+        return f"{provider}-{effort}"
+    if type_ == "ollama":
+        return "ollama" if provider == "ollama" else provider
+    return provider
+
+
+def unique_id(base: str, existing: set) -> str:
+    if base not in existing:
+        return base
+    n = 2
+    while f"{base}-{n}" in existing:
+        n += 1
+    return f"{base}-{n}"
+
+
+def build_entry(
+    catalog_entry: Dict[str, Any],
+    model: Optional[str],
+    effort: Optional[str],
+) -> Dict[str, Any]:
+    # フィールド順は既存 models.yaml に合わせる: type / provider / model / effort / enabled
+    entry: Dict[str, Any] = {"type": catalog_entry["type"]}
+    if catalog_entry["type"] != "ollama":
+        entry["provider"] = catalog_entry["id"]
+    if model:
+        entry["model"] = model
+    if effort:
+        entry["effort"] = effort
+    entry["enabled"] = True
+    return entry
+
+
+def find_provider(provider: str) -> Optional[Dict[str, Any]]:
+    return PROVIDER_INDEX.get((provider or "").strip().lower())
+
+
+def normalize_effort(raw: Optional[str]) -> Optional[str]:
+    if raw is None:
+        return None
+    value = str(raw).strip().lower()
+    if not value:
+        return None
+    if value not in ALLOWED_EFFORTS:
+        raise ValueError(
+            f'effort must be one of: {"/".join(ALLOWED_EFFORTS)} (got "{raw}")'
+        )
+    return value
+
+
+def default_effort_for(catalog_entry: Dict[str, Any], role: str) -> Optional[str]:
+    if not catalog_entry.get("needs_effort"):
+        return None
+    if role == "chat":
+        return catalog_entry.get("default_chat_effort") or "low"
+    return catalog_entry.get("default_builder_effort") or "xhigh"
+
+
+def entry_summary(entry_id: str, entry: Dict[str, Any]) -> str:
+    parts: List[str] = []
+    provider = entry.get("provider") or entry.get("type") or ""
+    parts.append(str(provider))
+    if entry.get("model"):
+        parts.append(str(entry["model"]))
+    if entry.get("effort"):
+        parts.append(f"effort={entry['effort']}")
+    return f"{entry_id} ({' / '.join(parts)})"
+
+
+def entries_equivalent(a: Dict[str, Any], b: Dict[str, Any]) -> bool:
+    keys = ("type", "provider", "model", "effort")
+    for k in keys:
+        if (a.get(k) or None) != (b.get(k) or None):
+            return False
+    return True

package/builtin-skills/_shared/_profile_lib.py

@@ -0,0 +1,98 @@
+"""Shared helpers for profile-edit / profile-delete skills.
+
+soul.md / user.md / memory.md は profile-save が `\n## <timestamp>\n\n<text>\n`
+形式で追記する。これを「ヘッダー = '## ...' を境にしたエントリ列」として
+扱うためのパーサと再シリアライズを提供する。
+"""
+
+from __future__ import annotations
+
+import os
+import re
+import tempfile
+from typing import List, Optional, Tuple
+
+
+_HEADER_RE = re.compile(r"^##\s+(.+?)\s*$")
+
+
+def profile_file(workspace: str, target: str) -> str:
+    return os.path.join(workspace, "memory", "profile", f"{target}.md")
+
+
+def parse_profile(raw: str) -> Tuple[str, List[dict]]:
+    """Split into (preamble, entries[{ timestamp, text }]).
+
+    preamble は最初の '## ' より前の部分(ファイル冒頭のコメント等)。
+    各 entry の `text` は本文のみ(前後の改行は trim 済み)。
+    """
+    lines = raw.splitlines(keepends=False)
+    pre: List[str] = []
+    entries: List[dict] = []
+    current: Optional[dict] = None
+    for line in lines:
+        m = _HEADER_RE.match(line)
+        if m:
+            if current is not None:
+                current["text"] = "\n".join(current["_body"]).strip()
+                del current["_body"]
+                entries.append(current)
+            current = {"timestamp": m.group(1).strip(), "_body": []}
+            continue
+        if current is None:
+            pre.append(line)
+        else:
+            current["_body"].append(line)
+    if current is not None:
+        current["text"] = "\n".join(current["_body"]).strip()
+        del current["_body"]
+        entries.append(current)
+    preamble = "\n".join(pre).rstrip("\n")
+    return preamble, entries
+
+
+def serialize_profile(preamble: str, entries: List[dict]) -> str:
+    out: List[str] = []
+    if preamble:
+        out.append(preamble)
+    for entry in entries:
+        out.append("")
+        out.append(f"## {entry['timestamp']}")
+        out.append("")
+        out.append(entry["text"])
+    text = "\n".join(out)
+    if not text.endswith("\n"):
+        text += "\n"
+    return text
+
+
+def write_atomic(path: str, content: str) -> None:
+    os.makedirs(os.path.dirname(path), exist_ok=True)
+    fd, tmp = tempfile.mkstemp(
+        prefix=".profile.", suffix=".md.tmp", dir=os.path.dirname(path),
+    )
+    try:
+        with os.fdopen(fd, "w", encoding="utf-8") as f:
+            f.write(content)
+        os.replace(tmp, path)
+    except Exception:
+        try:
+            os.unlink(tmp)
+        except Exception:
+            pass
+        raise
+
+
+def find_entry_index(entries: List[dict], *, index: Optional[int], timestamp: Optional[str]) -> int:
+    if timestamp:
+        for i, entry in enumerate(entries):
+            if entry["timestamp"] == timestamp:
+                return i
+        raise LookupError(f'timestamp "{timestamp}" にマッチするエントリがありません')
+    if index is not None:
+        if index < 1 or index > len(entries):
+            raise LookupError(
+                f"index {index} は範囲外です (entries={len(entries)})",
+            )
+        return index - 1
+    raise LookupError("index か timestamp のどちらかを指定してください")