feat(my-deepagent): v0.1.0 Step 0~5 — scaffolding through deepagent + OpenRouter

Python rewrite of the agent harness on top of deepagents 0.6.1 + langchain 1.x,
replacing the abandoned TS attempt in packages/. 388 unit/integration tests pass.

Steps
-----
0. Scaffolding — uv workspace, ruff/mypy/pre-commit/alembic, src/tests/docs
   trees with docs/schemas/ seeded from my-deepagent-seed/.
1. Core — config (pydantic-settings with MYDEEPAGENT_ env prefix and TOML
   source), enums (Backend, Capability, RiskLevel, ApprovalDecisionAction,
   ApprovalState, RunState, RunPhaseState, SessionState, ErrorClass),
   errors (MyDeepAgentError + BudgetExhaustedError with PEP-3134 cause +
   context suppression), hash (canonical JSON + sha256).
2. Persona/Workflow/Binding — pydantic v2 schemas with tuple-based deep
   immutability (post-construction hash drift prevented), YAML loaders,
   deterministic auto-select (preferred_backends → version → name → hash),
   override resolution with ineligibility diagnostics, PersonaConsentStore
   with fcntl.flock + tmp+fsync+rename atomic write.
3. Artifact schema registry — Draft202012Validator, multi-root resolution,
   structured ValidationFinding output.
4. Persistence — 18 SQLAlchemy 2.0 async ORM models with FK CASCADE/RESTRICT,
   WAL + busy_timeout + foreign_keys PRAGMA, alembic baseline +
   ux_active_run_repo_base partial unique index, LangGraph SqliteSaver as
   context manager only (lifecycle safety).
5. DeepAgent session — build_agent wires Persona → create_deep_agent with
   LocalShellBackend / FilesystemBackend / StateBackend / CompositeBackend,
   ChatOpenAI(base_url=openrouter) for openrouter: model strings, and 4
   middleware classes (cost / audit-tool / safety-shell / fallback-model).

Critical workarounds
--------------------
- deepagents 0.6.1 rejects FilesystemPermission together with backends that
  implement SandboxBackendProtocol (LocalShellBackend). SafetyShellMiddleware
  enforces destructive-command and secret-path policy at the tool layer
  instead, and build_agent strips the permissions kwarg when the persona's
  deepagents_backend is local_shell.
- FilesystemOperation in deepagents is Literal['read', 'write'] only;
  _map_operations collapses our richer schema (read/write/edit/ls) safely.

Real OpenRouter smoke
---------------------
test_openrouter_deepagents_local_shell_smoke calls DeepSeek via deepagents +
LocalShellBackend + SafetyShellMiddleware end-to-end. PASS, ~$0.000001 cost,
input=9 / output=1 tokens with content "OK".

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
chungyeong
2026-05-15 19:40:02 +09:00
parent 1fe59d16ca
commit 17ba5d723b
100 changed files with 12408 additions and 0 deletions

View File

@@ -0,0 +1,238 @@
"""Unit tests for src/my_deepagent/config.py."""
from __future__ import annotations
from pathlib import Path
import pytest
from pydantic import ValidationError
from my_deepagent.config import Config, load_config
# ---------------------------------------------------------------------------
# Default values (no env, no file)
# ---------------------------------------------------------------------------
def test_default_log_level(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
cfg = Config()
assert cfg.log_level == "info"
def test_default_lang(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
cfg = Config()
assert cfg.lang == "ko"
def test_default_budget_daily_usd(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
cfg = Config()
assert cfg.budget_daily_usd == pytest.approx(5.0)
def test_default_budget_run_usd(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
cfg = Config()
assert cfg.budget_run_usd == pytest.approx(1.0)
def test_default_budget_on_hit(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
cfg = Config()
assert cfg.budget_on_hit == "prompt"
def test_default_persona(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
cfg = Config()
assert cfg.default_persona == "default-interactive"
def test_default_openrouter_api_key_is_none(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
# _env_file=None bypasses any .env that may exist in the cwd (e.g. dev keys).
cfg = Config(_env_file=None) # type: ignore[call-arg]
assert cfg.openrouter_api_key is None
# ---------------------------------------------------------------------------
# Env var overrides
# ---------------------------------------------------------------------------
def test_env_budget_daily_usd(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_BUDGET_DAILY_USD", "10")
cfg = Config()
assert cfg.budget_daily_usd == pytest.approx(10.0)
def test_env_lang_en(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_LANG", "en")
cfg = Config()
assert cfg.lang == "en"
def test_env_log_level_debug(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_LOG_LEVEL", "debug")
cfg = Config()
assert cfg.log_level == "debug"
def test_env_openrouter_api_key(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_OPENROUTER_API_KEY", "sk-test-abc")
cfg = Config()
assert cfg.openrouter_api_key == "sk-test-abc"
def test_env_langsmith_tracing(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_LANGSMITH_TRACING", "true")
cfg = Config()
assert cfg.langsmith_tracing is True
# ---------------------------------------------------------------------------
# Validation errors for invalid values
# ---------------------------------------------------------------------------
def test_invalid_lang_raises(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_LANG", "fr")
with pytest.raises(ValidationError):
Config()
def test_invalid_log_level_raises(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_LOG_LEVEL", "verbose")
with pytest.raises(ValidationError):
Config()
def test_invalid_budget_on_hit_raises(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_BUDGET_ON_HIT", "explode")
with pytest.raises(ValidationError):
Config()
def test_negative_budget_raises(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
with pytest.raises(ValidationError):
Config(budget_daily_usd=-1.0)
# ---------------------------------------------------------------------------
# Frozen check
# ---------------------------------------------------------------------------
def test_frozen_prevents_mutation(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
cfg = Config()
with pytest.raises((ValidationError, TypeError)):
cfg.budget_daily_usd = 99 # type: ignore[misc]
# ---------------------------------------------------------------------------
# Path expansion (~ → absolute path)
# ---------------------------------------------------------------------------
def test_tilde_expansion_workspace_root(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_WORKSPACE_ROOT", "~/foo/bar")
cfg = Config()
assert cfg.workspace_root.is_absolute()
assert "~" not in str(cfg.workspace_root)
def test_tilde_expansion_data_dir(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
monkeypatch.setenv("MYDEEPAGENT_DATA_DIR", "~/mydata")
cfg = Config()
assert cfg.data_dir.is_absolute()
# ---------------------------------------------------------------------------
# TOML priority
# ---------------------------------------------------------------------------
def test_toml_overrides_default(monkeypatch: pytest.MonkeyPatch, tmp_path: Path) -> None:
_clear_env(monkeypatch)
toml_file = tmp_path / "config.toml"
toml_file.write_text('lang = "en"\nbudget_daily_usd = 7.5\n')
# Patch the toml_file location via init override
# Config reads toml via SettingsConfigDict; we pass via class-level override trick:
# Easiest approach: pass budget_daily_usd and lang directly to assert TOML *can* set them.
# For true TOML path injection, subclass Config temporarily.
class PatchedConfig(Config):
model_config = Config.model_config.copy()
PatchedConfig.model_config["toml_file"] = str(toml_file)
cfg = PatchedConfig()
assert cfg.lang == "en"
assert cfg.budget_daily_usd == pytest.approx(7.5)
# ---------------------------------------------------------------------------
# load_config helper
# ---------------------------------------------------------------------------
def test_load_config_with_overrides(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
cfg = load_config(budget_daily_usd=20.0, lang="en")
assert cfg.budget_daily_usd == pytest.approx(20.0)
assert cfg.lang == "en"
def test_load_config_default(monkeypatch: pytest.MonkeyPatch) -> None:
_clear_env(monkeypatch)
cfg = load_config()
assert cfg.log_level == "info"
# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------
_ENV_KEYS = [
"MYDEEPAGENT_BUDGET_DAILY_USD",
"MYDEEPAGENT_BUDGET_DAILY_WARN_USD",
"MYDEEPAGENT_BUDGET_RUN_USD",
"MYDEEPAGENT_BUDGET_RUN_WARN_USD",
"MYDEEPAGENT_BUDGET_ON_HIT",
"MYDEEPAGENT_LANG",
"MYDEEPAGENT_LOG_LEVEL",
"MYDEEPAGENT_OPENROUTER_API_KEY",
"MYDEEPAGENT_OPENROUTER_BASE_URL",
"MYDEEPAGENT_LANGSMITH_TRACING",
"MYDEEPAGENT_LANGSMITH_API_KEY",
"MYDEEPAGENT_LANGSMITH_PROJECT",
"MYDEEPAGENT_DATABASE_URL",
"MYDEEPAGENT_WORKSPACE_ROOT",
"MYDEEPAGENT_DATA_DIR",
"MYDEEPAGENT_CONFIG_DIR",
"MYDEEPAGENT_STATE_DIR",
"MYDEEPAGENT_DEFAULT_PERSONA",
]
def _clear_env(monkeypatch: pytest.MonkeyPatch) -> None:
"""Remove all MYDEEPAGENT_ env vars to isolate tests from the real environment."""
for key in _ENV_KEYS:
monkeypatch.delenv(key, raising=False)
# Also prevent dotenv file from being loaded
monkeypatch.setenv("MYDEEPAGENT_ENV_FILE", "")