pulse-engine 0.2.18__tar.gz → 0.2.19__tar.gz

{pulse_engine-0.2.18 → pulse_engine-0.2.19}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pulse-engine
-Version: 0.2.18
+Version: 0.2.19
 Summary: Pulse Engine — Hybrid framework for building Pulse products
 Author: Pulse Team
 Requires-Python: >=3.11,<3.13

{pulse_engine-0.2.18 → pulse_engine-0.2.19}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "pulse-engine"
-version = "0.2.18"
+version = "0.2.19"
 description = "Pulse Engine — Hybrid framework for building Pulse products"
 authors = ["Pulse Team"]
 readme = "README.md"

{pulse_engine-0.2.18 → pulse_engine-0.2.19}/src/pulse_engine/adapters/report_generator.py

@@ -104,9 +104,9 @@ class ReportGeneratorAdapter:
         version_number?}) pins a skill; omit it to let the engine's agent auto-select
         from the tenant's active catalog. ``marts`` (list of mart-name stems) restricts
         which staged fixtures the agent may read; omit for all. ``artifact_type``
-        (html|pdf|docx|xlsx|pptx) pins the output format; omit to let the agent decide.
-        The fixture source is derived from ``tenant_id`` inside the service — never
-        caller-supplied.
+        (html|pdf|docx|xlsx|pptx|md) pins the output format; omit to let the agent
+        decide. The fixture source is derived from ``tenant_id`` inside the service —
+        never caller-supplied.
         """
         from pulse_engine.reporting.models import ReportRequest
 

{pulse_engine-0.2.18 → pulse_engine-0.2.19}/src/pulse_engine/reporting/models.py

@@ -8,7 +8,7 @@ from typing import Literal
 from pydantic import BaseModel, Field, field_validator
 
 # The artifact formats the reporting pipeline can emit (mirrors wiring._CONTENT_TYPES).
-ArtifactType = Literal["html", "pdf", "docx", "xlsx", "pptx"]
+ArtifactType = Literal["html", "pdf", "docx", "xlsx", "pptx", "md"]
 
 
 class SkillRef(BaseModel):
@@ -38,7 +38,7 @@ class ReportRequest(BaseModel):
     # injected into the kickoff and the run fails if the agent emits a different type.
     artifact_type: ArtifactType | None = Field(
         default=None,
-        description="Output format (html|pdf|docx|xlsx|pptx); None = agent decides",
+        description="Output format (html|pdf|docx|xlsx|pptx|md); None = agent decides",
     )
     # NOTE: the fixture source is derived server-side from the authenticated tenant
     # (see wiring.make_reader). It is deliberately NOT caller-supplied — accepting a
@@ -54,7 +54,7 @@ class ReportRequest(BaseModel):
 
 class ReportResult(BaseModel):
     artifact_url: str
-    artifact_type: str  # "html" (v1 default) | "pdf" | "docx" | "xlsx" | "pptx"
+    artifact_type: str  # "html" (v1 default) | "pdf" | "docx" | "xlsx" | "pptx" | "md"
     skill_id: str
     skill_version: int | None
     tokens: int

{pulse_engine-0.2.18 → pulse_engine-0.2.19}/src/pulse_engine/reporting/tools.py

@@ -33,7 +33,9 @@ class _WriteReportInput(BaseModel):
     path: str = Field(
         ..., description="Absolute path of the finished artifact in the sandbox"
     )
-    artifact_type: str = Field("html", description="html | pdf | docx | xlsx | pptx")
+    artifact_type: str = Field(
+        "html", description="html | pdf | docx | xlsx | pptx | md"
+    )
 
 
 def build_report_tools(

{pulse_engine-0.2.18 → pulse_engine-0.2.19}/src/pulse_engine/reporting/wiring.py

@@ -46,6 +46,7 @@ _CONTENT_TYPES = {
     "pptx": (
         "application/vnd.openxmlformats-officedocument.presentationml.presentation"
     ),
+    "md": "text/markdown",
 }
 
 

pulse_engine-0.2.19/src/pulse_engine/skills/seed.py

@@ -0,0 +1,141 @@
+"""Idempotent seeding of the Agent Skills catalog from captured JSON snapshots.
+
+Extracted from Alembic migration ``0017`` (see DD-0xx). The migration only ever
+inserted rows if absent (``ON CONFLICT (id) DO NOTHING``), so once seeded the
+catalog could never be refreshed without authoring a new migration. This seeder
+**upserts** by primary key instead: re-running inserts missing rows and refreshes
+existing ones, so the catalog stays current independently of schema versioning.
+
+Schema lives in migration ``0016``; this module owns the data. The catalog is
+reference content, not per-tenant data, so a single global seed is correct.
+
+Run it with ``make seed`` or ``python -m pulse_engine.skills.seed``.
+"""
+
+from __future__ import annotations
+
+import json
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any, cast
+
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from pulse_engine.skills.models import Skill, SkillFile
+
+# Columns each snapshot row carries (mirror migration 0017's table constructs).
+# ``id`` is handled separately as the conflict key; created_at/updated_at are left
+# to their server defaults / onupdate so refreshes never clobber creation time.
+_SKILL_COLS = ("frontmatter", "body_content", "is_active", "visibility")
+_FILE_COLS = ("skill_id", "relative_path", "s3_key", "content_type")
+
+DEFAULT_DATA_DIR = Path(__file__).resolve().parent / "seed_data"
+
+
+@dataclass
+class SeedResult:
+    """Counts from one seed run — insufficient to be interesting, handy for logs."""
+
+    skills_inserted: int = 0
+    skills_updated: int = 0
+    files_inserted: int = 0
+    files_updated: int = 0
+
+
+def _load(data_dir: Path, name: str) -> list[dict[str, Any]]:
+    return cast("list[dict[str, Any]]", json.loads((data_dir / name).read_text()))
+
+
+async def _upsert(
+    session: AsyncSession,
+    model: type[Skill] | type[SkillFile],
+    rows: list[dict[str, Any]],
+    cols: tuple[str, ...],
+) -> tuple[int, int]:
+    """Insert-or-refresh ``rows`` by primary key; return (inserted, updated).
+
+    ORM get-then-set rather than a dialect-specific ``ON CONFLICT`` so the same
+    path runs on Postgres (prod) and SQLite (tests). The row count (~150) makes
+    per-row SELECTs irrelevant to performance. Flushes at the end so a subsequent
+    child upsert sees the parents (FK order).
+    """
+    inserted = updated = 0
+    for row in rows:
+        existing = await session.get(model, row["id"])
+        if existing is None:
+            session.add(model(id=row["id"], **{c: row[c] for c in cols}))
+            inserted += 1
+        else:
+            for c in cols:
+                setattr(existing, c, row[c])
+            updated += 1
+    await session.flush()
+    return inserted, updated
+
+
+async def seed_skills(
+    session: AsyncSession, *, data_dir: Path = DEFAULT_DATA_DIR
+) -> SeedResult:
+    """Upsert the skills + skill_files catalog. Idempotent; refreshes on re-run.
+
+    Parents (``skills``) are upserted before children (``skill_files``) to satisfy
+    the FK. The caller owns the transaction — commit (or roll back) after this
+    returns. ``skill_versions`` is intentionally not seeded (empty at capture).
+    """
+    skills_inserted, skills_updated = await _upsert(
+        session, Skill, _load(data_dir, "skills.json"), _SKILL_COLS
+    )
+    files_inserted, files_updated = await _upsert(
+        session, SkillFile, _load(data_dir, "skill_files.json"), _FILE_COLS
+    )
+    return SeedResult(
+        skills_inserted=skills_inserted,
+        skills_updated=skills_updated,
+        files_inserted=files_inserted,
+        files_updated=files_updated,
+    )
+
+
+async def _run(data_dir: Path) -> SeedResult:
+    from pulse_engine.config import get_settings
+    from pulse_engine.database import build_async_engine, build_session_factory
+
+    settings = get_settings()
+    if not settings.database_url:
+        raise SystemExit("DATABASE_URL is required to seed the skills catalog")
+
+    engine = build_async_engine(settings.database_url)
+    try:
+        maker = build_session_factory(engine)
+        async with maker() as session:
+            result = await seed_skills(session, data_dir=data_dir)
+            await session.commit()
+        return result
+    finally:
+        await engine.dispose()
+
+
+def main() -> None:
+    import argparse
+    import asyncio
+
+    parser = argparse.ArgumentParser(
+        description="Seed/refresh the Agent Skills catalog (idempotent upsert)"
+    )
+    parser.add_argument(
+        "--data-dir",
+        type=Path,
+        default=DEFAULT_DATA_DIR,
+        help="Directory holding skills.json + skill_files.json "
+        f"(default: {DEFAULT_DATA_DIR})",
+    )
+    args = parser.parse_args()
+    result = asyncio.run(_run(args.data_dir))
+    print(
+        f"skills: +{result.skills_inserted} ~{result.skills_updated}  "
+        f"skill_files: +{result.files_inserted} ~{result.files_updated}"
+    )
+
+
+if __name__ == "__main__":
+    main()