cauldron/cauldron/forge.py
Kayos 130f96a34f v0.1 — backend bones + ingredient sterilizer
LAN-only Flask API that consumes Mealie (source of truth for recipes / plans
/ lists) and clawdforge (centralized claude -p runner) to do AI work.

v0.1 surface:
  GET  /healthz                          liveness + clawdforge upstream
  GET  /api/recipes                      proxy Mealie recipe list
  POST /api/sterilize/preview/<slug>     dry-run AI parse, return proposals
  POST /api/sterilize/apply/<slug>       write parses back to Mealie

Why sterilizer first: Mealie's CRF parser is mediocre and Cobb's hand-typed
recipes have lots of free-form ingredient strings ("about 2 cups cooked
white rice", "a pinch of salt") that don't aggregate cleanly into a
shopping list. We batch all ingredients of one recipe into a single Sonnet
call via clawdforge, get back parallel structured parses, then on apply
link each to Mealie food/unit records (creating missing by name) and PUT
the recipe back. Preview is non-destructive.

No UI in v0.1 — bearer-auth API only. Frontend + Authentik OIDC + Abby's
swamp/meadow/forest palette arrives in v0.2.

Auth: simple shared bearer in env (ADMIN_BEARER) until OIDC lands. LAN-only
deploy means the bearer is the only gate; no public exposure.

Stack: python:3.12-slim + Flask 3 + gunicorn + requests. No DB in v0.1.
2026-04-28 16:59:11 -07:00

64 lines
2 KiB
Python

"""Thin HTTP client for clawdforge — we're a consumer."""
import requests
class ForgeError(RuntimeError):
pass
class Forge:
def __init__(self, *, base_url: str, token: str, default_model: str, default_timeout: int):
self.base_url = base_url.rstrip("/")
self.token = token
self.default_model = default_model
self.default_timeout = default_timeout
def _headers(self) -> dict:
return {"Authorization": f"Bearer {self.token}"}
def healthz(self) -> dict:
r = requests.get(f"{self.base_url}/healthz", headers=self._headers(), timeout=10)
r.raise_for_status()
return r.json()
def run(
self,
prompt: str,
*,
model: str | None = None,
system: str | None = None,
files: list[str] | None = None,
timeout_secs: int | None = None,
) -> dict:
"""POST /run. Returns parsed result dict on success.
Raises ForgeError on transport or upstream failure. The 'result' field
in the return is whatever clawdforge parsed out of `claude -p` — usually
a dict (when the prompt asked for JSON), occasionally a string.
"""
body = {"prompt": prompt, "model": model or self.default_model}
if system:
body["system"] = system
if files:
body["files"] = files
if timeout_secs:
body["timeout_secs"] = timeout_secs
# HTTP timeout = subprocess timeout + a 30s margin so we don't bail
# while clawdforge is still doing work for us.
http_timeout = (timeout_secs or self.default_timeout) + 30
try:
r = requests.post(
f"{self.base_url}/run",
headers=self._headers(),
json=body,
timeout=http_timeout,
)
except requests.RequestException as e:
raise ForgeError(f"transport: {e}") from e
if r.status_code >= 400:
raise ForgeError(f"upstream {r.status_code}: {r.text[:500]}")
return r.json()