Buckets:

download
raw
15.6 kB
# /// script
# dependencies = []
# ///
"""Generate values.schema.json for the Fred Helm chart.
Inlines the three backend configuration JSON schemas (already generated by
generate_config_schema.py) into the correct positions of a hand-written
chart-level schema. The result is a single self-contained JSON Schema
(draft-07, no $ref) that Helm and IDE plugins can use for validation and
autocompletion.
Usage (called automatically by python-config-schema.mk):
python scripts/generate_chart_schema.py \\
--fred-agents apps/fred-agents/config/schema/configuration.schema.json \\
--knowledge-flow apps/knowledge-flow-backend/config/schema/configuration.schema.json \\
--control-plane apps/control-plane-backend/config/schema/configuration.schema.json \\
--output deploy/charts/fred/values.schema.json
"""
import argparse
import copy
import json
import sys
from pathlib import Path
# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------
def _load(path: Path) -> dict:
with open(path) as f:
return json.load(f)
def _inline_refs(schema: dict) -> dict:
"""Resolve all $ref pointers within a self-contained schema that uses $defs.
Uses a `seen` set to break cycles: if a definition is currently being
expanded, replace the back-reference with an empty object rather than
looping indefinitely.
"""
schema = copy.deepcopy(schema)
defs = schema.get("$defs", {})
def _resolve(node: object, seen: frozenset[str] = frozenset()) -> object:
if isinstance(node, dict):
if "$ref" in node:
ref = node["$ref"]
if ref.startswith("#/$defs/"):
name = ref[len("#/$defs/"):]
if name in seen:
return {} # cycle guard: emit empty object
if name not in defs:
raise ValueError(
f"Invalid schema: $ref '{ref}' points to missing definition '{name}' in $defs"
)
return _resolve(copy.deepcopy(defs[name]), seen | {name})
return node
return {k: _resolve(v, seen) for k, v in node.items() if k != "$defs"}
if isinstance(node, list):
return [_resolve(i, seen) for i in node]
return node
resolved = _resolve(schema)
resolved.pop("$defs", None)
return resolved
# ---------------------------------------------------------------------------
# Shared Kubernetes / Helm sub-schemas
# ---------------------------------------------------------------------------
_STRING = {"type": "string"}
_BOOL = {"type": "boolean"}
_INT = {"type": "integer"}
_OBJECT_FREE = {"type": "object"} # used only where the shape is truly open (annotations, labels)
_ANY = {}
def _obj(props: dict, required: list[str] | None = None, additional: bool = False) -> dict:
s: dict = {"type": "object", "properties": props, "additionalProperties": additional}
if required:
s["required"] = required
return s
def _arr(items: dict) -> dict:
return {"type": "array", "items": items}
# ── Kubernetes primitives ────────────────────────────────────────────────────
_IMAGE = _obj({
"repository": _STRING,
"tag": _STRING,
"pullPolicy": {"type": "string", "enum": ["Always", "IfNotPresent", "Never"]},
})
_ROLLING_UPDATE = _obj({
"maxSurge": {"anyOf": [_INT, _STRING]},
"maxUnavailable": {"anyOf": [_INT, _STRING]},
})
_PORT_DEF = _obj({
"name": _STRING,
"containerPort": _INT,
"protocol": _STRING,
}, additional=True)
_SERVICE_PORT = _obj({
"name": _STRING,
"port": _INT,
"targetPort": {"anyOf": [_INT, _STRING]},
"protocol": _STRING,
}, additional=True)
_SERVICE_MONITOR = _obj({
"enabled": _BOOL,
"annotations": _OBJECT_FREE,
"labels": _OBJECT_FREE,
"interval": _STRING,
"scrapeTimeout": _STRING,
"path": _STRING,
"scheme": _STRING,
"honorLabels": _BOOL,
"relabelings": _arr(_ANY),
"metricRelabelings": _arr(_ANY),
"namespaceSelector": _OBJECT_FREE,
})
_SERVICE = _obj({
"enabled": _BOOL,
"annotations": _OBJECT_FREE,
"extraLabels": _OBJECT_FREE,
"type": _STRING,
"ports": _arr(_SERVICE_PORT),
})
_METRICS_SERVICE = _obj({
"enabled": _BOOL,
"annotations": _OBJECT_FREE,
"extraLabels": _OBJECT_FREE,
"type": _STRING,
"port": _INT,
"targetPort": {"anyOf": [_INT, _STRING]},
"serviceMonitor": _SERVICE_MONITOR,
})
_INGRESS_HOST_PATH = _obj({
"path": _STRING,
"pathType": _STRING,
"service": _obj({
"name": _STRING,
"port": {"anyOf": [_INT, _STRING]},
}, additional=True),
}, additional=True)
_INGRESS_HOST = _obj({
"host": _STRING,
"paths": _arr(_INGRESS_HOST_PATH),
}, additional=True)
_INGRESS_TLS = _obj({
"secretName": _STRING,
"hosts": _arr(_STRING),
}, additional=True)
_INGRESS = _obj({
"enabled": _BOOL,
"className": _STRING,
"hosts": _arr(_INGRESS_HOST),
"tls": _arr(_INGRESS_TLS),
}, additional=True)
# ── Gateway API HTTPRoute ────────────────────────────────────────────────────
# parentRefs and rules are passed through to the HTTPRoute spec verbatim
# (templates/http-route.yaml uses `toYaml`), so their items stay free-form to
# accept the full Gateway API shape without re-modelling it here.
_HTTP_ROUTE_PARENT_REF = _obj({
"group": _STRING,
"kind": _STRING,
"name": _STRING,
"namespace": _STRING,
"sectionName": _STRING,
"port": _INT,
}, additional=True)
_HTTP_ROUTE = _obj({
"enabled": _BOOL,
"annotations": _OBJECT_FREE,
"parentRefs": _arr(_HTTP_ROUTE_PARENT_REF),
"hostnames": _arr(_STRING),
"rules": _arr(_ANY),
}, additional=True)
_VOLUME_MOUNT = _obj({
"name": _STRING,
"mountPath": _STRING,
"subPath": _STRING,
"readOnly": _BOOL,
}, additional=True)
_VOLUME = _obj({
"name": _STRING,
}, additional=True)
_PROBE_DATA = _obj({}, additional=True)
_PROBE_ENTRY = _obj({
"enabled": _BOOL,
"data": _PROBE_DATA,
})
_PROBES = _obj({
"lifecycle": _obj({"enabled": _BOOL}, additional=True),
"livenessProbe": _PROBE_ENTRY,
"readinessProbe": _PROBE_ENTRY,
"startupProbe": _PROBE_ENTRY,
})
_SECURITY_CONTEXT = _obj({
"allowPrivilegeEscalation": _BOOL,
"capabilities": _obj({
"drop": _arr(_STRING),
"add": _arr(_STRING),
}),
"runAsUser": _INT,
"runAsNonRoot": _BOOL,
"readOnlyRootFilesystem": _BOOL,
"seccompProfile": _obj({"type": _STRING}, additional=True),
})
_RBAC_RULE = _obj({
"apiGroups": _arr(_STRING),
"resources": _arr(_STRING),
"verbs": _arr(_STRING),
}, additional=True)
_RBAC = _obj({
"enabled": _BOOL,
"permissions": _obj({
"namespaced": _arr(_RBAC_RULE),
"cluster": _arr(_RBAC_RULE),
}, additional=True),
})
_SERVICE_ACCOUNT = _obj({
"enabled": _BOOL,
"name": _STRING,
"annotations": _OBJECT_FREE,
"labels": _OBJECT_FREE,
"automountServiceAccountToken": _BOOL,
"rbac": _RBAC,
})
_ENV_VAR = _obj({
"name": _STRING,
"value": _STRING,
"valueFrom": _obj({}, additional=True),
}, additional=True)
_MIGRATION = _obj({
"enabled": _BOOL,
"scaleDown": _BOOL,
"backoffLimit": _INT,
"command": _arr(_STRING),
"args": _arr(_STRING),
"resources": _OBJECT_FREE,
# Annotations for the migration Job pod only. Distinct from the app-level
# podAnnotations, which are shared with the Deployment: a service-mesh
# sidecar that is desirable on a long-running Deployment keeps a hook Job
# from ever completing, so the opt-out must target the Job alone.
"podAnnotations": _OBJECT_FREE,
})
_COMMAND = _obj({
"enabled": _BOOL,
"data": _arr(_STRING),
})
_CONFIGURATION_TYPE = _obj({
"backend": _BOOL,
"frontend": _BOOL,
})
_KUBECONFIG = _obj({
"enabled": _BOOL,
})
def _base_app_props(extra: dict | None = None) -> dict:
"""Properties common to every application entry."""
props = {
"enabled": _BOOL,
"applicationName": _STRING,
"configurationFileName": _STRING,
"deployment": _obj({"enabled": _BOOL}),
"replicaCount": _INT,
"statefulset": _obj({"enabled": _BOOL}),
"job": _obj({"enabled": _BOOL}, additional=True),
"migration": _MIGRATION,
"rollingUpdate": _ROLLING_UPDATE,
"image": _IMAGE,
"command": _COMMAND,
"env": _arr(_ENV_VAR),
"extraEnvVars": _arr(_ENV_VAR),
"ports": _arr(_PORT_DEF),
"service": _SERVICE,
"metricsService": _METRICS_SERVICE,
"ingress": _INGRESS,
"httpRoute": _HTTP_ROUTE,
"volumeMounts": _arr(_VOLUME_MOUNT),
"volumes": _arr(_VOLUME),
# Pod- and container-level keys below are read by templates/deployment.yaml,
# job.yaml and hook-migration.yaml, but were absent from the schema, so
# values files legitimately using them were rejected by `helm template`.
"extraVolumeMounts": _arr(_VOLUME_MOUNT),
"extraVolumes": _arr(_VOLUME),
"resources": _obj({"limits": _OBJECT_FREE, "requests": _OBJECT_FREE}),
"imagePullSecrets": _arr(_obj({"name": _STRING}, additional=True)),
"podAnnotations": _OBJECT_FREE,
"podSecurityContext": _OBJECT_FREE,
"envFrom": _arr(_obj({}, additional=True)),
"nodeSelector": _OBJECT_FREE,
"tolerations": _arr(_obj({}, additional=True)),
"probes": _PROBES,
"securityContext": _SECURITY_CONTEXT,
"serviceAccount": _SERVICE_ACCOUNT,
"kubeconfig": _KUBECONFIG,
"configuration_type": _CONFIGURATION_TYPE,
}
if extra:
props.update(extra)
return props
# ---------------------------------------------------------------------------
# Application-specific schemas
# ---------------------------------------------------------------------------
def _fred_agents_app(fa_config: dict) -> dict:
# models_catalog: free-form (large catalog structure, fully user-defined)
# mcp_catalog: free-form
# dotenv: key/value env secrets
return _obj(_base_app_props({
"configuration": fa_config,
"dotenv": _obj({}, additional=True),
"models_catalog": _obj({}, additional=True),
"mcp_catalog": _obj({}, additional=True),
}))
def _frontend_app() -> dict:
_frontend_config = _obj({
"config_json": _obj({
"frontend_basename": _STRING,
}, additional=True),
}, additional=True)
return _obj(_base_app_props({
"configuration": _frontend_config,
}))
def _kf_app(kf_config: dict) -> dict:
return _obj(_base_app_props({
"configuration": kf_config,
"dotenv": _obj({}, additional=True),
}))
def _kf_worker_app(kf_config: dict) -> dict:
return _obj(_base_app_props({
"configuration": kf_config,
"dotenv": _obj({}, additional=True),
}))
def _cp_app(cp_config: dict) -> dict:
_policy_entry = _obj({
"mode": _STRING,
"retention": _STRING,
"cancel_on_rejoin": _BOOL,
}, additional=True)
_policy_rule = _obj({
"rule_id": _STRING,
"match": _obj({
"team_id": _STRING,
"trigger": _STRING,
}, additional=True),
"action": _obj({
"retention": _STRING,
}, additional=True),
}, additional=True)
_purge_policy = _obj({
"default": _policy_entry,
"rules": _arr(_policy_rule),
})
_conversation_policy_catalog = _obj({
"version": _STRING,
"conversation_policies": _obj({
"purge": _purge_policy,
}, additional=True),
})
return _obj(_base_app_props({
"configuration": cp_config,
"conversation_policy_catalog": _conversation_policy_catalog,
"dotenv": _obj({}, additional=True),
}))
def _cp_worker_app(cp_config: dict) -> dict:
return _obj(_base_app_props({
"configuration": cp_config,
"dotenv": _obj({}, additional=True),
}))
# ---------------------------------------------------------------------------
# Global schema
# ---------------------------------------------------------------------------
_GLOBAL = _obj({
"kubeconfig": _STRING,
"persistence": _obj({
"enabled": _BOOL,
"dynamic": _BOOL,
"accessModes": _STRING,
"reclaimPolicy": _STRING,
"storageClass": _STRING,
"hostPath": _STRING,
}),
"whitelist": _obj({
"enabled": _BOOL,
"configMapName": _STRING,
"filePath": _STRING,
"user_list_inline": _STRING,
"mountPath": _STRING,
"subPath": _STRING,
}),
})
# ---------------------------------------------------------------------------
# Assembly
# ---------------------------------------------------------------------------
def build(fa_schema_path: Path, kf_schema_path: Path, cp_schema_path: Path) -> dict:
fa_config = _inline_refs(_load(fa_schema_path))
kf_config = _inline_refs(_load(kf_schema_path))
cp_config = _inline_refs(_load(cp_schema_path))
return {
"$schema": "https://json-schema.org/draft-07/schema#",
"title": "Fred Helm Chart Values",
"type": "object",
"additionalProperties": False,
# x-* keys are YAML anchors used for DRY value reuse. Helm resolves
# them before template rendering but they remain visible to schema
# validators, so we allow any x-* key with any value at the root.
"patternProperties": {"^x-": {}},
"properties": {
"global": _GLOBAL,
"applications": _obj({
"fred-agents": _fred_agents_app(fa_config),
"frontend": _frontend_app(),
"knowledge-flow-backend": _kf_app(kf_config),
"knowledge-flow-worker": _kf_worker_app(kf_config),
"control-plane-backend": _cp_app(cp_config),
"control-plane-worker": _cp_worker_app(cp_config),
}),
},
}
# ---------------------------------------------------------------------------
# CLI
# ---------------------------------------------------------------------------
def main() -> None:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--fred-agents", required=True, type=Path, metavar="SCHEMA")
parser.add_argument("--knowledge-flow", required=True, type=Path, metavar="SCHEMA")
parser.add_argument("--control-plane", required=True, type=Path, metavar="SCHEMA")
parser.add_argument("--output", required=True, type=Path, metavar="OUTPUT")
args = parser.parse_args()
for p in (args.fred_agents, args.knowledge_flow, args.control_plane):
if not p.exists():
print(f"ERROR: schema not found: {p}", file=sys.stderr)
sys.exit(1)
schema = build(args.fred_agents, args.knowledge_flow, args.control_plane)
args.output.parent.mkdir(parents=True, exist_ok=True)
with open(args.output, "w") as f:
json.dump(schema, f, indent=2)
f.write("\n")
print(f"Chart values schema written to {args.output}")
if __name__ == "__main__":
main()

Xet Storage Details

Size:
15.6 kB
·
Xet hash:
cf21b0d520ee29225ee6f795340ccf90909cd7cc5266dc991001864311d6aeeb

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.