ODS V3 Pre-Release: Public testing and refinement ahead of the official V3 launch. Turn your PC, Mac, or Linux box into a private AI server.
Python ai-agentsamdcomfyuidockerllama-cppllmlocal-ain8nnvidiaopen-webuiragself-hostedspeech-to-textstrix-halotext-to-speechworkflow-automation
"""Closed public projection for context status; no session text crosses here."""
import re
from image_message import matches_archived_message
def _text(value, maximum, pattern=r"^[^\x00-\x1f\x7f]+$"):
return isinstance(value, str) and 0 < len(value) <= maximum and re.fullmatch(pattern, value) is not None
def _number(value, maximum=100_000_000):
return type(value) is int and 0 <= value <= maximum
def project_context(value):
if not isinstance(value, dict) or type(value.get("schemaVersion")) is not int or value["schemaVersion"] != 1:
raise ValueError("invalid context")
if value.get("status") not in {"ready", "missing", "busy", "unavailable"}:
raise ValueError("invalid status")
revision = value.get("sessionRevision")
if revision is not None and not _text(revision, 160, r"^[A-Za-z0-9:_-]+$"):
raise ValueError("invalid revision")
result = {"schemaVersion": 1, "status": value["status"], "sessionRevision": revision, "context": None, "model": None}
context, model = value.get("context"), value.get("model")
if context is not None:
if (not isinstance(context, dict) or not _number(context.get("used"))
or not _number(context.get("window"), 10_000_000) or context["window"] < 1
or not _text(context.get("measuredAt"), 64, r"^[0-9TZ: .+\-]+$")):
raise ValueError("invalid context measurement")
result["context"] = {key: context[key] for key in ("used", "window", "measuredAt")}
if model is not None:
if (not isinstance(model, dict) or not _text(model.get("id"), 512) or not _text(model.get("provider"), 128)
or not _number(model.get("contextWindow"), 10_000_000) or model["contextWindow"] < 1
or context is not None and context["window"] != model["contextWindow"]):
raise ValueError("invalid model")
result["model"] = {key: model[key] for key in ("id", "provider", "contextWindow")}
if "routeFingerprint" in model:
if not _text(model["routeFingerprint"], 64, r"^[a-f0-9]{64}$"):
raise ValueError("invalid model route")
result["model"]["routeFingerprint"] = model["routeFingerprint"]
if "imageInput" in model:
if model["imageInput"] not in ("supported", "unsupported", "unknown"):
raise ValueError("invalid image capability")
result["model"]["imageInput"] = model["imageInput"]
if "imageRouteFingerprint" in model:
if not _text(model["imageRouteFingerprint"], 64, r"^[a-f0-9]{64}$"):
raise ValueError("invalid image route")
result["model"]["imageRouteFingerprint"] = model["imageRouteFingerprint"]
compact = value.get("compaction")
if (not isinstance(compact, dict) or compact.get("status") not in {"idle", "running", "completed", "skipped", "failed", "unknown"}
or not _number(compact.get("count"))):
raise ValueError("invalid compaction")
result["compaction"] = {key: compact[key] for key in ("status", "count")}
for key in ("requestId", "reason", "tokensBefore", "tokensAfter"):
item = compact.get(key)
if item is None:
continue
valid = (_text(item, 128, r"^[A-Za-z0-9_-]+$") if key == "requestId" else
_text(item, 96, r"^[a-z][a-z0-9-]*$") if key == "reason" else _number(item))
if not valid:
raise ValueError("invalid compaction detail")
result["compaction"][key] = item
history = value.get("history")
if (not isinstance(history, dict) or history.get("status") not in {"ready", "pending", "unknown"}
or not _number(history.get("acknowledgedMessages"), 2000)
or history.get("revision") is not None and not _text(history["revision"], 64, r"^[a-f0-9]{64}$")):
raise ValueError("invalid history")
result["history"] = {key: history[key] for key in ("status", "revision", "acknowledgedMessages")}
if history.get("reason") is not None:
if not _text(history["reason"], 96, r"^[a-z][a-z0-9-]*$"):
raise ValueError("invalid history reason")
result["history"]["reason"] = history["reason"]
return result
def valid_history_snapshot(data):
snapshot = data.get("history_snapshot")
if snapshot is None:
return True # Existing OpenAI-compatible clients remain supported.
if (not _text(data.get("request_id"), 128, r"^[A-Za-z0-9_-]+$")
or not isinstance(snapshot, dict) or set(snapshot) != {"schemaVersion", "messages"}
or type(snapshot["schemaVersion"]) is not int or snapshot["schemaVersion"] not in (1, 2)
or not isinstance(snapshot["messages"], list) or not 1 <= len(snapshot["messages"]) <= 2000):
return False
size = 0
identities = {}
for message in snapshot["messages"]:
if (not isinstance(message, dict) or set(message) not in ({"role", "content"}, {"role", "content", "images"})
or message["role"] not in {"user", "assistant"} or not isinstance(message["content"], str)):
return False
if "images" in message:
images = message["images"]
if (snapshot["schemaVersion"] != 2 or message["role"] != "user"
or not isinstance(images, list) or not 1 <= len(images) <= 4):
return False
seen = set()
for image in images:
if (not isinstance(image, dict) or set(image) != {"id", "sha256"}
or not _text(image["id"], 36, r"^img-[a-f0-9]{32}$")
or not _text(image["sha256"], 64, r"^[a-f0-9]{64}$")
or image["id"] in seen
or image["id"] in identities and identities[image["id"]] != image["sha256"]):
return False
seen.add(image["id"])
identities[image["id"]] = image["sha256"]
size += len(image["id"]) + len(image["sha256"])
size += len(message["content"].encode("utf-8"))
if size > 4 * 1024 * 1024:
return False
return bool(data.get("messages")) and matches_archived_message(data["messages"][-1], snapshot["messages"][-1]) and snapshot["messages"][-1]["role"] == "user"