feat: harden chat state by tenant

This commit is contained in:
thanhnv
2026-07-08 23:38:22 +09:00
parent b3b0544ba8
commit c800f7edf7
20 changed files with 380 additions and 54 deletions
@@ -13,6 +13,7 @@ import argparse
import hashlib
import json
import os
import subprocess
import sys
from datetime import datetime, timezone
@@ -31,6 +32,7 @@ def project_root() -> str:
ROOT = project_root()
HARNESS_ROOT = os.path.join(ROOT, "packages", "casan-harness")
TENANT_STORE = os.path.join(HARNESS_ROOT, "scripts", "bash", "tenant-store.sh")
def now_iso() -> str:
@@ -38,9 +40,26 @@ def now_iso() -> str:
def inbox_path() -> str:
return os.environ.get("CASAN_APPROVAL_INBOX_FILE") or os.path.join(
ROOT, ".specify", "level5", "approval-inbox.json"
)
if os.environ.get("CASAN_APPROVAL_INBOX_FILE"):
return guarded_override(os.environ["CASAN_APPROVAL_INBOX_FILE"])
if os.environ.get("CASAN_TENANT_ID"):
return tenant_path("approvals/approval-inbox.json")
return os.path.join(ROOT, ".specify", "level5", "approval-inbox.json")
def tenant_path(logical: str) -> str:
r = subprocess.run(["bash", TENANT_STORE, "resolve", logical], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return r.stdout.strip()
def guarded_override(path: str) -> str:
if path and os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "guard", path], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return path
def policy_path() -> str:
@@ -29,6 +29,7 @@ ROOT = project_root()
HARNESS_ROOT = os.path.join(ROOT, "packages", "casan-harness")
HARNESS_BIN = os.path.join(HARNESS_ROOT, "scripts", "bash")
RBAC = os.path.join(HARNESS_BIN, "rbac-check.py")
TENANT_STORE = os.path.join(HARNESS_BIN, "tenant-store.sh")
GENESIS_HASH = "0" * 64
@@ -44,16 +45,33 @@ def state_root() -> str:
return os.environ.get("CASAN_STATE_ROOT") or os.path.join(ROOT, ".specify")
def tenant_path(logical: str, fallback: str) -> str:
if os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "resolve", logical], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return r.stdout.strip()
return os.path.join(state_root(), fallback)
def guarded_override(path: str) -> str:
if path and os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "guard", path], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return path
def registry_path() -> str:
return os.environ.get("CASAN_AGENT_REGISTRY_FILE") or os.path.join(HARNESS_ROOT, "config", "agent-registry.yaml")
def audit_path() -> str:
return os.environ.get("CASAN_AGENT_BIND_AUDIT_LOG") or os.path.join(state_root(), "logs", "chat", "agent-bindings.jsonl")
return guarded_override(os.environ["CASAN_AGENT_BIND_AUDIT_LOG"]) if os.environ.get("CASAN_AGENT_BIND_AUDIT_LOG") else tenant_path("chat/agent-bindings.jsonl", "logs/chat/agent-bindings.jsonl")
def head_path() -> str:
return os.environ.get("CASAN_AGENT_BIND_AUDIT_HEAD") or os.path.join(state_root(), "logs", "chat", "agent-bindings-head.txt")
return guarded_override(os.environ["CASAN_AGENT_BIND_AUDIT_HEAD"]) if os.environ.get("CASAN_AGENT_BIND_AUDIT_HEAD") else tenant_path("chat/agent-bindings-head.txt", "logs/chat/agent-bindings-head.txt")
def load_registry():
@@ -149,6 +167,8 @@ def run_rbac(args):
def bind(args) -> int:
if args.tenant and args.tenant != "default":
os.environ["CASAN_TENANT_ID"] = args.tenant
registry = load_registry()
if registry.get("_error"):
out = decision_payload(args, "DENIED", registry["_error"])
@@ -35,30 +35,49 @@ HARNESS_BIN = os.path.join(HARNESS_ROOT, "scripts", "bash")
ROUTER = os.path.join(HARNESS_BIN, "prompt-mode-router.py")
SECURITY = os.path.join(HARNESS_BIN, "security-check.sh")
ACTION_GATE = os.path.join(HARNESS_BIN, "action-gate.sh")
TENANT_STORE = os.path.join(HARNESS_BIN, "tenant-store.sh")
TENANT_CRYPT = os.path.join(HARNESS_BIN, "tenant-crypt.sh")
def state_root() -> str:
return os.environ.get("CASAN_STATE_ROOT") or os.path.join(ROOT, ".specify")
def tenant_path(logical: str, fallback: str) -> str:
if os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "resolve", logical], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return r.stdout.strip()
return os.path.join(state_root(), fallback)
def guarded_override(path: str) -> str:
if path and os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "guard", path], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return path
def config_path() -> str:
return os.environ.get("CASAN_OPERATOR_ACTIONS_FILE") or os.path.join(HARNESS_ROOT, "config", "operator-actions.yaml")
def audit_path() -> str:
return os.environ.get("CASAN_CHAT_AUDIT_LOG") or os.path.join(state_root(), "logs", "chat", "chat-turns.jsonl")
return guarded_override(os.environ["CASAN_CHAT_AUDIT_LOG"]) if os.environ.get("CASAN_CHAT_AUDIT_LOG") else tenant_path("chat/chat-turns.jsonl", "logs/chat/chat-turns.jsonl")
def head_path() -> str:
return os.environ.get("CASAN_CHAT_AUDIT_HEAD") or os.path.join(state_root(), "logs", "chat", "chat-head.txt")
return guarded_override(os.environ["CASAN_CHAT_AUDIT_HEAD"]) if os.environ.get("CASAN_CHAT_AUDIT_HEAD") else tenant_path("chat/chat-head.txt", "logs/chat/chat-head.txt")
def metrics_path() -> str:
return os.environ.get("CASAN_CHAT_METRICS_LOG") or os.path.join(state_root(), "logs", "cost", "metrics.jsonl")
return guarded_override(os.environ["CASAN_CHAT_METRICS_LOG"]) if os.environ.get("CASAN_CHAT_METRICS_LOG") else tenant_path("telemetry/cost/metrics.jsonl", "logs/cost/metrics.jsonl")
def artifact_dir() -> str:
return os.path.join(state_root(), "logs", "chat", "operator-artifacts")
return tenant_path("chat/operator-artifacts", "logs/chat/operator-artifacts")
def now_iso() -> str:
@@ -181,9 +200,18 @@ def record_turn(base):
os.makedirs(os.path.dirname(head_path()), exist_ok=True)
with open(head_path(), "w", encoding="utf-8") as fh:
fh.write(record_hash + "\n")
encrypt_chat_audit_snapshot(path)
return rec
def encrypt_chat_audit_snapshot(path: str):
if not os.environ.get("CASAN_TENANT_ID"):
return
if not os.path.isfile(path):
return
subprocess.run(["bash", TENANT_CRYPT, "encrypt", path, path + ".enc"], cwd=ROOT, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
def record_metrics(trace_id: str, message: str, answer: str, status: str, latency_ms: int, action_id: str):
input_tokens = len(message.split())
output_tokens = len(answer.split())
@@ -287,6 +315,8 @@ def list_actions(args) -> int:
def run(args) -> int:
if args.tenant and args.tenant != "default":
os.environ["CASAN_TENANT_ID"] = args.tenant
started = datetime.now(timezone.utc)
trace_id = str(uuid.uuid4())
chat_id = args.chat_id or "chat-default"
@@ -39,22 +39,41 @@ ROOT = project_root()
HARNESS_BIN = os.path.join(ROOT, "packages", "casan-harness", "scripts", "bash")
ROUTER = os.path.join(HARNESS_BIN, "prompt-mode-router.py")
SECURITY = os.path.join(HARNESS_BIN, "security-check.sh")
TENANT_STORE = os.path.join(HARNESS_BIN, "tenant-store.sh")
TENANT_CRYPT = os.path.join(HARNESS_BIN, "tenant-crypt.sh")
def state_root() -> str:
return os.environ.get("CASAN_STATE_ROOT") or os.path.join(ROOT, ".specify")
def tenant_path(logical: str, fallback: str) -> str:
if os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "resolve", logical], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return r.stdout.strip()
return os.path.join(state_root(), fallback)
def guarded_override(path: str) -> str:
if path and os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "guard", path], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return path
def audit_path() -> str:
return os.environ.get("CASAN_CHAT_AUDIT_LOG") or os.path.join(state_root(), "logs", "chat", "chat-turns.jsonl")
return guarded_override(os.environ["CASAN_CHAT_AUDIT_LOG"]) if os.environ.get("CASAN_CHAT_AUDIT_LOG") else tenant_path("chat/chat-turns.jsonl", "logs/chat/chat-turns.jsonl")
def head_path() -> str:
return os.environ.get("CASAN_CHAT_AUDIT_HEAD") or os.path.join(state_root(), "logs", "chat", "chat-head.txt")
return guarded_override(os.environ["CASAN_CHAT_AUDIT_HEAD"]) if os.environ.get("CASAN_CHAT_AUDIT_HEAD") else tenant_path("chat/chat-head.txt", "logs/chat/chat-head.txt")
def metrics_path() -> str:
return os.environ.get("CASAN_CHAT_METRICS_LOG") or os.path.join(state_root(), "logs", "cost", "metrics.jsonl")
return guarded_override(os.environ["CASAN_CHAT_METRICS_LOG"]) if os.environ.get("CASAN_CHAT_METRICS_LOG") else tenant_path("telemetry/cost/metrics.jsonl", "logs/cost/metrics.jsonl")
def now_iso() -> str:
@@ -223,9 +242,18 @@ def record_turn(base):
os.makedirs(os.path.dirname(head_path()), exist_ok=True)
with open(head_path(), "w", encoding="utf-8") as fh:
fh.write(record_hash + "\n")
encrypt_chat_audit_snapshot(path)
return rec
def encrypt_chat_audit_snapshot(path: str):
if not os.environ.get("CASAN_TENANT_ID"):
return
if not os.path.isfile(path):
return
subprocess.run(["bash", TENANT_CRYPT, "encrypt", path, path + ".enc"], cwd=ROOT, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
def record_metrics(trace_id: str, message: str, answer: str, status: str, latency_ms: int):
input_tokens = len(message.split())
output_tokens = len(answer.split())
@@ -253,6 +281,8 @@ def record_metrics(trace_id: str, message: str, answer: str, status: str, latenc
def ask(args):
if args.tenant and args.tenant != "default":
os.environ["CASAN_TENANT_ID"] = args.tenant
started = datetime.now(timezone.utc)
trace_id = str(uuid.uuid4())
message = args.message
@@ -33,14 +33,32 @@ def project_root() -> str:
ROOT = project_root()
BIN = os.path.join(ROOT, "packages", "casan-harness", "scripts", "bash")
LOOP_TRACE = os.path.join(BIN, "loop-trace.py")
TENANT_STORE = os.path.join(BIN, "tenant-store.sh")
def state_root() -> str:
return os.environ.get("CASAN_STATE_ROOT") or os.path.join(ROOT, ".specify")
def tenant_path(logical: str, fallback: str) -> str:
if os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "resolve", logical], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return r.stdout.strip()
return os.path.join(state_root(), fallback)
def guarded_override(path: str) -> str:
if path and os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "guard", path], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return path
def audit_path() -> str:
return os.environ.get("CASAN_CHAT_AUDIT_LOG") or os.path.join(state_root(), "logs", "chat", "chat-turns.jsonl")
return guarded_override(os.environ["CASAN_CHAT_AUDIT_LOG"]) if os.environ.get("CASAN_CHAT_AUDIT_LOG") else tenant_path("chat/chat-turns.jsonl", "logs/chat/chat-turns.jsonl")
def sha_text(text: str) -> str:
@@ -87,12 +105,12 @@ def resolve_path(path: str) -> str:
def loop_state_root(loop_run):
explicit = os.environ.get("CASAN_LOOP_STATE_ROOT")
if explicit:
return explicit
return guarded_override(explicit)
artifact = loop_run.get("artifact") or ""
marker = os.sep + "logs" + os.sep + "chat" + os.sep + "loop-runs" + os.sep
if marker in artifact:
return artifact.split(marker, 1)[0] + os.sep + "logs" + os.sep + "chat" + os.sep + "loop-state"
return os.path.join(state_root(), "logs", "chat", "loop-state")
return tenant_path("chat/loop-state", "logs/chat/loop-state")
def run_loop_trace(loop_run, cmd: str):
@@ -38,22 +38,43 @@ PREFLIGHT = os.path.join(BIN, "harness-preflight.sh")
CONTEXT_SCAN = os.path.join(BIN, "context-assemble-scan.sh")
TOOL_OUTPUT_SCAN = os.path.join(BIN, "tool-output-scan.sh")
ARTIFACT_SCAN = os.path.join(BIN, "artifact-scan.sh")
TENANT_STORE = os.path.join(BIN, "tenant-store.sh")
TENANT_CRYPT = os.path.join(BIN, "tenant-crypt.sh")
KILL_SWITCH = os.path.join(BIN, "kill-switch.sh")
COST_SPIKE = os.path.join(BIN, "cost-spike-detect.sh")
def state_root() -> str:
return os.environ.get("CASAN_STATE_ROOT") or os.path.join(ROOT, ".specify")
def tenant_path(logical: str, fallback: str) -> str:
if os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "resolve", logical], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return r.stdout.strip()
return os.path.join(state_root(), fallback)
def guarded_override(path: str) -> str:
if path and os.environ.get("CASAN_TENANT_ID"):
r = subprocess.run(["bash", TENANT_STORE, "guard", path], cwd=ROOT, capture_output=True, text=True)
if r.returncode != 0:
raise SystemExit((r.stderr or r.stdout or "TENANT_DENIED").strip())
return path
def now_iso() -> str:
return datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
def audit_path() -> str:
return os.environ.get("CASAN_CHAT_AUDIT_LOG") or os.path.join(state_root(), "logs", "chat", "chat-turns.jsonl")
return guarded_override(os.environ["CASAN_CHAT_AUDIT_LOG"]) if os.environ.get("CASAN_CHAT_AUDIT_LOG") else tenant_path("chat/chat-turns.jsonl", "logs/chat/chat-turns.jsonl")
def head_path() -> str:
return os.environ.get("CASAN_CHAT_AUDIT_HEAD") or os.path.join(state_root(), "logs", "chat", "chat-head.txt")
return guarded_override(os.environ["CASAN_CHAT_AUDIT_HEAD"]) if os.environ.get("CASAN_CHAT_AUDIT_HEAD") else tenant_path("chat/chat-head.txt", "logs/chat/chat-head.txt")
def sha(text: str) -> str:
@@ -147,11 +168,11 @@ def write_text(path: str, text: str):
def loop_dir(run_id: str) -> str:
return os.path.join(state_root(), "logs", "chat", "loop-runs", run_id)
return tenant_path(f"chat/loop-runs/{run_id}", f"logs/chat/loop-runs/{run_id}")
def codegen_dir(run_id: str) -> str:
return os.path.join(state_root(), "logs", "chat", "codegen-artifacts", run_id)
return tenant_path(f"chat/codegen-artifacts/{run_id}", f"logs/chat/codegen-artifacts/{run_id}")
def load_chat_head() -> str:
@@ -177,9 +198,65 @@ def append_chat_turn(base):
os.makedirs(os.path.dirname(head_path()), exist_ok=True)
with open(head_path(), "w", encoding="utf-8") as fh:
fh.write(record_hash + "\n")
encrypt_chat_audit_snapshot(path)
return rec
def encrypt_chat_audit_snapshot(path: str):
if not os.environ.get("CASAN_TENANT_ID"):
return
if not os.path.isfile(path):
return
enc = path + ".enc"
subprocess.run(["bash", TENANT_CRYPT, "encrypt", path, enc], cwd=ROOT, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
def tenant_runtime_guard(args):
if args.tenant and args.tenant != "default":
os.environ["CASAN_TENANT_ID"] = args.tenant
if os.environ.get("CASAN_TENANT_ID"):
ks = subprocess.run(["bash", KILL_SWITCH, "check", "tenant", os.environ["CASAN_TENANT_ID"]], cwd=ROOT, capture_output=True, text=True)
if ks.returncode != 0:
print(json.dumps({
"success": False,
"mode": "BLOCK",
"risk": "high",
"decision": "HALTED",
"answer": f"Tenant kill-switch active for {os.environ['CASAN_TENANT_ID']}.",
"sources": [],
"certified": False,
"audit": {},
"router": {"reason": "tenant_kill_switch_active", "matched_rules": ["tenant_kill_switch"]},
}, ensure_ascii=False))
return 3
if os.environ.get("CASAN_COST_CUMULATIVE_BUDGET_TOKENS"):
quota = subprocess.run(["bash", COST_SPIKE], cwd=ROOT, capture_output=True, text=True)
if quota.returncode == 2:
print(json.dumps({
"success": False,
"mode": "BLOCK",
"risk": "high",
"decision": "HALTED",
"answer": f"Tenant quota exceeded for {os.environ['CASAN_TENANT_ID']}.",
"sources": [],
"certified": False,
"audit": {},
"router": {"reason": "tenant_quota_exceeded", "matched_rules": ["tenant_quota"]},
}, ensure_ascii=False))
return 3
return 0
def loop_env_for(args):
env = {
**os.environ,
"CASAN_LOOP_STATE_ROOT": os.environ.get("CASAN_LOOP_STATE_ROOT") or tenant_path("chat/loop-state", "logs/chat/loop-state"),
}
if os.environ.get("CASAN_TENANT_ID"):
env["CASAN_TENANT_ID"] = os.environ["CASAN_TENANT_ID"]
return env
def run_preflight_and_context(run_id: str, draft_path: str):
preflight_out = os.path.join(loop_dir(run_id), "preflight.json")
r = subprocess.run(["bash", PREFLIGHT, draft_path, preflight_out, "--model", "local:chat-turn"], cwd=ROOT, capture_output=True, text=True)
@@ -235,11 +312,7 @@ def certify_operator_draft(args, router, binding):
"success_criteria": criteria_path,
}, 2
loop_env = {
**os.environ,
"CASAN_LOOP_STATE_ROOT": os.environ.get("CASAN_LOOP_STATE_ROOT") or os.path.join(state_root(), "logs", "chat", "loop-state"),
"CASAN_TENANT_ID": args.tenant,
}
loop_env = loop_env_for(args)
dlevel = f"L{binding.get('delegation_level', 0)}"
r = subprocess.run([
"bash", LOOP_RUN,
@@ -318,11 +391,7 @@ def certify_codegen_draft(args, router, binding):
"artifact_scan": {"ok": False, "output": scan_out},
}, 2
loop_env = {
**os.environ,
"CASAN_LOOP_STATE_ROOT": os.environ.get("CASAN_LOOP_STATE_ROOT") or os.path.join(state_root(), "logs", "chat", "loop-state"),
"CASAN_TENANT_ID": args.tenant,
}
loop_env = loop_env_for(args)
dlevel = f"L{binding.get('delegation_level', 0)}"
r = subprocess.run([
"bash", LOOP_RUN,
@@ -530,6 +599,9 @@ def submit_escalation(args, router, binding):
def ask(args) -> int:
guard_rc = tenant_runtime_guard(args)
if guard_rc != 0:
return guard_rc
router = classify(args.message)
bind_rc, binding = bind_agent(args, router)
if bind_rc != 0:
@@ -155,6 +155,7 @@ run "phase-chat-stream-hold" bash "$TESTS/phase-chat-stream-hold-tests.sh"
run "phase-chat-replay" bash "$TESTS/phase-chat-replay-tests.sh"
run "phase-chat-approval" bash "$TESTS/phase-chat-approval-tests.sh"
run "phase-chat-codegen" bash "$TESTS/phase-chat-codegen-tests.sh"
run "phase-chat-tenant" bash "$TESTS/phase-chat-tenant-tests.sh"
# ARCH-02: coverage cannot silently drop; ARCH-01: harness/policy cannot silently
# drift. Both SKIP cleanly when no manifest is provisioned (non-strict dev/CI).