senteniel-core-fks (Python)

senteniel-core-fks is the library-first Sentinel package:

• no required port
• no required HTTP server
• pluggable MCP/tool backends
• pluggable orchestrators
• pluggable audit sinks

The existing SentinelClient (GraphQL remote client) is still available. Both imports are supported:

• from sentinel_core import ... (preferred)
• from sentinel_sdk import ... (backward compatible)

Install (editable)

pip install -e sdk

Install from PyPI:

pip install senteniel-core-fks

Published package name: senteniel-core-fks.

Local Library Usage (No Server)

from sentinel_core import (
    SentinelEngine,
    StaticToolBackend,
    InMemoryAuditSink,
    make_policy_engine_from_dict,
)

backend = StaticToolBackend(
    {
        "openbnb_airbnb.airbnb_search": lambda args: {"ok": True, "args": args},
    }
)

policy = make_policy_engine_from_dict(
    {
        "openbnb_airbnb.": {
            "decision": "ALLOW",
            "risk": 0.0,
            "reason": "Airbnb read tools allowed",
        }
    }
)

audit = InMemoryAuditSink()
engine = SentinelEngine(tool_backend=backend, policy_engine=policy, audit_sinks=[audit])
engine.sync_tools()

decision = engine.propose_tool_call(
    "openbnb_airbnb.airbnb_search",
    {"location": "Accra"},
)
print(decision)
print(audit.events[-1])

Audit Sinks

• InMemoryAuditSink: captures events in memory (events list).
• JsonlAuditSink: appends one JSON event per line to a file.
• HttpAuditSink: sends events to an external endpoint.

Orchestrator Plugin Example

from sentinel_core import SentinelEngine, ExplicitToolCallOrchestrator, StaticToolBackend

backend = StaticToolBackend({"demo.echo": lambda args: args})
engine = SentinelEngine(tool_backend=backend, allow_unknown_tools=True)
engine.register_orchestrator(ExplicitToolCallOrchestrator())
engine.sync_tools()

result = engine.run("explicit", 'demo.echo {"msg":"hello"}')
print(result)

Remote GraphQL Client (Optional)

from sentinel_core import SentinelClient

client = SentinelClient("http://localhost:8000/graphql")
result = client.propose_tool_call("fs.list_dir", {"path": "/sandbox"})
print(result)