Files
fgAI/src/fgai/monitor.py
2026-07-06 11:27:56 +02:00

740 lines
41 KiB
Python

from __future__ import annotations
import contextlib
import json
import signal
import threading
import time
from pathlib import Path
from .anomaly import anomaly_summary, detect_source_anomalies
from .baseline import BaselineStore
from .config import ConfigStore
from .correlation import correlate_source_ips
from .event_context import build_event_context
from .feedback import FeedbackStore
from .graylog_aggregate import GraylogAggregateSource
from .graylog_mcp import GraylogMcpClient
from .graylog_source import GraylogStreamSource
from .history import FieldDiscoveryStore, HistoryStore, StatusSnapshotStore
from .incidents import IncidentStore, build_incidents
from .data_quality import assess_data_quality
from .llm import ollama_dashboard_assessment, ollama_profile_advice
from .logs import local_in_failures, read_events, summarize_events, top_field_values
from .mitigation import parse_allowlist, suggest_block_candidates
from .policies import audit_policies, read_policies
from .profile_suggestions import apply_profile_advice, suggest_stream_profiles
from .recommendations import build_recommendations
from .sequences import detect_sequences
from .threat_intel import ThreatIntelClient, enrich_ips, is_public_ip
from .triage import build_triage_queue
from .stream_profiles import parse_profiles
@contextlib.contextmanager
def _cycle_timeout(seconds: int):
if seconds <= 0 or threading.current_thread() is not threading.main_thread() or not hasattr(signal, "SIGALRM"):
yield
return
previous_handler = signal.getsignal(signal.SIGALRM)
previous_timer = signal.getitimer(signal.ITIMER_REAL)
started = time.monotonic()
def _raise_timeout(_signum, _frame):
raise TimeoutError(f"status_cycle_timeout_{seconds}s")
signal.signal(signal.SIGALRM, _raise_timeout)
signal.setitimer(signal.ITIMER_REAL, seconds)
try:
yield
finally:
elapsed = time.monotonic() - started
remaining = max(0.0, previous_timer[0] - elapsed) if previous_timer[0] else 0.0
signal.setitimer(signal.ITIMER_REAL, remaining, previous_timer[1])
signal.signal(signal.SIGALRM, previous_handler)
def _stream_titles(config: dict[str, object]) -> dict[str, str]:
return {
str(item.get("id", "")): str(item.get("title", "") or item.get("id", ""))
for item in config.get("graylog_streams", [])
if isinstance(item, dict) and item.get("id")
}
def _public_runtime_config(config: dict[str, object]) -> dict[str, object]:
public = {
key: value
for key, value in config.items()
if key not in {"graylog_mcp_token", "abuseipdb_api_key", "virustotal_api_key"}
}
public["graylog_mcp_token_configured"] = bool(config.get("graylog_mcp_token"))
public["abuseipdb_api_key_configured"] = bool(config.get("abuseipdb_api_key"))
public["virustotal_api_key_configured"] = bool(config.get("virustotal_api_key"))
public["enabled_streams"] = sum(
1
for item in config.get("graylog_streams", [])
if isinstance(item, dict) and item.get("enabled")
)
return public
def _profile_field_set(profile: object | None) -> set[str]:
if profile is None:
return set()
fields = {
str(getattr(profile, "entity_field", "") or ""),
str(getattr(profile, "timestamp_field", "") or ""),
}
fields.update(str(field) for field in getattr(profile, "entity_fields", ()) if field)
fields.update(str(field) for field in getattr(profile, "categorical_fields", ()) if field)
fields.update(str(field) for field in getattr(profile, "numeric_fields", ()) if field)
for relation in getattr(profile, "relationship_fields", ()):
fields.add(str(getattr(relation, "left", "") or ""))
fields.add(str(getattr(relation, "right", "") or ""))
return {field for field in fields if field}
def _suggestion_field_set(suggestion: dict[str, object]) -> set[str]:
profile = suggestion.get("profile", {}) if isinstance(suggestion.get("profile"), dict) else {}
fields = {
str(profile.get("entity_field", "") or ""),
str(profile.get("timestamp_field", "") or ""),
}
for key in ("entity_fields", "categorical_fields", "numeric_fields"):
value = profile.get(key, [])
if isinstance(value, list):
fields.update(str(field) for field in value if field)
relationships = profile.get("relationship_fields", [])
if isinstance(relationships, list):
for relation in relationships:
if isinstance(relation, dict):
fields.add(str(relation.get("left", "") or ""))
fields.add(str(relation.get("right", "") or ""))
return {field for field in fields if field}
def _needs_profile_advisor(suggestion: dict[str, object], stream_profiles: dict[str, object]) -> bool:
if not suggestion.get("profile_exists"):
return True
stream_id = str(suggestion.get("stream_id", ""))
suggested_fields = _suggestion_field_set(suggestion)
existing_fields = _profile_field_set(stream_profiles.get(stream_id))
return bool(suggested_fields - existing_fields)
def _stream_name(stream_id: str, stream_titles: dict[str, str], profile: object | None = None) -> str:
return stream_titles.get(stream_id) or getattr(profile, "name", "") or stream_id
def _profile_name(stream_id: str, stream_titles: dict[str, str], profile: object | None = None) -> str:
name = str(getattr(profile, "name", "") or "").strip()
if not name or name == stream_id:
return f"{_stream_name(stream_id, stream_titles, profile)} profile"
return name
def _range_seconds(value: object) -> int:
try:
return max(60, int(value))
except (TypeError, ValueError):
return 300
def _graylog_fields_from_result(result: dict[str, object]) -> list[dict[str, object]]:
content = result.get("result", {}).get("content", []) if isinstance(result.get("result"), dict) else []
text = next((item.get("text", "") for item in content if isinstance(item, dict)), "")
if not text:
return []
try:
payload = json.loads(text)
except json.JSONDecodeError:
return []
fields = payload.get("fields", payload) if isinstance(payload, dict) else payload
if isinstance(fields, dict) and isinstance(fields.get("fields"), list):
fields = fields["fields"]
return [item for item in fields if isinstance(item, dict)] if isinstance(fields, list) else []
def _stream_coverage(runtime_values: dict[str, object], stream_profiles: dict[str, object], stream_status: dict[str, object], profile_readiness: list[dict[str, object]], stream_titles: dict[str, str]) -> list[dict[str, object]]:
configured = [
item for item in runtime_values.get("graylog_streams", [])
if isinstance(item, dict) and item.get("id")
]
status_by_id = {
str(item.get("stream_id", "")): item
for item in stream_status.get("streams", [])
if isinstance(item, dict) and item.get("stream_id")
}
readiness_by_stream: dict[str, list[dict[str, object]]] = {}
for item in profile_readiness:
readiness_by_stream.setdefault(str(item.get("stream_id", "")), []).append(item)
ids = list(dict.fromkeys([str(item.get("id", "")) for item in configured] + list(stream_profiles) + list(status_by_id)))
rows = []
for stream_id in ids:
profile = stream_profiles.get(stream_id)
readiness = readiness_by_stream.get(stream_id, [])
ready_fields = sum(1 for item in readiness if item.get("ready"))
total_fields = len(readiness)
status = status_by_id.get(stream_id, {})
enabled = next((bool(item.get("enabled")) for item in configured if str(item.get("id", "")) == stream_id), False)
events_fetched = int(status.get("events_fetched", 0) or 0)
health = "not_enabled" if not enabled else "poll_budget_skipped" if status.get("error") == "skipped_poll_budget" else "partial_fetch" if status.get("partial") else "missing_profile" if not profile else "no_events" if events_fetched == 0 else "learning" if total_fields and ready_fields < total_fields else "ready" if total_fields else "profile_needs_fields"
rows.append({
"stream_id": stream_id,
"stream_name": _stream_name(stream_id, stream_titles, profile),
"enabled": enabled,
"profile": _profile_name(stream_id, stream_titles, profile) if profile else "",
"profile_ready": bool(profile),
"entity_field": ", ".join(getattr(profile, "entity_fields", ()) or (str(getattr(profile, "entity_field", "")),)) if profile else "",
"tracked_fields": len(getattr(profile, "categorical_fields", ())) + len(getattr(profile, "numeric_fields", ())) + len(getattr(profile, "relationship_fields", ())) if profile else 0,
"ready_fields": ready_fields,
"total_fields": total_fields,
"readiness": f"{ready_fields}/{total_fields}" if total_fields else "0/0",
"events_fetched": events_fetched,
"aggregate_events": int(status.get("aggregate_events", 0) or 0),
"aggregate_status": str(status.get("aggregate_status", "")),
"aggregate_schema_properties": ", ".join(str(item) for item in status.get("aggregate_schema_properties", []) if item),
"latest_event_time": str(status.get("latest_event_time", "")),
"truncated": bool(status.get("truncated")),
"partial": bool(status.get("partial")),
"raw_error": str(status.get("error", "")),
"aggregate_error": str(status.get("aggregate_error", "")),
"error": str(status.get("aggregate_error", "") or status.get("error", "")),
"health": health,
"health_detail": str(status.get("health_detail", "")) or ("No raw events returned for this stream in the current MCP poll window." if enabled and events_fetched == 0 else ""),
})
return rows
def build_status(
log_path: str,
*,
policy_path: str | None = None,
min_block_events: int = 3,
min_block_score: int = 7,
anomaly_limit: int = 20,
baseline_path: str | None = None,
config_path: str | None = None,
history_path: str | None = None,
incident_path: str | None = None,
status_cache_path: str | None = None,
) -> dict[str, object]:
config_store = ConfigStore(config_path) if config_path else None
config_exists = bool(config_store and config_store.path.exists())
runtime_values = config_store.read() if config_exists and config_store else {}
runtime_config = _public_runtime_config(runtime_values) if config_store else {}
stream_profiles = parse_profiles(runtime_values.get("graylog_stream_profiles", []))
stream_titles = _stream_titles(runtime_values)
events = read_events(log_path) if Path(log_path).exists() else []
mcp_status: dict[str, object] = {"status": "not_configured"}
if runtime_values.get("log_source") == "graylog_mcp":
url, token = str(runtime_values.get("graylog_mcp_url", "")), str(runtime_values.get("graylog_mcp_token", ""))
verify_tls = bool(runtime_values.get("graylog_tls_verify", True))
if not url or not token:
mcp_status = {"status": "missing_configuration"}
events = []
else:
try:
configured_streams = runtime_values.get("graylog_streams", [])
stream_configs = [item for item in configured_streams if isinstance(item, dict) and item.get("enabled") and item.get("id")]
stream_ids = [str(item.get("id")) for item in stream_configs]
if not stream_ids:
legacy_stream = str(runtime_values.get("graylog_stream", "") or "")
if legacy_stream:
stream_configs = [{"id": legacy_stream, "title": "Graylog"}]
else:
mcp_status = {"status": "no_streams_enabled", "streams": [], "events_fetched": 0, "coverage_status": "no_streams_enabled"}
events = []
raise StopIteration
stream_statuses = []
events = []
range_seconds = _range_seconds(runtime_values.get("graylog_range_seconds", 300))
max_events_per_stream = max(1, int(runtime_values.get("graylog_max_events_per_stream", 5000) or 5000))
raw_sample_events = max(1, int(runtime_values.get("graylog_raw_sample_events", 5000) or 5000))
mcp_call_timeout = max(1, int(runtime_values.get("graylog_mcp_call_timeout_seconds", 8) or 8))
mcp_poll_timeout = max(60, int(runtime_values.get("graylog_mcp_poll_timeout_seconds", 120) or 120))
fetch_mode = str(runtime_values.get("graylog_fetch_mode", "auto") or "auto")
use_aggregate = fetch_mode == "aggregate" or (fetch_mode == "auto" and max_events_per_stream > raw_sample_events)
aggregate_events_total = 0
poll_started_monotonic = time.monotonic()
poll_deadline = poll_started_monotonic + mcp_poll_timeout
client = GraylogMcpClient(url, token, timeout=mcp_call_timeout, verify_tls=verify_tls)
probe_status = client.probe()
discovery_store = FieldDiscoveryStore(history_path) if history_path else None
catalog_fields_total = 0
def budget_exceeded() -> bool:
return time.monotonic() >= poll_deadline
def skipped_status(stream_id: str, stream_name: str) -> dict[str, object]:
return {
"stream_id": stream_id,
"stream_name": stream_name,
"source": "graylog_mcp",
"events_fetched": 0,
"aggregate_events": 0,
"pages": 0,
"partial": True,
"error": "skipped_poll_budget",
"truncated": False,
"latest_event_time": "",
"raw_sample_limit": min(raw_sample_events, 10_000) if use_aggregate else max_events_per_stream,
"health_detail": "Skipped because the MCP poll time budget was reached before this stream.",
}
for stream_config in stream_configs:
stream_id = str(stream_config["id"])
stream_name = str(stream_config.get("title", "") or stream_titles.get(stream_id) or stream_id)
if budget_exceeded():
stream_statuses.append(skipped_status(stream_id, stream_name))
continue
profile = stream_profiles.get(stream_id)
profile_fields = (
str(getattr(profile, "entity_field", "")),
*tuple(str(field) for field in getattr(profile, "entity_fields", ())),
str(getattr(profile, "timestamp_field", "")),
*tuple(str(field) for field in getattr(profile, "categorical_fields", ())),
*tuple(str(field) for field in getattr(profile, "numeric_fields", ())),
*tuple(str(getattr(relation, "left", "")) for relation in getattr(profile, "relationship_fields", ())),
*tuple(str(getattr(relation, "right", "")) for relation in getattr(profile, "relationship_fields", ())),
) if profile else ()
if discovery_store and not budget_exceeded():
try:
catalog_fields_total += discovery_store.ingest_catalog(stream_id, stream_name, _graylog_fields_from_result(client.call_tool("list_fields", {"streams": [stream_id]})))
except RuntimeError:
pass
aggregate_status: dict[str, object] = {}
if use_aggregate and not budget_exceeded():
aggregate_status = GraylogAggregateSource(client, stream_id, str(runtime_values.get("graylog_query", "*"))).fetch_count(range_seconds=range_seconds, probe_status=probe_status, deadline_monotonic=poll_deadline)
aggregate_events_total += int(aggregate_status.get("aggregate_events", 0) or 0)
raw_limit = min(raw_sample_events, 10_000) if use_aggregate else max_events_per_stream
if budget_exceeded():
stream_statuses.append({**skipped_status(stream_id, stream_name), **aggregate_status})
continue
stream_events, stream_status = GraylogStreamSource(client, stream_id, str(runtime_values.get("graylog_query", "*")), str(runtime_values.get("graylog_field_mapping", "")), stream_name, profile_fields).fetch(max_events=raw_limit, range_seconds=range_seconds, probe_status=probe_status, deadline_monotonic=poll_deadline)
events.extend(stream_events)
stream_statuses.append({"stream_id": stream_id, "stream_name": stream_name, **aggregate_status, **stream_status, "raw_sample_limit": raw_limit})
sample_limited_streams = [item for item in stream_statuses if item.get("truncated") and use_aggregate]
truncated_streams = [item for item in stream_statuses if item.get("truncated") and not use_aggregate]
partial_streams = [item for item in stream_statuses if item.get("partial")]
skipped_streams = [item for item in stream_statuses if item.get("error") == "skipped_poll_budget"]
aggregate_errors = [item for item in stream_statuses if item.get("aggregate_status") == "error"]
warnings = []
if skipped_streams:
warnings.append(f"{len(skipped_streams)} stream(s) skipped because the MCP poll time budget was reached.")
if aggregate_errors:
warnings.append(f"{len(aggregate_errors)} stream(s) returned aggregate MCP errors.")
if partial_streams:
warnings.append(f"{len(partial_streams)} stream(s) returned a partial MCP fetch; Graylog likely timed out or rejected a large paged query.")
if truncated_streams:
warnings.append(f"{len(truncated_streams)} stream(s) hit max_events_per_stream; high EPS means the analysis window is only partially sampled.")
mcp_status = {
"status": "partial" if partial_streams or aggregate_errors else "connected",
"streams": stream_statuses,
"events_fetched": len(events),
"raw_events_fetched": len(events),
"aggregate_events": aggregate_events_total,
"poll_completed_at": int(time.time()),
"poll_duration_seconds": round(time.monotonic() - poll_started_monotonic, 2),
"fetch_mode": "aggregate" if use_aggregate else "raw",
"range_seconds": range_seconds,
"max_events_per_stream": max_events_per_stream,
"raw_sample_events": raw_sample_events,
"call_timeout_seconds": mcp_call_timeout,
"poll_timeout_seconds": mcp_poll_timeout,
"partial_streams": len(partial_streams),
"skipped_streams": len(skipped_streams),
"sample_limited_streams": len(sample_limited_streams),
"truncated_streams": len(truncated_streams),
"aggregate_error_streams": len(aggregate_errors),
"catalog_fields": catalog_fields_total,
"coverage_status": "partial" if partial_streams else "truncated" if truncated_streams else "complete_window",
"coverage_warning": " ".join(warnings),
}
except StopIteration:
pass
except RuntimeError as exc:
mcp_status = {"status": "error", "error": str(exc)}
events = []
baseline = BaselineStore(baseline_path) if baseline_path else None
profiles = baseline.profiles({event.src_ip for event in events if event.src_ip}) if baseline else {}
baseline_training_days = int(runtime_values.get("baseline_training_days", 7) or 7)
field_deviations = baseline.profile_deviations(events, stream_profiles, min_training_days=baseline_training_days) if baseline else {}
sequence_findings = detect_sequences(events)
for entity, findings in sequence_findings.items():
field_deviations.setdefault(entity, []).extend(findings)
for deviations in field_deviations.values():
for deviation in deviations:
stream_id = str(deviation.get("stream_id", ""))
profile = stream_profiles.get(stream_id)
name = _stream_name(stream_id, stream_titles, profile)
deviation["stream_name"] = name
deviation["stream_title"] = name
deviation["profile_name"] = _profile_name(stream_id, stream_titles, profile)
deviation["sample_events"] = [
{"stream": name, **sample} if isinstance(sample, dict) and not sample.get("stream") else sample
for sample in deviation.get("sample_events", [])
]
feedback = FeedbackStore().entries()
for entity, deviations in field_deviations.items():
for deviation in deviations:
match = next((
item for item in feedback
if item.get("entity") == entity
and item.get("stream_id") == deviation.get("stream_id")
and item.get("field") == deviation.get("field")
and (not item.get("value") or item.get("value") == deviation.get("value", ""))
), None)
if match:
deviation["feedback"] = match["status"]
if match["status"] in {"false_positive", "expected"}:
deviation["score"] = 0
anomalies = detect_source_anomalies(events, limit=anomaly_limit, baselines=profiles, field_deviations=field_deviations)
baseline_events = baseline.ingest(events) if baseline else 0
profile_baseline_fields = baseline.ingest_profile_fields(events, stream_profiles) if baseline else 0
baseline_maintenance = (
baseline.maintenance(
retention_days=int(runtime_values.get("baseline_retention_days", 14) or 14),
value_retention_days=int(runtime_values.get("baseline_value_retention_days", 7) or 7),
max_values_per_field=int(runtime_values.get("baseline_max_values_per_field", 2000) or 2000),
)
if baseline
else {}
)
profile_readiness = baseline.profile_readiness(stream_profiles, min_training_days=baseline_training_days) if baseline else []
profile_readiness = [
{
**item,
"profile_name": _profile_name(str(item.get("stream_id", "")), stream_titles, stream_profiles.get(str(item.get("stream_id", "")))),
"stream_name": _stream_name(str(item.get("stream_id", "")), stream_titles, stream_profiles.get(str(item.get("stream_id", "")))),
}
for item in profile_readiness
]
stream_coverage = _stream_coverage(runtime_values, stream_profiles, mcp_status, profile_readiness, stream_titles)
discovery_cache_events = []
discovered_profile_fields = 0
if history_path:
discovery_store = FieldDiscoveryStore(history_path)
discovered_profile_fields = discovery_store.ingest(events)
discovery_cache_events = discovery_store.synthetic_events()
profile_suggestions = suggest_stream_profiles([*discovery_cache_events, *events], existing_profiles=stream_profiles)
profile_advisor_status = {"enabled": bool(runtime_values.get("profile_advisor_enabled")), "status": "disabled"}
advisor_candidates = [item for item in profile_suggestions if _needs_profile_advisor(item, stream_profiles)]
if runtime_values.get("profile_advisor_enabled") and profile_suggestions and not advisor_candidates:
profile_advisor_status = {"enabled": True, "status": "skipped_no_profile_changes", "reason": "No missing profiles or newly discovered profile fields need advisor review."}
elif runtime_values.get("profile_advisor_enabled") and profile_suggestions:
try:
advice = ollama_profile_advice(
advisor_candidates[:10],
model=str(runtime_values.get("profile_advisor_model", "") or "qwen3:8b"),
timeout=int(runtime_values.get("profile_advisor_timeout", 120) or 120),
)
profile_suggestions = apply_profile_advice(profile_suggestions, advice)
profile_advisor_status = {"enabled": True, "status": "ok" if advice else "empty", "profiles_returned": len(advice), "model": str(runtime_values.get("profile_advisor_model", "") or "qwen3:8b")}
except Exception as exc:
profile_advisor_status = {"enabled": True, "status": "error", "error": str(exc), "model": str(runtime_values.get("profile_advisor_model", "") or "qwen3:8b")}
for suggestion in profile_suggestions:
suggestion.setdefault("profile_advisor", {"status": "heuristic", "error": str(exc)})
intel_ips = sorted(
{
ip
for event in events
for ip in (event.src_ip, event.dst_ip)
if is_public_ip(ip)
}
)
threat_enabled = bool(runtime_values.get("threat_intel_enabled")) if runtime_values else None
reputation = enrich_ips(intel_ips, limit=25, enabled=threat_enabled, config=runtime_values)
threat_intel_status = ThreatIntelClient(
enabled=threat_enabled,
provider=str(runtime_values.get("threat_intel_provider", "auto")),
abuseipdb_key=str(runtime_values.get("abuseipdb_api_key", "") or "") or None,
virustotal_key=str(runtime_values.get("virustotal_api_key", "") or "") or None,
daily_limit=int(runtime_values.get("threat_intel_daily_limit", 100) or 100),
ttl_seconds=int(runtime_values.get("threat_intel_ttl_seconds", 604800) or 604800),
error_ttl_seconds=int(runtime_values.get("threat_intel_error_ttl_seconds", 3600) or 3600),
abuseipdb_max_age_days=int(runtime_values.get("abuseipdb_max_age_days", 90) or 90),
).status()
recommendations = build_recommendations(events, anomalies, reputation)
correlations = correlate_source_ips(events)
incidents = IncidentStore(incident_path or "state/signalscope-incidents.json").apply(build_incidents(anomalies, field_deviations, correlations))
triage_queue = build_triage_queue(incidents, field_deviations, correlations, recommendations)
block_candidates = suggest_block_candidates(
events,
min_events=min_block_events,
min_score=min_block_score,
allowlist=parse_allowlist(),
)
policy_findings: list[dict[str, str | None]] = []
policy_error: str | None = None
if policy_path and Path(policy_path).exists():
try:
policy_findings = [finding.__dict__ for finding in audit_policies(read_policies(policy_path))]
except Exception as exc:
policy_error = str(exc)
summary = summarize_events(events)
if isinstance(mcp_status, dict) and int(mcp_status.get("aggregate_events", 0) or 0) > summary.get("total", 0):
summary["total"] = int(mcp_status.get("aggregate_events", 0) or 0)
summary["raw_sample_total"] = len(events)
summary["aggregate_backed"] = True
status = {
"status_schema": 2,
"stale": False,
"generated_at": int(time.time()),
"log_path": log_path,
"policy_path": policy_path,
"summary": summary,
"anomaly_summary": anomaly_summary(anomalies),
"baseline": {"enabled": bool(baseline), "sources_ready": len(profiles), "training_days": baseline_training_days, "new_events_recorded": baseline_events, "profile_fields_recorded": profile_baseline_fields, "discovery_fields_recorded": discovered_profile_fields, "discovery_cache_events": len(discovery_cache_events), "maintenance": baseline_maintenance, "size_bytes": baseline_maintenance.get("size_bytes", 0) if isinstance(baseline_maintenance, dict) else 0},
"capabilities": {"threat_intel": threat_intel_status, "graylog_mcp": mcp_status, "profile_advisor": profile_advisor_status},
"configuration": runtime_config,
"stream_profiles": [{"stream_id": item.stream_id, "name": _profile_name(item.stream_id, stream_titles, item), "stream_name": _stream_name(item.stream_id, stream_titles, item), "entity_field": item.entity_field, "entity_fields": list(item.entity_fields), "timestamp_field": item.timestamp_field, "categorical_fields": list(item.categorical_fields), "numeric_fields": list(item.numeric_fields), "detectors": item.detectors, "field_weights": item.field_weights, "relationship_fields": [{"left": relation.left, "right": relation.right, "name": relation.name} for relation in item.relationship_fields]} for item in stream_profiles.values()],
"stream_coverage": stream_coverage,
"profile_suggestions": profile_suggestions,
"profile_readiness": profile_readiness,
"diagnostics": {
"top_source_ips": top_field_values(events, "srcip", limit=10),
"top_destination_ips": top_field_values(events, "dstip", limit=10),
"top_policy_ids": top_field_values(events, "policyid", limit=10),
"top_destination_ports": top_field_values(events, "dstport", limit=10),
"top_source_ports": top_field_values(events, "srcport", limit=10),
"top_services": top_field_values(events, "service", limit=10),
"top_actions": top_field_values(events, "action", limit=10),
"top_subtypes": top_field_values(events, "subtype", limit=10),
"local_in_failures": local_in_failures(events, limit=10),
},
"event_context": build_event_context(events),
"field_deviations": field_deviations,
"triage_queue": triage_queue,
"sequence_findings": sequence_findings,
"feedback": feedback,
"cross_source_correlations": correlations,
"incidents": incidents,
"data_quality": assess_data_quality(events, mcp_status),
"status_cache": {"served_from_cache": False, "reason": ""},
"anomalies": [
{
"subject": finding.subject,
"score": finding.score,
"severity": finding.severity,
"confidence": finding.confidence,
"reasons": finding.reasons,
"evidence": finding.evidence,
}
for finding in anomalies
],
"recommendations": [
{
"subject": item.subject,
"score": item.score,
"severity": item.severity,
"title": item.title,
"recommendation": item.recommendation,
"reasons": item.reasons,
"related_policy_ids": item.related_policy_ids,
"related_services": item.related_services,
}
for item in recommendations
],
"reputation": reputation,
"block_candidates": [
{"src_ip": candidate.src_ip, "score": candidate.score, "reasons": candidate.reasons}
for candidate in block_candidates
],
"policy_findings": policy_findings,
"policy_error": policy_error,
}
if history_path:
history = HistoryStore(history_path)
history.record(status)
status["history"] = history.recent()
if status_cache_path and not (isinstance(mcp_status, dict) and mcp_status.get("status") == "error"):
StatusSnapshotStore(status_cache_path).save("last_good", status)
return status
def cached_status_with_error(cache_path: str, error_status: dict[str, object]) -> dict[str, object] | None:
cached = StatusSnapshotStore(cache_path).load("last_good")
if not cached:
return None
for row in cached.get("stream_coverage", []) if isinstance(cached.get("stream_coverage"), list) else []:
if isinstance(row, dict):
row.setdefault("raw_error", "")
row.setdefault("aggregate_error", "")
row.setdefault("error", row.get("aggregate_error") or row.get("raw_error") or "")
if row.get("aggregate_status") == "error" and not row.get("aggregate_error") and row.get("error"):
row["aggregate_error"] = str(row.get("error", ""))
if row.get("partial") and not row.get("raw_error") and row.get("error"):
row["raw_error"] = str(row.get("error", ""))
cached["status_schema"] = 2
cached["generated_at"] = int(time.time())
cached["stale"] = True
cached["stale_reason"] = "live_mcp_error"
capabilities = cached.setdefault("capabilities", {})
if isinstance(capabilities, dict):
capabilities["graylog_mcp"] = error_status
cached.setdefault("status_cache", {})
if isinstance(cached["status_cache"], dict):
cached["status_cache"].update({"served_from_cache": True, "reason": "live_mcp_error"})
return cached
def add_llm_assessment(status: dict[str, object], *, previous: str | None = None, model: str | None = None, timeout: int | None = None) -> None:
try:
status["llm_assessment"] = {
"enabled": True,
"status": "ok",
"generated_at": int(time.time()),
"text": ollama_dashboard_assessment(status, model=model, timeout=timeout),
}
except Exception as exc:
status["llm_assessment"] = {
"enabled": True,
"status": "error",
"generated_at": int(time.time()),
"error": str(exc),
"text": previous or "",
}
def write_status(status: dict[str, object], output: str) -> None:
output_path = Path(output)
output_path.parent.mkdir(parents=True, exist_ok=True)
tmp_path = output_path.with_suffix(f"{output_path.suffix}.tmp")
tmp_path.write_text(json.dumps(status, indent=2, sort_keys=True), encoding="utf-8")
tmp_path.replace(output_path)
def write_refreshing_status(output: str, *, cache_path: str | None = None, call_timeout_seconds: int = 0, poll_timeout_seconds: int = 0, runtime_values: dict[str, object] | None = None) -> None:
output_path = Path(output)
now = int(time.time())
try:
current = json.loads(output_path.read_text(encoding="utf-8")) if output_path.exists() else {}
except (json.JSONDecodeError, OSError):
current = {}
if not isinstance(current, dict):
current = {}
used_cache = False
if cache_path:
cached = StatusSnapshotStore(cache_path).load("last_good")
if cached:
current = cached
used_cache = True
previous_mcp = current.get("capabilities", {}).get("graylog_mcp", {}) if isinstance(current.get("capabilities"), dict) else {}
previous_completed_at = previous_mcp.get("poll_completed_at", current.get("generated_at", 0)) if isinstance(previous_mcp, dict) else current.get("generated_at", 0)
current.setdefault("status_schema", 2)
current["generated_at"] = now
current["stale"] = False
current["stale_reason"] = ""
if runtime_values is not None:
stream_profiles = parse_profiles(runtime_values.get("graylog_stream_profiles", []))
current["configuration"] = _public_runtime_config(runtime_values)
current["stream_coverage"] = _stream_coverage(runtime_values, stream_profiles, {"streams": []}, [], _stream_titles(runtime_values))
capabilities = current.setdefault("capabilities", {})
if isinstance(capabilities, dict):
previous_events = previous_mcp.get("events_fetched", 0) if isinstance(previous_mcp, dict) else 0
previous_raw_events = previous_mcp.get("raw_events_fetched", 0) if isinstance(previous_mcp, dict) else 0
previous_aggregate_events = previous_mcp.get("aggregate_events", 0) if isinstance(previous_mcp, dict) else 0
previous_fetch_mode = previous_mcp.get("fetch_mode", "") if isinstance(previous_mcp, dict) else ""
previous_coverage_status = previous_mcp.get("coverage_status", "") if isinstance(previous_mcp, dict) else ""
capabilities["graylog_mcp"] = {
"status": "refreshing",
"previous_status": previous_mcp.get("status", "") if isinstance(previous_mcp, dict) else "",
"previous_error": previous_mcp.get("error", "") if isinstance(previous_mcp, dict) else "",
"previous_events_fetched": previous_events,
"previous_raw_events_fetched": previous_raw_events,
"previous_aggregate_events": previous_aggregate_events,
"previous_fetch_mode": previous_fetch_mode,
"previous_coverage_status": previous_coverage_status,
"events_fetched": previous_events,
"raw_events_fetched": previous_raw_events,
"aggregate_events": previous_aggregate_events,
"poll_started_at": now,
"previous_poll_completed_at": previous_completed_at,
"fetch_mode": previous_fetch_mode,
"coverage_status": previous_coverage_status or "refreshing",
"call_timeout_seconds": call_timeout_seconds or previous_mcp.get("call_timeout_seconds", 0),
"poll_timeout_seconds": poll_timeout_seconds or previous_mcp.get("poll_timeout_seconds", 0),
}
current["status_cache"] = {"served_from_cache": used_cache, "reason": "refreshing"}
write_status(current, output)
def monitor_loop(
log_path: str,
output: str,
*,
policy_path: str | None = None,
interval: int = 10,
anomaly_limit: int = 20,
llm: bool = False,
llm_interval: int = 300,
llm_model: str | None = None,
llm_timeout: int | None = None,
baseline_path: str | None = None,
config_path: str | None = None,
history_path: str | None = None,
status_cache_path: str | None = None,
) -> None:
print(f"Monitoring {log_path}", flush=True)
print(f"Writing status to {output}", flush=True)
last_llm_at = 0
last_llm_text: str | None = None
while True:
runtime = ConfigStore(config_path).read() if config_path and Path(config_path).exists() else {}
effective_llm = bool(runtime.get("llm_enabled")) if runtime else llm
effective_model = str(runtime.get("llm_model") or llm_model or "")
mcp_call_timeout = max(1, int(runtime.get("graylog_mcp_call_timeout_seconds", 8) or 8))
mcp_poll_timeout = max(60, int(runtime.get("graylog_mcp_poll_timeout_seconds", 120) or 120))
if runtime.get("log_source") == "graylog_mcp":
write_refreshing_status(output, cache_path=status_cache_path, call_timeout_seconds=mcp_call_timeout, poll_timeout_seconds=mcp_poll_timeout, runtime_values=runtime)
try:
status_timeout = mcp_poll_timeout + max(30, mcp_call_timeout * 2)
with _cycle_timeout(status_timeout if runtime.get("log_source") == "graylog_mcp" else 0):
status = build_status(
log_path, policy_path=policy_path, anomaly_limit=anomaly_limit,
baseline_path=baseline_path, config_path=config_path, history_path=history_path,
status_cache_path=status_cache_path,
)
except Exception as exc:
error_status = {"status": "error", "error": f"monitor_error: {exc}", "call_timeout_seconds": mcp_call_timeout, "poll_timeout_seconds": mcp_poll_timeout}
status = cached_status_with_error(status_cache_path, error_status) if status_cache_path else None
if not status:
status = {
"status_schema": 2,
"generated_at": int(time.time()),
"stale": True,
"stale_reason": "monitor_error",
"capabilities": {"graylog_mcp": error_status},
"summary": {"total": 0},
"status_cache": {"served_from_cache": False, "reason": "monitor_error"},
}
mcp = status.get("capabilities", {}).get("graylog_mcp", {}) if isinstance(status.get("capabilities"), dict) else {}
if status_cache_path and isinstance(mcp, dict) and mcp.get("status") == "error":
cached = cached_status_with_error(status_cache_path, mcp)
if cached:
status = cached
if effective_llm:
now = int(time.time())
if now - last_llm_at >= llm_interval:
add_llm_assessment(status, previous=last_llm_text, model=effective_model or None, timeout=llm_timeout)
assessment = status.get("llm_assessment", {})
if isinstance(assessment, dict):
last_llm_text = str(assessment.get("text", "") or last_llm_text or "")
last_llm_at = now
else:
status["llm_assessment"] = {
"enabled": True,
"status": "cached",
"generated_at": last_llm_at,
"text": last_llm_text or "",
}
else:
status["llm_assessment"] = {"enabled": False, "status": "disabled", "text": ""}
if status_cache_path and isinstance(mcp, dict) and mcp.get("status") not in {"error", "refreshing"}:
StatusSnapshotStore(status_cache_path).save("last_good", status)
write_status(status, output)
time.sleep(interval)