Open-source AI Security Operations Center: alert fusion, LLM-agent triage, MITRE ATT&CK investigation, and a replayable decision ledger for every agent step. Self-hostable, runs with no API keys, MIT licensed. Ships an MCP server for Claude, Cursor and Continue.
Python agentic-aiai-securityalert-triageblue-teamcybersecuritydetection-engineeringincident-responselanggraphllm-agentsmcpmitre-attackmodel-context-protocolpurple-teamsecurity-operationsself-hostedsiemsoarsocthreat-detectionthreat-intelligence
docs(roadmap): date the v8 tracker to v12.2.0 and record why #362 closed (#1015)
Beenu Arora9d77fb94b6cb
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex cfb21c694..e90a516e7 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -120,6 +120,33 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0its synthetic labelling stated. The UI-only preview is still mentioned,labelled as what it is.+- **The v8 tracker is dated to v12.2.0, and the v8.0 wave-2 tracking issue is+ closed rather than refreshed.** Issue #362 existed so the wave-2 backlog+ would be readable "without needing to read `AISOC_V8_PROGRESS.md`", and then+ pointed readers at that file anyway — a link that has returned **404 since+ 2026-06-30**, when #368 deleted the root tracker and created+ `docs/roadmap/v8-progress.md` in the same commit. The issue is the same shape+ as the defect this repository has been burned by before: a tracker whose+ own reference does not resolve.++ Almost nothing needed migrating — the six wave-2 T-IDs are already audited+ against the tree in `docs/roadmap/v8-progress.md`, in a table that cites #362+ as its source, and every wave-3 candidate already carries a disposition in+ `ROADMAP.md`. Re-auditing before closing was still worth doing rather than+ assumed, because it found two things wrong in the tracker itself. **Lacework+ has no `get_resource_config`**, so T1.2 covers three of the four providers+ the ticket named and not four — five connectors implement the method and+ `lacework.py`, which is registered, does not; that gap was recorded nowhere+ and is now in the T1.2 row. And T3.6 still listed a durable approval store as+ outstanding after it had landed: `PostgresTimerStore` is wired in+ `services/slack-bot/app/main.py` behind migration `062_approval_timers.sql`,+ so the genuine remainder there is only the Teams proactive card push.++ The tracker header also read `v10.0.0` while the tree read `12.2.0`, so+ redirecting to it meant making it true first: it is now dated, and carries+ the reason #362 was closed so the next reader does not have to reconstruct+ it.+## [12.2.0] - 2026-09-28### Fixeddiff --git a/docs/roadmap/v8-progress.md b/docs/roadmap/v8-progress.mdindex b119d8da6..99de03295 100644--- a/docs/roadmap/v8-progress.md+++ b/docs/roadmap/v8-progress.md@@ -1,12 +1,13 @@# AiSOC v8 — progress tracker-**Last updated:** 2026-09-25-**Current release:** `v10.0.0` (2026-09-25) · **Next:** unscheduled.+**Last updated:** 2026-09-28+**Current release:** `v12.2.0` (2026-09-28) · **Next:** unscheduled.Packaging is no longer named against a version: it slipped v8.0 to v8.1 tov8.2 for the same reason each time, and the blocker is registry credentials-rather than code. See `[10.0.0]` in `CHANGELOG.md` for what v10.0 closed and-what it knowingly left open — it is a major because upgrading requires an-operator action, not because the v8 line is superseded.+rather than code. Each major since v10.0 has been a major for the same kind of+reason — upgrading requires an operator action — not because the v8 line is+superseded. See the per-version sections of `CHANGELOG.md` for what each one+closed and what it knowingly left open.This tracker is the at-a-glance view of what has landed across the v8 line andwhat is still open. It is deliberately short and dated. When it disagrees with@@ -18,6 +19,23 @@ the tree, the tree wins — and the tracker is the thing to fix.> linked here for "v8.1 packaging work in flight" and a reader found a June> snapshot with no packaging in it. A tracker that is not updated is worse than> no tracker, because it is read as current.+>+> Its root-level predecessor `AISOC_V8_PROGRESS.md` is the sharper version of+> the same lesson, and the reason issue+> [#362](https://github.com/beenuar/AiSOC/issues/362) was closed rather than+> refreshed. That issue existed so the wave-2 backlog would be readable+> "without needing to read `AISOC_V8_PROGRESS.md`", and pointed at the file+> anyway — a link that has returned **404 since 2026-06-30**, when+> [#368](https://github.com/beenuar/AiSOC/pull/368) deleted the root tracker+> and created this one in the same commit. Every item #362 listed is now+> recorded here (wave-2, in the table below) or in+> [`ROADMAP.md`](../../ROADMAP.md) (wave-3). Re-auditing before closing it was+> worth doing rather than assuming: one gap it named — Lacework having no+> config-snapshot writer — was recorded in neither file and is migrated into+> the T1.2 row below, and the T3.6 row was itself wrong, still listing a+> durable approval store that had already landed. A tracking issue for a+> release four majors old is a tracker nobody reads; this file is the one to+> keep current.---@@ -78,12 +96,12 @@ optional one.| T-ID | Item | What the audit found | Done ||------|------|----------------------|------|-| T1.2 | Versioned config-snapshot writers, Neo4j `:CONFIGURED_AS {ts}` | The Neo4j writer is fully built. The Go provider called a route nobody serves, and could not have called the real one (it has no vault), so snapshots silently never ran while reporting themselves enabled. `is_current` / `valid_from` / `valid_to` were declared and never written, so the documented O(1) lookup matched zero edges. | yes |+| T1.2 | Versioned config-snapshot writers, Neo4j `:CONFIGURED_AS {ts}` | The Neo4j writer is fully built. The Go provider called a route nobody serves, and could not have called the real one (it has no vault), so snapshots silently never ran while reporting themselves enabled. `is_current` / `valid_from` / `valid_to` were declared and never written, so the documented O(1) lookup matched zero edges. | partial — **Lacework has no `get_resource_config`**, so three of the four providers the ticket named are covered. Five connectors implement it (AWS Security Hub, Azure Defender, GCP SCC, GitHub, Okta); `lacework.py` is registered and does not. `AISOC_SNAPSHOT_ENABLED` is also off by default and no compose file sets it, so this runs on operator opt-in only. || T2.3 | `LLMInputContract` across every sub-agent | Real and fail-closed in `services/agents`, covering 15 of 16 call sites. `services/api` had **no contract at all** across seven endpoints, and the module described as living there did not exist. The no-bypass gate could not see a raw-HTTP LLM call. | yes || T3.2 | Effective-permissions resolvers (Azure / GCP / Okta / GWS) | **Already shipped** — all five exist, are registered and report `coverage: "full"`. The gap is the snapshot, not the resolver: no connector answers `__posture_snapshot__`, so four of five return 412. Docstrings still called them scaffolds. | audited; gap gated || T3.3 | Attack-chain ranking + timeline UI | Grouping (fusion, Redis) and weighted ranking (API, Postgres) both exist as independent implementations that never exchange data. `AttackChainPanel` and `AttackStory.tsx` are honest about empty states; `InvestigationTimeline.tsx` rendered a fabricated investigation with no demo gate. | partial — fabrication gated, the two implementations remain separate || T3.5 | Business-context rule engine | **Already shipped**, and the v8.0 Postgres fix is real. But the two evaluators disagreed on `not`, so one console-accepted rule silently discarded that tenant's entire rule set at triage, suppressions included. | yes |-| T3.6 | ChatOps coverage | Every piece existed; three had no caller. Slack Block Kit approvals worked only as a reply to the analyst's own slash command, the Teams card factory had zero senders, and the signed email link pointed at a route that did not exist. Worse: approvals **authorized nobody**. | partial — authorization and the email route fixed; proactive card push and a durable approval store remain |+| T3.6 | ChatOps coverage | Every piece existed; three had no caller. Slack Block Kit approvals worked only as a reply to the analyst's own slash command, the Teams card factory had zero senders, and the signed email link pointed at a route that did not exist. Worse: approvals **authorized nobody**. | partial — authorization, the email route and the durable approval store are all done; the Teams proactive card push remains. The durable store was recorded here as outstanding after it had landed: `PostgresTimerStore` is wired at `services/slack-bot/app/main.py` behind migration `062_approval_timers.sql`. The genuine remainder is narrower than "Teams" — `approval_card()` in `services/teams-bot/app/cards.py` has no caller outside its own tests, and `services/teams-bot/app/main.py` exposes only inbound routes, so there is no outbound path for an Adaptive Card. The proactive Teams path that does exist (`services/actions/app/executors/chatops.py`) sends a legacy Connector Card. |### Release integrity