Adds a self-hosted (Ollama, qwen2.5-coder) model provider abstraction
with a pluggable opt-in cloud adapter, schema grounding, and a shared
cost/safety guard every AI-suggested query is assessed against --
compiling to and executing through the same unchanged Phase 2 IR/
compiler and Phase 4 tenant scoping as a hand-written query, no
parallel execution path.
Track A (built into the query bar): inline ghost-text autocomplete,
"Explain this query", "Fix this query" with a diff view, and a
rule-based "Optimize" suggestion. Track B: natural-language-to-query
translation, always a separate review step from execution, with
`sentryctl query --nl` requiring explicit confirmation to run.
Every accepted/dismissed translate-fix-optimize interaction is logged
into the same append-only audit_log table Phase 4 built.
Two real product bugs were found and fixed via live browser
verification (a Svelte effect re-running on every keystroke that
silently cancelled the ghost-text debounce; a ghost-text widget
positioned at document offset 0 instead of the cursor), and a real
costguard logic bug (unbounded-aggregation vs. raw-row) was caught by
its own test suite. New integration tests wire a real Ollama client
through the real HTTP handler against a mock server matching Ollama's
wire contract (hack/mock-ollama), keeping model-quality verification
out of CI as a disclosed, periodic human-run check instead.
See /docs/phase-7-ai-design.md and /docs/phase-7-runbook.md.
PUT/DELETE /dashboards/{id}/permissions/{userId} (per-resource dashboard
grants, built earlier this phase) had no caller but Go tests and curl --
named as a real, disclosed gap in docs/phase-4-runbook.md. Adds a CLI
surface: sentryctl dashboards permissions list/grant/revoke, following
the existing dashboards subcommand pattern.
grant/revoke needed a new httpclient.go helper (httpMutateNoBody) since
both endpoints respond 204 No Content -- the existing helpers all expect
a JSON body to pretty-print. grant validates the role client-side
(viewer/editor only, mirroring api/dashboards.validGrantRole) before
making a request, since Admin/Owner already have tenant-wide dashboard
access and a resource-level grant can never raise someone past Editor.
Verified with real httptest.Server round trips (method, path, request
body, and error-body parsing on a 501 from a deployment with no
enterprise permission service wired in) -- the same pattern every other
sentryctl subcommand's tests already use, no fake/mock client needed
since sentryctl itself is just an HTTP client with no store of its own.
RBAC (api/internal/authz) is live on /query and /dashboards, backed by a
new enterprise/ module (session issuance, audit logging, RBAC storage,
OIDC/SAML protocol wiring) that core never imports -- only calls over
HTTP. Found and fixed a real cross-tenant vulnerability in dashboards
(no tenant_id filtering at all) while writing the threat model doc.
Two things are explicitly NOT done, documented rather than hidden:
tenant isolation for log data itself (/query still shares one ClickHouse
connection and Tantivy index across every tenant -- RBAC controls who
can query, not what a query can see), and human SSO login (protocol
wiring exists, no HTTP handler calls it yet). See
docs/security/threat-model.md and docs/phase-4-runbook.md.
Also adds deploy/ (Go Operator + Helm chart, validated offline only --
no cluster was reachable in this environment).
Saved, shareable multi-panel dashboards (table/line/bar/single-stat
panels via gridstack + uPlot, global + per-panel time range, JSON
export/import) and threshold/absence alert rules with an
ok/pending/firing evaluator and webhook/Slack/PagerDuty delivery.
- New /metadata component: Postgres control-plane store for dashboards,
panels, notification targets, alert rules/state, and delivery log --
see docs/phase-3-dashboard-design.md for why ClickHouse's MergeTree
family isn't a fit for this access pattern (needs real row-level
locking and read-your-writes consistency).
- api/internal/dashboards: dashboard/panel CRUD, pure -- panel query
execution stays client-side, reusing the existing /query endpoint.
- New /alerting service: rule/target CRUD, a ticker-driven evaluator
(claim-then-evaluate concurrency control, transactional-outbox
delivery, query errors and threshold zero-rows never coerced into a
false transition) and webhook/Slack/PagerDuty delivery with
retry/backoff. See docs/phase-3-alerting-design.md for the full
state-machine design and the four correctness properties it
implements.
- web: /dashboards and /alerts UIs; cli: sentryctl dashboards/alerts
list/get/apply, seeding a future Terraform provider's JSON contract.
- hack/alert-load-test: 500 rules against real ClickHouse data, real
measured results in docs/phase-3-runbook.md.
Five real bugs found by actually running this against a live stack
(documented in the runbook, not just fixed silently): a latent Phase 2
bug where ClickHouse rejected the timestamp format used for
earliest=/latest= queries; a "now" literal token injected into query
text; a GridStack/uPlot layout-timing race; JS's Date.parse being too
lenient to use as a timestamp-detection heuristic; a rule's "enabled"
field silently defaulting to false when omitted; and the evaluator's
claim-batch-size and worker-pool-concurrency defaulting to the same
value, causing 500 concurrently-due rules to take 125s to cycle through
instead of the configured 60s.
Replaces the separate SQL-only /query and text-only /search endpoints
with one pipe-syntax query language (plus raw SQL escape hatch) that
compiles to a single IR and execution plan across both backends, so a
query like `message:"connection refused" | stats count by host` runs
as one request instead of two disjoint tools.
- api/internal/querylang: lexer -> ast -> parser -> ir -> planner ->
executor, each layer independently tested.
- Execution generalizes Phase 1's proven Tantivy-prefilter pattern
into a 4-way routing table (pure ClickHouse / text-only / text +
aggregation / raw SQL passthrough).
- Unified web query page and `sentryctl query`, both hitting the same
POST /query endpoint.
- Benchmarked against a real 1,022,000-row dataset
(hack/benchmark-fixture); caught and fixed a real bug where the
Tantivy prefilter cap (10,000) produced an IN-clause exceeding
ClickHouse's default max_query_size -- lowered to 5,000, documented
in docs/query-language-design.md and docs/phase-2-runbook.md.
- docs/query-language-reference.md: customer-facing syntax reference.
End-to-end log pipeline for Linux hosts, per /docs/architecture.md:
- proto: shared gRPC contract (agent <-> ingest), Go bindings checked in
- agent: Rust, musl-targeted, journald/file sourcing, RFC5424 parser,
mTLS gRPC client, no required config for the common case
- ingest: Go, single binary with --mode server|consumer|all; gRPC front
end forwards to Redpanda unchanged, consumer normalizes and
batch-writes to ClickHouse with at-least-once delivery
- storage: ClickHouse schema + a plain SQL-file migration runner
- api: minimal SELECT-only query endpoint, plain REST (not gRPC+gateway
yet -- see api/README.md)
- web: SvelteKit static SPA, one query page
- transport: Redpanda compose + topic provisioning
- cli: sentryctl ping stub
- hack/dev-certs: throwaway CA + cert generation for local mTLS
- root docker-compose.yml + docs/phase-0-runbook.md tie it together
Not yet run end-to-end against real Docker/ClickHouse/Redpanda -- see the
runbook's caveats section before relying on this working as-is.
jcoffey-dev