feat(stuck): stuck-state pipeline overlay (cad-atlas#11)
CD / Build & Import (push) Failing after 3s
CD / Detect unsubstituted template (push) Successful in 1s
CD / Lint / Test / Vet (push) Successful in 5s
CD / var-go/oath (push) Skipped
CD / Deploy via GitOps (push) Skipped

Consumes mathias/cad#14's cad-stuck query directly — cad's app/domain
packages are public Go modules, so this reuses the exact tested
classification logic (import, not re-implementation) rather than
duplicating the taxonomy across two repos.

ApplyStuckOverlay (internal/atlas/stuck.go, fixture-tested) maps each
finding's category onto the existing 9-stage pipeline: armed_never_fired to
the human dispatch gate, claimed_never_ran/ceiling_hit/refire_loop/
near_attempt_cap to the one Execute stage (there is no separate Claim
stage), done_not_closed to the loop-back stage. Adds a StuckCount/
AwaitingCount badge per stage plus one Node per finding, linking out to the
underlying issue/comment — read-only, no controls, matching cad#14's own
constraint.

Two visual classes, kept deliberately distinct per the issue's own framing:
a coral 'N stuck' badge (needs intervention) vs a green 'N awaiting' badge
(a functioning control waiting on a human) — done_not_closed can never
render as a bottleneck.

Verified against the live estate: STAGE 05 shows 2 stuck (cobalt-dingo#30
ceiling_hit, cad-atlas#10 near_attempt_cap), STAGE 08 shows 5 awaiting
(cobalt-dingo #3/#7/#12/#16/#19 done_not_closed) — screenshotted via a
koala k3s Playwright job against a live-token server, matching cad#14's own
proven query output exactly.

go.mod: added git.d-ma.be/mathias/cad v0.10.0 (app/domain/adapters-gitearead
only — no mcp-chassis pulled in, that's only imported by cad's own cmd/cad).
Dockerfile: added the same GOPRIVATE/insteadOf fetch config mathias/cad's
own Dockerfile uses, now needed for this new private-module dependency.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
2026-08-28 13:27:02 +02:00
co-authored by Claude Sonnet 5
parent 4ab65a3b1b
commit 97d0f56383
9 changed files with 327 additions and 8 deletions
+69
View File
@@ -1,9 +1,12 @@
package web
import (
"context"
_ "embed"
"encoding/json"
"net/http"
"os"
"strings"
"sync"
"time"
@@ -12,6 +15,10 @@ import (
"git.d-ma.be/mathias/cad-atlas/internal/cluster"
"git.d-ma.be/mathias/cad-atlas/internal/gitea"
"git.d-ma.be/mathias/cad-atlas/internal/version"
"git.d-ma.be/mathias/cad/adapters/gitearead"
cadapp "git.d-ma.be/mathias/cad/app"
caddomain "git.d-ma.be/mathias/cad/domain"
)
// atlasHTML is the Phase-A/B static shell. It fetches /api/atlas.json at load
@@ -64,6 +71,10 @@ func NewHandler() http.Handler {
}
}
}
// cad-atlas#11: overlay mathias/cad#14's cad-stuck findings as
// per-stage badges. Read-only — ApplyStuckOverlay only adds Nodes
// with a link out; it never writes anything back to Gitea.
a.Stages = atlas.ApplyStuckOverlay(a.Stages, ld.stuck)
if ld.deploy != nil || ld.flux != nil {
var live []atlas.Node
if ld.deploy != nil {
@@ -93,6 +104,30 @@ type liveOverlayData struct {
flux *atlas.Flux
timeline []atlas.RunDot
issues []atlas.Node
stuck []caddomain.StuckFinding
}
// stuckRepos is the estate cad-stuck queries — mirrors mathias/dispatch's
// dispatch-repos.txt scope. Override with CAD_ATLAS_STUCK_REPOS (comma
// separated "owner/name" entries) so this doesn't need a redeploy every time
// the estate's repo list changes; cad-atlas has no filesystem access to the
// canonical dispatch-repos.txt at runtime.
var defaultStuckRepos = []string{
"mathias/dispatch-sandbox", "mathias/cobalt-dingo", "mathias/parallax",
"mathias/cad-atlas", "mathias/brain-gardener", "mathias/hyperguild", "mathias/brain",
}
func stuckRepos() []string {
if v := os.Getenv("CAD_ATLAS_STUCK_REPOS"); v != "" {
var repos []string
for _, r := range strings.Split(v, ",") {
if r = strings.TrimSpace(r); r != "" {
repos = append(repos, r)
}
}
return repos
}
return defaultStuckRepos
}
// live cache: query the cluster at most once per TTL; fall back to the authored
@@ -147,6 +182,40 @@ func liveOverlay() liveOverlayData {
d.flux = &f
}
}
d.stuck = fetchStuck()
liveCache = d
return d
}
// fetchStuck re-derives cad#14's stuck worklist fresh on every liveOverlay
// call (no cache beyond the shared liveTTL) — the same "re-derive, never
// recall" discipline cad-stuck and cad-babysit both hold to. Read-only:
// gitearead.Client has no write method in its package. Degrades to nil
// (no badges rendered) on any error or when GITEA_TOKEN is unset, exactly
// like every other liveOverlay source — a forge outage costs a missing
// overlay, never a broken page.
func fetchStuck() []caddomain.StuckFinding {
token := os.Getenv("GITEA_TOKEN")
if token == "" {
return nil
}
base := os.Getenv("GITEA_BASE")
if base == "" || strings.Contains(base, ".svc.cluster.local") {
// The in-cluster Gitea service (internal/gitea's default) has no
// public HTTPS endpoint for gitearead's client to call with a
// bearer token in the same way; cad-stuck's own default (the public
// git.d-ma.be) is the correct base here regardless of what
// GITEA_BASE is set to for the read-only in-cluster fetchers above.
base = "https://git.d-ma.be"
}
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
defer cancel()
client := gitearead.New(base, token)
cfg := cadapp.DefaultStuckConfig()
cfg.BaseURL = base
findings, _, err := cadapp.RunStuckQuery(ctx, client, stuckRepos(), cfg)
if err != nil {
return nil
}
return findings
}