rose-ash/plans/business-logic-fed-flows.md

# Business logic as composition — a content-addressed DAG over pluggable substrates

**Vision (elevated 2026-07-02):** business logic IS art-dag. An object's behavior is a
**content-addressed DAG** (lib/artdag), declared on its **type** alongside content grammar +
allowed relations. Everything else is a pluggable ADAPTER — the same fold/adapter principle as
render-vs-execute-vs-deps, applied to execution/communication/deployment:

- **Behavior = an artdag DAG** — the invariant, content-addressed (`artdag/dag`, analyze/plan/
  optimize/schedule). Business logic, art media pipelines, workflows — all the same abstraction.
- **Execution = an injected RUNNER** (`artdag/run dag RUNNER cache`; `artdag/op-table-runner`).
  Substrates are runners on a **capability ladder**, same DAG throughout — a node DECLARES the
  capabilities it needs, a runner ADVERTISES what it supports, and the match is checked at bind
  time (fail fast, not a mystery at run time). artdag/analyze computes a DAG's required-capability
  set → its MINIMUM runner. So "simple in SX / durable in Erlang / distributed in celery-sx" is a
  DERIVED property of the DAG, not a human judgment; a trivial rule is a one-node DAG needing only
  `{effect}` and runs synchronously with zero ceremony. Business logic and media pipelines are the
  SAME structure (a content-addressed op-DAG); they differ only in the capabilities their nodes
  require, hence the runner. The ladder (capabilities each runner adds):
  - **op-table / execute-fold runner** — `{effect, branch, each}` — synchronous, local, in-request. Covers P0.
  - **Erlang runner** — `+ suspend/resume/wait` (durable), deterministic replay (flow-on-erlang).
  - **celery-sx runner** — `+ parallel, retry, offload` — distributed/durable task executor, "Celery the way it should have
    been" on erlang-on-sx, ZERO packages. It's a LEAN GLUE of parts we already have, not a
    reimplementation: broker = lib/persist KV (durable enqueue/claim/ack/visibility-timeout) ·
    worker pool = the er-scheduler / Erlang processes · result backend = content-addressed results
    (artdag keys by content-id → dedup/memoization FREE — Celery bolts this on badly) · retries/
    replay = flow-on-erlang · scheduling/fan-out/chords = artdag/schedule (minikanren CLP(FD)) +
    the DAG's topo batches · the plug point = artdag/op-table-runner. The genuinely-new code is
    small (~a few hundred lines): a durable queue + a worker loop (pull node → runner → write
    content-addressed result) + retry/backoff. **BUILD WHEN A DAG DEMANDS IT** — heavy compute,
    long-running/retryable tasks, or fan-out across machines — NOT for the synchronous P0.
  - **real-Celery over artdag/L1** — `+ gpu/heavy-compute` — the existing Python media pipeline (JAX/IPFS) as a runner.
- **Communication = an injected TRANSPORT** (`artdag/federation`, transport injected). Substrates:
  fed-sx (ActivityPub/next/), internal HMAC HTTP (services), IPFS (content-addressed). Because
  content-ids are global, a result computed on one instance is reusable on another by id.
- **Deployment = PLACEMENT** — a subdomain service, a fed-sx peer, an L1 worker: just where a
  runner runs. Not the essence.
- **State change → triggers a DAG** (over a transport) → executed by a runner → effects (data) a
  driver dispatches. fed-sx + Erlang is ONE adapter set (durable/federated), not THE architecture.

So: the TYPE carries content-grammar + allowed-relations + a **behavior DAG (+ triggers)**; the
object's state changes emit activities; the platform picks runner/transport/placement per context.

**Design (decided 2026-07-02; corrected after review):**
- **Activity log = every OBSERVABLE object-level state change** — the event source. NOT just CID
  deltas: relations write `edge:*` rows, NOT the record, so a relation change does NOT shift the
  CID. Two ActivityPub-faithful event classes: content/status change → a CID-carrying
  `Create`/`Update`; relation change → an `Add`/`Remove` referencing the edge.
- **Verbs are TRANSITIONS, not raw deltas.** `on-publish` = draft→published (fire-once), not every
  CID delta of a published post. Emitter picks the verb (Create/Update/Add/Remove/Delete). Global
  fire-once is the emitter's job (+ a durable inbox for federation); the seam only dedups in-call.
- **Triggers = declared subscriptions** — a type declares named triggers; flows fire only on
  matching ones. Log complete, execution precise.
- **Flows split by CAPABILITY (DERIVED, not a human "complexity" call):** every flow is a
  content-addressed op-DAG. A node declares the capabilities it needs (a `wait` needs `suspend`, a
  fan-out needs `parallel`, a heavy op needs `offload`); a runner advertises what it supports;
  artdag/analyze computes the DAG's required set → its MINIMUM runner; bind-time checks required ⊆
  runner (fail fast). So the sync/durable/distributed choice falls out of the DAG — a `{effect}`-only
  DAG runs synchronously (no ceremony); a DAG with a `wait` node auto-requires the Erlang runner.
  Business logic and media pipelines are the SAME structure — they differ only in node capabilities.
- **Effects are DATA; a DRIVER dispatches them** (perform no IO — a blocking call deadlocks the
  er-scheduler). The host is the driver for P0.
- **The type carries its whole contract:** fields+grammar (content) · allowed relations (external)
  · **behavior bindings** (triggers → DAG). All composition, all editable in the type-def editor.
- **CANONICAL ACTIVITY = the seam shape** `{:verb :actor :object <cid> :object-type :delta :prev
  :ts :id}`; each runner adapter MARSHALS it to its substrate (e.g. the Erlang runner → next/'s
  `[{type,…},{object,[…]}]` proplist via host/blog--activity->erl). DONE (P0.4): host/blog--publish-
  activity emits the canonical shape; host/blog--activity->erl is the RA marshaller (staged).

**Reference (one durable-runner substrate, verified):** `next/tests/triggers_e2e.sh` = 10/10 —
next/'s trigger_registry + flow_dispatch + blog_publish_digest (suspend/resume/guard/dedup). This
is what the ERLANG RUNNER ADAPTER wraps (Phase RA), not what P0 wires directly.

## The ADAPTER SEAM (design first — the contract every substrate plugs into)
The invariant is an **activity** (state-change event) + a **behavior DAG**. Everything between is
a swappable adapter — each a dict-of-functions (SX-native, like the fold domains). Six contracts:

1. **Activity** (content-addressed event): `{:verb "create"|"update"|"add"|"remove"|"delete"
   :actor :object <cid> :object-type :delta :prev <cid> :ts}`.
2. **Behavior binding** (declared on a TYPE): `{:on {:verb :object-type :guard} :dag <ref>}`. The
   TYPE carries content-grammar + allowed-relations + these bindings. NO :runner hint — the runner
   is DERIVED: the DAG's required-capability set (artdag/analyze over its nodes; a node declares
   `:needs` e.g. `#{suspend}`) selects the minimum runner that advertises them.
3. **Trigger-registry adapter** — `{:register! (fn spec dag hint) :match (fn activity -> [binding])}`.
   Impls: a local SX matcher / next/'s Erlang trigger_registry.
4. **Runner adapter** — `{:capabilities <set>  :run (fn dag env -> {:status "done"|"suspended"|
   "failed" :effects :resume :error})}`. :capabilities is what it ADVERTISES (op-table `{effect,
   branch, each}`; Erlang `+ suspend`; celery-sx `+ parallel, retry, offload`); the binder checks
   dag-required ⊆ runner-caps and fails fast otherwise. env = `{:activity :actor :ctx :effects
   :binding}` (:effects = injected external-read interfaces, deterministic for replay). `:results`
   is runner-INTERNAL. Durability = the runner's, not the DAG's. A durable runner that SUSPENDS is
   wired at construction to the transport's INBOUND channel and injects its out-of-band completion
   there (pump drains it).
5. **Transport adapter** — `{:emit (fn activity) :deliver (fn -> [activity])}`. `:emit` = outbound
   (log/publish), `:deliver` = inbound (peers + async runner completions). Impls: in-process /
   fed-sx (next/) / HTTP / IPFS. Content-ids global → results move by id.
6. **Effect driver** — `{:dispatch (fn effect -> [activity])}`. Perform the effect-as-data; may emit
   NEW activities (loop closure).

**Engine** (`behavior/make-engine {:triggers :runner :transport :driver :effects? :ctx-of?}`) →
`behavior/process(engine, activity)` / `behavior/pump(engine)`:
```
step(activity):  if activity.id already :seen → skip (dedup)      ; in-call cycle guard
  emit(transport, activity)                                       ; log
  for b in match(triggers, activity):
     r = run(runner, b.dag, {:activity :actor :ctx :effects :binding})
     case r.status:
       "done"      → for eff in r.effects: for a in dispatch(driver, eff): step(a)   ; loop closes
       "suspended" → record (+ r.resume); a durable runner completes out-of-band → inbound → pump
       "failed"    → record (+ r.error) for flow-level retry / dead-letter
pump():  for a in transport.deliver(): step(a)                    ; inbound drain, shared :seen
```
Every stage injected → swap runner (sync→Erlang→celery-sx), transport (in-proc→fed-sx), registry,
driver — DAG + engine unchanged. **Global** idempotency = emitter fire-once + a durable inbox.

**Build order:** (a) **DONE** — seam as SX contracts + reference engine, tested substrate-agnostic:
:status branch (done/suspended/failed), injected env (:ctx + :effects), dedup by activity :id,
behavior/pump for inbound, async-completion loop (suspend → out-of-band inject → pump). **behavior
10/10.** (b) P0 supplies the REAL SYNCHRONOUS adapters; durable/federated runners+transports are
later adapter phases (RA/TA below).

## P0 — publish workflow on the seam (SYNCHRONOUS, all-SX)
Prove: live host publish → the seam engine (in-process transport · local-SX trigger registry ·
sync op-table runner over an SX publish-DAG · host driver) → the effect surfaces. NO Erlang, NO
fed-sx yet — those are adapter phases (RA/TA). Every piece swaps later; the DAG + engine don't.

- [x] **P0.1 — publish-activity contract (SX side).** host/blog--publish-activity + post-category.
  blog 200/200. NOTE: emits the next/-Erlang shape today; P0.4 reconciles to the canonical seam shape.
- [x] **P0.2 — the publish-DAG + execute-fold runner + the CAPABILITY check. DONE 2026-07-02.**
  **HYPOTHESIS-TEST FINDING:** the synchronous business flow expresses NATURALLY as an EXECUTE-FOLD
  composition (host/execute.sx: seq/effect/alt — the branch on category IS `alt`, exactly what it's
  for), NOT an artdag DAG — artdag is pure DATAFLOW with no control flow. So "business logic = art-
  dag" is confirmed at the ABSTRACTION (both content-addressed op-DAGs) and REFINED at the vocabulary:
  the SYNCHRONOUS control-flow runner is the execute-fold (caps {effect,branch,each}); artdag is the
  DATAFLOW sibling (a different runner). Two instances of one thing, run very differently — as
  predicted. Built: lib/host/flows.sx (host/flow--{node-cap, required-caps, subset?, exec-runner,
  bind}); host/blog--publish-dag + publish-ctx. Verified: publish-DAG required-caps derived =
  {effect,branch} → binds to exec-runner; runs → newsletter→[validate,digest]/urgent→[validate,
  notify]/other→[validate,skip]; a `wait` node → required {suspend} → binds FAIL-FAST against the
  exec-runner (would need Erlang, RA). flows 7/7, blog 203/203, conformance 591/591.
  IMPLICATION for RA/TA: the Erlang runner isn't a "different flow language" — it's the SAME op-DAG
  with +{suspend} nodes; RA is the runner that advertises suspend + wraps flow_dispatch.
  CAVEAT (don't calcify this finding): execute-fold-vs-artdag is a CURRENT capability SNAPSHOT, NOT
  a permanent boundary. artdag MAY GROW control-flow node-kinds (a runner advertising +{effect,
  branch, each}), and business logic then MIGRATES to artdag to inherit content-addressed
  memoization / optimize (fuse/dedup/dce) / schedule / FEDERATION (a flow result reused across peers
  by content-id — the federation vision, free). The capability model makes that migration seamless
  (same DAGs, richer runner; the execute-fold is just the pragmatic sync runner NOW). See phase AX.
- [x] **P0.3 — wire the seam on the live host. DONE + LIVE-VERIFIED 2026-07-02.** host/blog--
  {transport, triggers (on-publish: create+article → publish-DAG), driver (records each effect),
  publish-engine, fire-publish!, maybe-publish!}. Both write handlers (form-submit POST /new,
  edit-submit POST /:slug/edit) call maybe-publish!(slug, prev-status, new-status) — a non-published
  → published TRANSITION fires the flow (fire-once), in the handler BODY. /flows renders the flow
  log. LIVE PROOF: logged in + POST /new on blog.rose-ash.com → /flows shows `validate` + `notify`
  (category defaulted to urgent). behavior→exec-runner→driver all real. blog 207/207, conformance
  595/595. GAP: the flow log is IN-MEMORY (clears on restart) — "durable record" is P0.3b (persist
  the log to the blog store + boot-load, string-keyed to dodge the keyword/persist split). Also: the
  live test post `p0.3-seam-live-test` persists (no delete route) — harmless, clean up if wanted.
- [x] **P0.4 — canonical activity + reconcile. DONE + LIVE-VERIFIED 2026-07-02.** host/blog--
  publish-activity now emits the CANONICAL seam shape {:verb :actor :object <cid> :object-type :slug
  :category :delta :id} — :object is a content-addressed REFERENCE (the CID, was an inlined dict),
  :id the dedup identity, :slug+:category the domain fields the DAG reads. Consumers reconciled: the
  trigger matches :verb+:object-type; publish-ctx reads top-level :category+:slug. Added the runner
  MARSHALLER host/blog--activity->erl (canonical → next/'s proplist, for RA — defined+tested, unused
  until RA). (:ts/:prev omitted — no clock primitive in the host; deferred.) LIVE: published on
  blog.rose-ash.com → /flows fired validate+notify with the canonical activity. blog 209/209,
  conformance 597/597.

**P0 COMPLETE** — the synchronous publish workflow runs end-to-end on the LIVE host through the
substrate-agnostic seam, durably, in the canonical activity shape, with the Erlang-runner marshaller
staged. Every piece is a swappable adapter: RA (Erlang runner) + TA (fed-sx transport) plug in next
without touching the DAG or the wiring.

### P0 REVIEW (2026-07-02) — findings + carried-forward debt
- **FIXED (was a live bug):** edit-submit ran maybe-publish! BEFORE set-field-values!, so an edit
  that set a category AND published branched on the STALE category. Reordered (fields first);
  regression test added. blog 210/210.
- **DEBT #1 (blocks P2): activity identity.** Dedup uses :id = the object CID. Relation Add/Remove
  events don't change the CID → multiple relation activities would share :id → false dedup. P2 needs
  a real activity id (verb+object+seq/edge), NOT the bare CID.
- **DEBT #2 (P1): the capability bind isn't wired into the live engine.** host/blog--publish-engine
  calls exec-runner directly; host/flow--bind is tested but decorative in the live path. P1's engine
  must DERIVE the runner via bind (required ⊆ advertised), not hardcode it.
- **DEBT #3 (RA): flow execution is synchronous in the request path.** Fine for 2 cheap effects, but
  a durable/suspend runner CANNOT block a request — RA needs dispatch moved OFF the request path
  (emit in-request → a background loop calls behavior/pump). The seam supports it (pump); the wiring
  doesn't exist. This is structural, not just the marshaller.
- **DEBT #4 (minor): the "urgent" category default** notifies everyone for an uncategorised post —
  demo-convenient, semantically wrong. Default to skip/none once the demo value is past.
- Acknowledged/known: unbounded flow log (cap/rotate), registry :register! is a stub (P1 makes it
  real), :actor "site" placeholder (TA needs the actor model), marshaller unverified-until-RA.

### Sequencing note (hidden prerequisites)
- **P1's "runner derived via caps" is near-vacuous until a SECOND runner exists** — with only
  exec-runner, everything derives to it. The derivation only MEANS something once RA lands.
- **RA is where the real risk lives** (SX→erlang-on-sx dispatch + suspend→resume→pump + the
  er-scheduler context bug) AND P1/P2 quietly depend on what RA forces (a 2nd runner, the async
  boundary). RECOMMENDATION: a narrow **RA SPIKE** next — prove one dispatch + one suspend/resume/
  pump cycle in isolation — de-risks the whole durable/federated half before building P1/P2 on it.

## P1 — types DECLARE behavior (generalize) — DONE + LIVE-VERIFIED 2026-07-02
- [x] The type carries :behavior — a list of flat string-keyed bindings {"verb" "type" "dag"}
  (persist-safe, like :type-relations), stored on the type-post (host/blog--type-behavior /
  set-type-behavior!). The "article" type declares {"verb" "create" "type" "article" "dag" "publish"}.
- [x] The behavior REGISTRY (host/blog--behaviors) is gathered at boot from ALL posts' declarations
  (host/blog--load-behaviors!, in serve.sh after seed-types!); the trigger match (host/blog--triggers
  :match = host/blog--match-behaviors) consults it. The hardcoded create+article trigger is GONE.
- [x] The runner is DERIVED, not hinted (DEBT #2 fixed): match-behaviors resolves the :dag via a DAG
  registry (host/blog--dag-registry) and picks the runner via host/flow--select-runner over the fleet
  (host/blog--runner-fleet = [exec-runner]; RA joins at RA-live). Each binding carries its :runner;
  behavior/-run-binding uses it. An {effect,branch} publish-DAG → exec-runner; a {suspend} DAG would
  route to RA (proven in ra 9/9 with a 2-runner fleet).
- [x] The type-def view SHOWS each behavior + its derived runner (host/blog--behavior-lines): LIVE at
  blog.rose-ash.com/article — "on create → publish DAG · needs {effect, branch} · runner:
  synchronous (exec-fold)". Derived + visible, not hand-set.
- LIVE PROOF: published on blog.rose-ash.com → /flows fired validate+notify via the DECLARED path
  (registry + derived runner). blog 213/213, full conformance 610/610. FINDING: load-behaviors! must
  scan ALL posts, not filter by is-type? (article didn't pass is-type? on the durable store though it
  did in-memory) — the type declaration is authoritative, the is-type? classification isn't reliable enough.

## P2 — state-change → activity emission (ALL events) — DONE + LIVE-VERIFIED 2026-07-02
- [x] TWO event classes emit canonical activities through the seam: CONTENT (host/blog--content-
  activity: Create on first publish, Update on a subsequent published edit — object-type DERIVED from
  is-a, not hardcoded) and RELATION (host/blog--relation-activity: Add/Remove, carrying :relation +
  :target). host/blog--emit! runs any activity through behavior/process (logged + matched);
  emit-content-change! (create/update) wired into form-submit + edit-submit; emit-relation!
  (add/remove) wired into relate-submit + unrelate-submit.
- [x] DEBT #1 FIXED — per-EVENT :id, not the bare CID. Content = "create:"/"update:"+cid; relation =
  "add:"/"remove:"+src:kind:dst (EDGE-based, since a relation change doesn't shift the CID, so a
  CID-based id would false-dedup different edges on one object). Verified: different edges → different ids.
- [x] The activity log is the DURABLE EVENT SOURCE (host/blog--activity-log, string-keyed records
  persisted under "activitylog", boot-loaded via host/blog-load-activitylog!). Surfaced at /activities.
  This is what TA will push to peers.
- LIVE PROOF: on blog.rose-ash.com — publish → /activities "create article <cid>"; relate → "add
  article p2-events — add welcome related"; unrelate → "remove …". All three classes, durable.
  blog 217/217 (+4 P2, reframed P0.3 fire-once tests for Update semantics), full conformance 614/614.

## RA — the ERLANG (durable) RUNNER adapter   ← the old "fed-sx spike", now an adapter
<!-- PREREQ (review): move dispatch OFF the request path (DEBT #3) — background loop calls behavior/pump; suspend can't block a request. -->
**SPIKE DONE 2026-07-02 — RA is VIABLE (plans/ra-spike.sh, 4/4).** Proven from the SX side:
(1) our canonical activity dict → Erlang activity-proplist SOURCE (ra/erl-src ≈ host/blog--activity->erl)
is valid Erlang the flow consumes; (2) it drives pipeline:apply_triggers → blog_publish_digest → done
+ 3 emails (urgent sync branch); (3) the newsletter activity SUSPENDS on the morning timer
(status =:= {ok,{suspended,morning}}); (4) flow_store:resume(Id, morning_ts) COMPLETES it → 3 emails
(the async cycle); (5) NO er-scheduler deadlock — flow-on-erlang's railway threading holds when
driven from SX. KEY FINDINGS for the build: erlang-eval-ast returns Erlang TERMS directly (integers
raw, atoms as {:tag atom :name …}) — the runner must parse results, not assume :name; flow_store
start→{done,V}|{suspended,Tag}, resume(Id,Res) maps 1:1 onto {:status done|suspended :effects :resume};
the flow instance Id is the resume handle.
- [x] **RA RUNNER BUILT + TESTED (module + integration) 2026-07-02.** lib/host/ra.sx — a PURE-SX
  seam runner (advertises {effect,branch,each,suspend}) with an INJECTED erl-eval (real =
  er-to-sx-deep ∘ erlang-eval-ast; mock in unit tests), so it loads in the plain host and is testable
  without the Erlang runtime. host/ra--{atom,bin,erl-src,start-expr,resume-expr,parse,make-runner,
  resume,real-eval}: marshals our canonical activity → Erlang source (CID as <<"…">> binary, atoms
  single-quoted), starts flow_store, parses (ok Id (flow_done V))→{:status done :effects V :flow-id} /
  (ok Id (flow_suspended T))→{:status suspended :resume {:id :tag}}. DUAL-RUNNER ROUTING in flows.sx:
  host/flow--required-caps handles a {:erl-flow :needs} DAG (declared caps); host/flow--select-runner
  picks the cheapest runner covering the DAG's needs — the capability model is now REAL (2 runners:
  an {effect,branch} composition → exec-runner; a {suspend} DAG → RA). ra 9/9 (mock) + plans/
  ra-integration.sh 4/4 (the REAL module driving live flow_store: urgent→done, newsletter→suspended
  with resume handle, effect-as-data carried). Full host conformance green. next/tests/triggers_e2e.sh
  10/10 baseline intact.
- [ ] **RA-LIVE (deferred — the deployment step, prerequisite now PRECISE).** KEY FINDING: gen_servers
  do NOT persist across separate erlang-eval-ast calls (flow README: "the scheduler doesn't preserve
  spawned processes across separate erlang-eval-ast invocations"). So a boot-per-call proves the
  module (done), but TRUE async (suspend → return the request → resume LATER in another call) needs a
  PERSISTENT next/ kernel PROCESS holding flow_store — the async boundary (DEBT #3) is deeper than
  "off the request path". REMAINING: (a) stand up a long-lived next/ kernel (nx_kernel/http_server
  already run persistently for TCP) that RA talks to; (b) wire a DURABLE behavior binding ({:erl-flow
  "blog_digest" :needs (effect branch suspend)}) into the live publish engine, routed to RA via
  select-runner; (c) the resumed completion re-enters via the transport inbound + behavior/pump.
  The runner + marshalling + suspend/resume mechanics are all proven; this is process lifecycle + wiring.

## TA — the FED-SX TRANSPORT adapter   ← federation proper
- [x] **TA TRANSPORT BUILT + the federation LOOP PROVEN 2026-07-02.** lib/host/ta.sx — a seam
  transport {:emit :deliver} over a DIRECTIONAL wire (out=outbox→followers, in=inbox←follows). The
  transport is the SERIALIZATION boundary: activities cross as SX-source strings via host/ta--
  serialize/deserialize (keyword-keyed activity ↔ flat string-keyed wire form — the P2 activity
  fields). host/ta--make-transport(out-wire, in-wire) + host/ta--make-mem-wire (an in-memory
  directional queue). PROVEN (ta 5/5): a content + a relation activity round-trip through the wire;
  the FEDERATION LOOP — instance A emits an activity → the wire carries it → instance B's
  behavior/pump delivers + processes it → B's engine fires ITS behavior on A's activity; DIRECTIONAL
  (B re-emits to its own outbox, not back into the inbox — no loop). This is "everything works over
  fed-sx" proven at the seam. Full host conformance green.
- [ ] **TA-LIVE (deferred — same shape as RA-live).** Swap the mem-wire for the REAL next/ delivery
  wire (outbox → http_server → peer inbox). Needs: (a) a PERSISTENT next/ kernel process (gen_servers
  don't survive across erlang-eval-ast calls — the RA-live finding; next/ outbox/http_server are the
  persistent side); (b) the ACTOR MODEL real (:actor is a "site" placeholder — peer_actors /
  follower_graph / per-author identity decide WHO the out-wire delivers to); (c) push /activities
  (the P2 event source) onto the out-wire. RISK: next/ delivery M2 blockers (er-scheduler context).
  The transport contract + serialization + the loop are proven; TA-live is the wire impl + placement.

## AX — artdag GROWS control-flow (business logic MIGRATES to artdag) [DEMAND-DRIVEN]
Today artdag is pure dataflow and the execute-fold is the synchronous control-flow runner. That's
a capability snapshot. When a business flow WANTS the artdag engine's benefits — content-addressed
memoization (recompute only on input-CID change), optimize (fuse/dedup/dce), schedule, and above
all FEDERATION (a flow result reused across peers by content-id) — grow artdag's node vocabulary
so its runner advertises `+{effect, branch, each}`, and the SAME behavior DAGs migrate onto it
(the capability model makes this seamless; the seam + DAGs don't change). Two real design pieces:
(1) DYNAMIC control in a static DAG — a `branch` PRUNES a path (conditional nodes make downstream
nodes live/dead; content-addressing holds on the taken path); (2) EFFECT nodes vs memoization —
pure nodes memoize, `{effect}` nodes are marked non-cacheable (must run / be idempotent). Build
when a flow's cost, reuse, or cross-peer sharing makes the execute-fold's re-run-everything
insufficient — not before. The execute-fold stays the lean default for cheap synchronous flows.

## RX — celery-sx runner (DEMAND-DRIVEN, not scheduled)
Build the distributed/durable runner adapter the moment a real DAG needs heavy compute /
long-running-retryable tasks / cross-machine fan-out (the artdag/JAX media case, or federated
flows that can't run in-request). New code is small — glue persist (durable queue: enqueue/claim/
ack/visibility-timeout) + er-scheduler (worker loop: pull node → op-table-runner → content-
addressed result) + artdag/schedule (fan-out) + retry/backoff. Slots in at artdag/op-table-runner
alongside the synchronous + Erlang runners. Zero packages. Do NOT pre-build; the op-table runner
covers everything until a DAG's cost/latency/placement forces the substrate.

## P4 — close the loop
- [ ] Flow effects mutate objects back durably (a flow's DescribeEffects → host writes / new
  activities), so business logic can change state, which federates, which triggers more flows.

## Progress log (newest first)
- 2026-07-02 — TA TRANSPORT built + the federation LOOP proven (lib/host/ta.sx, ta 5/5). A seam
  transport over a directional wire (serialization boundary; activities cross as SX-source). Proven
  in-memory: A emits → wire → B pump → B's engine fires ITS behavior on A's activity (directional, no
  loop). "Everything works over fed-sx" proven at the seam. TA-live (real next/ delivery wire) deferred
  — needs the persistent kernel (RA-live finding) + the actor model (who to deliver to). NEXT: the
  DISTRIBUTED HALF is now all proven-at-the-seam (RA + TA); the remaining live steps (RA-live, TA-live)
  share ONE prerequisite — a persistent next/ kernel process + the actor model. Or P4/AX/RX.
- 2026-07-02 — P2 DONE + LIVE-VERIFIED. All observable state changes now emit canonical activities
  through the seam: content Create/Update + relation Add/Remove. DEBT #1 fixed (per-event ids; edge-
  based for relations). The activity log is the durable event source, surfaced at /activities. Live:
  publish→create, relate→add, unrelate→remove all logged. blog 217/217, conformance 614/614. The
  event source is now complete + federatable — NEXT: RA-live (persistent kernel) or TA (fed-sx
  transport pushes /activities to peers → federation).
- 2026-07-02 — P1 DONE + LIVE-VERIFIED. Types DECLARE :behavior (stored on the type-post, gathered
  into a registry at boot); the trigger match consults the registry; the runner is DERIVED via
  host/flow--select-runner over the fleet (DEBT #2 fixed — no hardcoded trigger, no runner hint). The
  /article page shows the declared behavior + derived runner. Published live → /flows fired via the
  declared path. blog 213/213, conformance 610/610. Finding: load-behaviors! scans ALL posts (not
  is-type?-filtered — unreliable on the durable store). NEXT: RA-live (persistent kernel wires RA into
  the fleet → durable bindings route to RA), or P2 (all state-change → activity emission).
- 2026-07-02 — RA RUNNER BUILT + tested (module + integration). lib/host/ra.sx = a pure-SX seam
  runner with injected erl-eval (loads in the plain host, mock-testable); marshals our activity →
  Erlang, drives flow_store, parses done/suspended → the runner contract. Dual-runner ROUTING in
  flows.sx (host/flow--select-runner + required-caps for {:erl-flow :needs} DAGs) makes the capability
  model REAL (2 runners). ra 9/9 (mock) + plans/ra-integration.sh 4/4 (REAL module → live flow_store).
  Full host conformance 607/607. FINDING: gen_servers don't persist across erlang-eval-ast calls, so
  RA-LIVE (true cross-call suspend/resume) needs a persistent next/ kernel process — the async
  boundary is deeper than "off the request path". Runner mechanics fully proven; RA-live = lifecycle
  + wiring. NEXT: RA-live (persistent kernel + a durable binding wired to RA), or P1 (capability model
  is now real, so it's no longer vacuous).
- 2026-07-02 — RA SPIKE DONE → RA is VIABLE (plans/ra-spike.sh, 4/4). From SX: our canonical activity
  serializes to valid Erlang, drives blog_publish_digest through flow_store (done + suspend + resume),
  no er-scheduler deadlock. De-risks the whole durable/federated half. Findings: erlang-eval-ast
  returns terms directly (parse, don't assume :name); flow_store start/resume maps 1:1 onto the runner
  contract. Remaining for full RA: load the Erlang runtime into serving (or out-of-process), the async
  dispatch boundary (DEBT #3), CID→binary marshalling, structured result parsing. NEXT: full RA, or
  P1 now that a real 2nd runner is proven reachable.
- 2026-07-02 — P0.4 DONE + LIVE-VERIFIED → **P0 COMPLETE**. Canonical seam activity shape
  ({:verb :object=cid :object-type :slug :category :delta :id}); consumers reconciled (trigger match,
  publish-ctx); host/blog--activity->erl marshaller staged for RA. Published live → /flows fired
  validate+notify with the canonical activity. blog 209/209, conformance 597/597. The synchronous
  publish workflow is end-to-end on the live host through the substrate-agnostic seam, durable, in
  the canonical shape. NEXT: P1 (types declare :behavior, engine built per type, runner derived via
  caps) — or RA (Erlang durable runner, the marshaller is ready).
- 2026-07-02 — P0.3b DONE + LIVE-VERIFIED. The flow log is now DURABLE: the driver
  persists string-keyed effect records to the blog store (dodging the keyword/persist top-level
  split); host/blog-load-flowlog! rebuilds it on boot (serve.sh). Proof: published on
  blog.rose-ash.com, RESTARTED the container, /flows still showed validate+notify (reloaded from the
  store). blog 208/208, conformance 596/596. Whole-list rewrite per effect — cap/rotate later.
- 2026-07-02 — P0.3 DONE + LIVE-VERIFIED. The seam wired into the live publish path: on-publish
  registry + in-process transport + host driver + the execute-fold runner, fired by the draft→
  published transition in both write handlers. Published a real post on blog.rose-ash.com → /flows
  surfaced validate + notify, driven by the actual behavior engine. blog 207/207, conformance
  595/595. NEXT: P0.4 (canonical activity shape) or P0.3b (durable flow log); then P1 (types declare
  behavior — build the engine per type from its :behavior bindings, runner derived via caps).
- 2026-07-02 — DON'T-CALCIFY note (user: "artdag may in the future contain business logic"). The
  execute-fold-vs-artdag split from P0.2 is a capability SNAPSHOT, not a boundary. Added phase AX:
  artdag grows +{effect,branch,each} node-kinds and business logic migrates onto it to inherit
  content-addressed memoization / optimize / FEDERATION (flow result reused across peers by CID —
  the federation vision, free). Design work named: dynamic control (branch prunes) in a static DAG;
  effect nodes non-cacheable vs pure nodes memoized. Demand-driven; execute-fold stays the lean
  default. P0.2 finding + flows.sx header annotated so the finding doesn't harden.
- 2026-07-02 — P0.2 DONE + the hypothesis CONFIRMED (and refined). The synchronous publish workflow
  is NATURAL as an execute-fold composition (seq/effect/alt), NOT artdag dataflow (no branch there).
  So business-logic = art-dag holds at the abstraction (content-addressed op-DAG) but the SYNCHRONOUS
  runner is the execute-fold, artdag the dataflow sibling — two instances, run differently, exactly
  the framing. lib/host/flows.sx (capability layer + exec-runner + bind) + host/blog--publish-dag.
  Runner DERIVED via required-caps ⊆ advertised; wait→fail-fast. flows 7/7, blog 203/203, 591/591.
- 2026-07-02 — folded in CAPABILITY-TYPED nodes / CAPABILITY-ADVERTISING runners. A node declares
  `:needs` (wait→suspend, fan-out→parallel, heavy→offload); a runner advertises `:capabilities`
  (op-table {effect,branch,each}; Erlang +suspend; celery-sx +parallel,retry,offload); artdag/analyze
  computes a DAG's min-runner; the binder checks required ⊆ runner-caps (fail fast). So the sync/
  durable/distributed split is DERIVED from the DAG, not a human call — a {effect}-only DAG runs with
  zero ceremony; a wait node auto-requires Erlang. Removed the :runner hint from the binding. P0.2
  gains the hypothesis test: does the publish workflow express naturally as a DAG, and does flipping
  a node to `wait` fail-fast against the op-table runner? Clarifies "business logic = art-dag" — same
  op-DAG structure, differing only in node capabilities, hence runner. (Insight: they're two
  instances of one thing, suitable to run very differently.)
- 2026-07-02 — whole-plan coherence review. The reframe (artdag+seam) had left the middle stale:
  P0.2/P0.3 still described the Erlang-bridge-first path; P0.1's activity didn't match the seam
  contract; the seam section predated the enrichment. Fixed: P0 rewritten around the seam + SX
  op-table runner (all-SX, no Erlang/fed-sx); the Erlang/fed-sx path DEMOTED to explicit adapter
  phases RA (durable runner) + TA (fed-sx transport); canonical activity shape + P0.4 reconcile;
  seam contract refreshed to behavior.sx (status/dedup/pump/async, behavior 10/10); stray bits cleared.
- 2026-07-02 — seam DONE + reviewed twice. lib/host/behavior.sx (engine + 4 adapters); enriched
  substrate-agnostic (status/env/dedup/pump); 2nd review corrected the async-completion contract
  (construction-wired inbound + pump, not env :emit) + proved it. behavior 10/10, conformance 580.
- 2026-07-02 — P0.1 done. host/blog--publish-activity + host/blog--post-category; the publish
  contract in SX, 200/200. Verified next/ triggers e2e baseline 10/10. Roadmap anchored. NEXT:
  P0.2 the dispatch bridge (in-process: serve.sh loads next/ kernel + registers the on-publish
  trigger; host emits the activity via the erlang-on-sx bridge to pipeline:apply_triggers).