toga-ai 1.0.973 → 1.0.974

package/CLAUDE.md

@@ -232,7 +232,7 @@ See `package.json` for full details.
 
 ## TOGA Technology Claude Harness
 
-Installed from: npm bundle v1.0.174
+Installed from: npm bundle v1.0.768
 
 ---
 

package/knowledge/1.0/apps/tools/features/talos-pricing-ui.md

@@ -25,9 +25,9 @@ related:
   - ./persona-gated-navigation.md
   - ./mvc-data-access-patterns.md
   - ../architecture.md
-  - ../../../2.0/apps/talos/architecture.md
-  - ../../../2.0/apps/talos/features/pricing-cogs-model.md
-  - ../../../2.0/apps/worker2/features/talos-pricing-automation.md
+  - ../../../../standalone/apps/talos-backend/architecture.md
+  - ../../../../2.0/apps/worker2/features/talos-pricing-cogs-model.md
+  - ../../../../2.0/apps/worker2/features/talos-pricing-automation.md
 ---
 
 The 1.0 (tools) face of the Talos Pricing Platform — Contracts / Pricing Dashboard / Usage / Pricing Settings pages + the estimate engine, reading the Team DB via `db_team`; open for the contract lifecycle, worker-owned read-only rows, or the fetchOne/PHP 8.5 traps.
@@ -43,7 +43,7 @@ The old Onboarding page could only **INSERT** (hardcoded `status='ONBOARDED'`, n
 - **A signed contract can never be hard-deleted.** `TalosPricingBands` is `ON DELETE CASCADE`, so deleting the client would destroy the agreed rate schedule and the margin history with it. Signed contracts **churn**; they do not delete.
 - **Post-signature edits append to `TalosContractAmendments`** rather than mutating the signed terms.
 - Includes an **Actual-vs-Estimate** comparison with variance chips, a month-by-month history table plus cost/margin chart, and a one-click **"apply recommended org fee"**.
-- **Pricing model toggle + contract terms (2026-09-17).** The page now sets `pricingModel` (`PER_USER` / `PER_INTERACTION`), the fixed 1/3/5-year term dropdown (`contractTermMonths`), the per-interaction inputs (`contractedInteractionCount`, voice-only `avgMinutesPerInteraction`, `perInteractionFee`), and the two one-time contract fees (`implementationFee`, `trainingFee`). Model design, term ramp, fee amortization, and the recurring-vs-blended margin split are all in [pricing-cogs-model](../../../2.0/apps/talos/features/pricing-cogs-model.md) — not restated here.
+- **Pricing model toggle + contract terms (2026-09-17).** The page now sets `pricingModel` (`PER_USER` / `PER_INTERACTION`), the fixed 1/3/5-year term dropdown (`contractTermMonths`), the per-interaction inputs (`contractedInteractionCount`, voice-only `avgMinutesPerInteraction`, `perInteractionFee`), and the two one-time contract fees (`implementationFee`, `trainingFee`). Model design, term ramp, fee amortization, and the recurring-vs-blended margin split are all in [pricing-cogs-model](../../../../2.0/apps/worker2/features/talos-pricing-cogs-model.md) — not restated here.
 
 ### Pricing Dashboard (`pricing/get.php`) — rebuilt for sales leadership
 Portfolio KPIs, a 12-month blended-margin trend with the target band shaded, an exception list with deterministic explanations, and licensed-vs-active utilisation.
@@ -59,7 +59,7 @@ Technical-only (persona-narrowed on the action, per `mvc-data-access-patterns`).
 - `mvc/talos/onboarding/*` and `mvc/talos/factors/*` — superseded above.
 
 ### Estimator (`App_Talos_Estimator`, `_/app/talos/estimator.php`)
-Now driven by the **token unit price** and **workload profiles** instead of the feature checklist and per-client calibration factor (methodology in `pricing-cogs-model`). Service offering (CHAT / VOICE_TO_VOICE / NATURAL_VOICE) describes **modality only** and says nothing about cost; cost is driven by tool-call **intensity** via the profile (retrieval / general / analytics). Per-interaction model adds `costPerInteraction`, `estimateByInteraction`, `recommendedPricePerInteraction`, `recommendedOrgFeeInteraction`, `orgFeeOptionsInteraction`. Constants live in Team-DB `TalosCostFactors` (shared with worker2); a golden-master parity test (`estimator.parity.test.php` ↔ worker2 `PricingParityTest.php`, 11/11 pass) guards drift — see [pricing-cogs-model](../../../2.0/apps/talos/features/pricing-cogs-model.md).
+Now driven by the **token unit price** and **workload profiles** instead of the feature checklist and per-client calibration factor (methodology in `pricing-cogs-model`). Service offering (CHAT / VOICE_TO_VOICE / NATURAL_VOICE) describes **modality only** and says nothing about cost; cost is driven by tool-call **intensity** via the profile (retrieval / general / analytics). Per-interaction model adds `costPerInteraction`, `estimateByInteraction`, `recommendedPricePerInteraction`, `recommendedOrgFeeInteraction`, `orgFeeOptionsInteraction`. Constants live in Team-DB `TalosCostFactors` (shared with worker2); a golden-master parity test (`estimator.parity.test.php` ↔ worker2 `PricingParityTest.php`, 11/11 pass) guards drift — see [pricing-cogs-model](../../../../2.0/apps/worker2/features/talos-pricing-cogs-model.md).
 
 ## Gotchas
 - **`App_Database::fetchOne()` is not an existence test** — it 500s on zero rows. See [mvc-data-access-patterns](./mvc-data-access-patterns.md); the original `onboarding/post.php` carried this exact bug and would have fataled on the first save anyone attempted, which is good evidence nobody ever completed an onboarding through it.
@@ -80,6 +80,6 @@ Now driven by the **token unit price** and **workload profiles** instead of the
 - [Persona-gated navigation](./persona-gated-navigation.md)
 - [Tools MVC — data-access patterns](./mvc-data-access-patterns.md)
 - [Tools architecture](../architecture.md)
-- [Talos architecture (2.0)](../../../2.0/apps/talos/architecture.md)
-- [Pricing COGS model (2.0)](../../../2.0/apps/talos/features/pricing-cogs-model.md)
-- [Talos pricing automation (worker2)](../../../2.0/apps/worker2/features/talos-pricing-automation.md)
+- [Talos architecture (standalone)](../../../../standalone/apps/talos-backend/architecture.md)
+- [Pricing COGS model (worker2)](../../../../2.0/apps/worker2/features/talos-pricing-cogs-model.md)
+- [Talos pricing automation (worker2)](../../../../2.0/apps/worker2/features/talos-pricing-automation.md)

package/knowledge/2.0/apps/_underscore/features/email-send-pipeline.md

@@ -18,7 +18,7 @@ related:
   - email-template-sending.md
   - string-html-entity-helpers.md
   - ../../worker2/features/all-client-email-queue-monitor.md
-  - ../../ai-bdr/features/web-funnel-app.md
+  - ../../../../standalone/apps/bdr/features/web-funnel-app.md
   - ../../../clients/compass-canada/features/french-order-email-localization.md
 ---
 
@@ -169,7 +169,7 @@ derive one from the other).
   be paired with updating BDR's env, or BDR's Send button can only reach its *failed* state (it
   has no mail-app fallback). Settings BDR mirrors: AWS SES **us-west-2**, port **587**, TLS;
   `DevTeam@goagilant.com` is a known-good verified sender identity in worker2 config. See
-  [ai-bdr web-funnel-app.md](../../ai-bdr/features/web-funnel-app.md).
+  [ai-bdr web-funnel-app.md](../../../../standalone/apps/bdr/features/web-funnel-app.md).
 
 - **⚠ READ-AFTER-WRITE ON THE READER REPLICA LOSES `send()`'S OWN LOG ROW (deterministic).**
   `_Email::send()` INSERTs the `PENDING` `Logs_[Client].Email` row, then reads `$emailLog->id`

package/knowledge/2.0/apps/api2/features/nested-relationship-writes.md

@@ -22,7 +22,7 @@ related:
   - ../../toga2-supply/features/currency-amount-lines-editor.md
   - cxml-shipnotice-gateway.md
   - ../../../clients/aig/features/entitlement-intake.md
-  - ../../ai-bdr/features/web-funnel-app.md
+  - ../../../../standalone/apps/bdr/features/web-funnel-app.md
   - ../../worker2/features/vapi-webhook-handler.md
   - ../../../clients/elite/features/netsuite-togasupply-sync.md
   - ../../dbchanges2/workflows/client-schema-drift-audit.md
@@ -52,7 +52,7 @@ Implemented in `Component/Api/V2/V2.php` (`getForeignKeyValue()` / the nested-wr
 To **create** a new related record, send its data **without** an identifier (and accept that a new row is written). Never rely on a non-identifier field (a custom field, a name, an email) to de-duplicate — the engine will not use it to match, and you accumulate duplicate empty child rows on every write. If you need to link by a business key (like an employee XID), **resolve that key to a uuid in your caller first** (e.g. a `key → uuid` map), then send `{uuid}`.
 
 **UPDATE-path back-reference injection defeats the uuid forced-MATCH (primary-pointer bug).** A **single-key identifier object** (`{uuid: <x>}` and nothing else) is supposed to hit the engine's **forced-MATCH fast-path** (`V2.php:6952`) — link the existing child, never create, and let the caller re-point a parent's forward FK to it. On the **CREATE** path this works: the set-loop (`V2.php:4601–4616`) applies the child object and the parent FK is assigned at `V2.php:4647` with no interference. On the **UPDATE** path it does **not**. When a child model carries a `contactId` FK *back* to the parent (as `ContactPhoneNumbers` → `Contacts` does), the update path **injects a reverse back-reference** (sets the child's `contactId = parent id`) into the nested child object at `V2.php:4985–4996` — added by commit `91d803f` "guard child FK set". That injection turns the single-key object into a **multi-key** object, so it no longer qualifies for forced-MATCH, and it diverts what `getForeignKeyValue()` returns **before** the parent's forward FK is assigned at `V2.php:5047`. Net effect: a `PUT {primaryContactPhoneNumber: {uuid}}` does **not** set `Contacts.primaryContactPhoneNumberId`, and the `PUT` response **echoes `primaryContactPhoneNumber: null`** on the miss.
-- **Scope.** Any **forward singular FK** whose child model has a back-FK to the parent is exposed on update — `primaryContactEmailAddress` / `primaryContactAddress` on `Contacts` are likely affected the same way. Root cause of the **BDR web-funnel "phone-less contact is never dialed"** bug: see [web-funnel-app.md](../../ai-bdr/features/web-funnel-app.md) and [vapi-webhook-handler.md](../../worker2/features/vapi-webhook-handler.md).
+- **Scope.** Any **forward singular FK** whose child model has a back-FK to the parent is exposed on update — `primaryContactEmailAddress` / `primaryContactAddress` on `Contacts` are likely affected the same way. Root cause of the **BDR web-funnel "phone-less contact is never dialed"** bug: see [web-funnel-app.md](../../../../standalone/apps/bdr/features/web-funnel-app.md) and [vapi-webhook-handler.md](../../worker2/features/vapi-webhook-handler.md).
 - **Suggested platform fix (deferred to its own PR/ticket):** gate the `4985–4996` injection to **true reverse / has-many** relations only, so a forward singular-FK child keeps its single-key forced-MATCH; confirm `RecordFields.childPolicy` for `Contacts.primaryContactPhoneNumberId` when fixing. Until then the caller-side workaround is a follow-up identifier-only `PUT` re-link with a self-heal that detects the null/mismatched echo (shipped on the BDR funnel side — see web-funnel-app.md).
 
 **⚠ CREATE path: a nested singular FK is matched TENANT-WIDE by VALUE and RE-PARENTED.** The forward-singular-FK back-reference machinery has a second, worse manifestation — on the **CREATE** path, and the damage is a **stolen child row** rather than a null pointer. A nested `primaryContactEmailAddress` / `primaryContactPhoneNumber` object is matched **across the whole tenant by its VALUE** (`emailAddress` / `phoneNumber`) — **not** by `uuid` — and the matched child's back-reference `contactId` is then **flipped to the newest writer**, silently detaching it from its previous owner. **Measured on beta `Client_Aig` 2026-08-17 (TRUE-80562).** Two sequential POSTs each created a new Contact (**854** at 10:19, **856** at 10:41). Afterwards: Contacts **854 AND 856** both have `primaryContactEmailAddressId = 250` and `primaryContactPhoneNumberId = 18`; `ContactEmailAddresses` row **250** has `contactId = 856`; so contact **854 still points at a child row that now belongs to contact 856** — 854's primary email is effectively **detached**. **Singular and collection paths DIVERGE — only the forward singular FK re-parents.** The collection members were **not** stolen: the nested `contactEmailAddresses` entries produced rows **251/252** owned by 854 and **separate new rows 255/256** owned by 856. Same payload, same request — different behavior per path.
@@ -212,5 +212,5 @@ behaviour — it bites any write to a nested child, not one feature. Worked exam
 - [assetType from NetSuite itemtype](../../../1.0/apps/library/features/netsuite-item-assettype-sync.md)
 - [Per-currency amount lines editor](../../toga2-supply/features/currency-amount-lines-editor.md)
 - [cXML ShipNotice Gateway](cxml-shipnotice-gateway.md)
-- [BDR web-funnel app](../../ai-bdr/features/web-funnel-app.md) / [VAPI webhook handler](../../worker2/features/vapi-webhook-handler.md)
+- [BDR web-funnel app](../../../../standalone/apps/bdr/features/web-funnel-app.md) / [VAPI webhook handler](../../worker2/features/vapi-webhook-handler.md)
 - [Client schema-drift audit](../../dbchanges2/workflows/client-schema-drift-audit.md)

package/knowledge/2.0/apps/toga-blox/features/talos-assistant.md

@@ -27,8 +27,8 @@ related:
   - ../workflows/dynamic-publish-pipeline.md
   - ../workflows/local-link-into-a-consumer-app.md
   - ../../toga25-supply/features/talos-integration.md
-  - ../../ai-bdr/features/landing-chat-drawer.md
-  - ../../talos/features/chat-frontend.md
+  - ../../../../standalone/apps/bdr/features/landing-chat-drawer.md
+  - ../../../../standalone/apps/talos/features/chat-frontend.md
 ---
 
 Shared pure-UI AI-assistant (launcher + docked/drawer chat panel, adapter-injected, fully tokenized); open for its adapter/controlled contracts, streaming limits, branch divergence, and the per-role message-field gotcha.
@@ -109,7 +109,7 @@ onSend: (text: string) => Promise<TalosResponse>   // resolves ONCE, with the fi
 
 `TalosMessageModel` / `TalosConversation` are documented "in-memory only". There is **no token delta, no `threadId`, no tool-call surface, and no interrupt representation** anywhere in `types.ts`.
 
-**So token streaming is NOT an adapter-implementation detail.** No adapter, however clever, can stream through a `Promise<TalosResponse>` — the promise resolves once. Wiring the live Aegra/LangGraph backend (which streams tokens, emits tool calls, surfaces HITL interrupts — see [Talos chat frontend](../../talos/features/chat-frontend.md)) means changing this component's props. That was accepted deliberately rather than shipping a non-streaming v1.
+**So token streaming is NOT an adapter-implementation detail.** No adapter, however clever, can stream through a `Promise<TalosResponse>` — the promise resolves once. Wiring the live Aegra/LangGraph backend (which streams tokens, emits tool calls, surfaces HITL interrupts — see [Talos chat frontend](../../../../standalone/apps/talos/features/chat-frontend.md)) means changing this component's props. That was accepted deliberately rather than shipping a non-streaming v1.
 
 ### Controlled mode — BUILT 2026-09-01 (the TRUE-80692 contract change)
 
@@ -185,7 +185,7 @@ So the launcher was **already painted under the backdrop and its clicks were alr
 - **A `renderMessages`-style slot can silently delete panel-synthesized UI.** The loading bubble is not in `messages`; it was lost exactly this way. Check what the panel mixes into a list before replacing that list.
 - **`onSend` cannot stream** — use **controlled mode** (above). A live streaming backend needs the controlled props, not a cleverer adapter.
 - **Tool calls and HITL interrupts still have no representation** in `TalosMessageModel`. A host mapping from LangGraph must drop them. Surfacing them needs new blox props.
-- The real chat backend is a later blox ticket — `onSend` is stubbed here; do not assume live responses from blox alone. **A working implementation exists outside blox and is now wired to this component:** `bdr/src/lib/talosChat.ts` implements the same `onSend(text) => Promise<TalosResponse>` seam against the live Talos campaign-chat backend (session token, fingerprint, strict message shaping, offline fallback) and plugs straight in — see [AI-BDR landing chat drawer](../../ai-bdr/features/landing-chat-drawer.md). **The seam held: no bridge/translation layer was needed.** Consider upstreaming it when the adapter ticket is picked up; the only env-ism to change is its Next-specific `process.env.NEXT_PUBLIC_TALOS_API_BASE` default, which becomes a required `baseUrl` option for Vite consumers.
+- The real chat backend is a later blox ticket — `onSend` is stubbed here; do not assume live responses from blox alone. **A working implementation exists outside blox and is now wired to this component:** `bdr/src/lib/talosChat.ts` implements the same `onSend(text) => Promise<TalosResponse>` seam against the live Talos campaign-chat backend (session token, fingerprint, strict message shaping, offline fallback) and plugs straight in — see [AI-BDR landing chat drawer](../../../../standalone/apps/bdr/features/landing-chat-drawer.md). **The seam held: no bridge/translation layer was needed.** Consider upstreaming it when the adapter ticket is picked up; the only env-ism to change is its Next-specific `process.env.NEXT_PUBLIC_TALOS_API_BASE` default, which becomes a required `baseUrl` option for Vite consumers.
 - **BDR consumes this component today, but only through a local link.** The React `^18` peer cap is not a hard blocker in practice — BDR (React 19.2.4) runs it via a junction-linked checkout with a single React (procedure: [Local-linking blox into a consumer app](../workflows/local-link-into-a-consumer-app.md); framer-motion 11 and Pro FontAwesome behave fine on React 19). **What is still blocked is deployment:** BDR cannot ship until this branch merges and publishes and BDR consumes the npm version.
 - **Hosts can import the Talos subpath to avoid the barrel's CSS.** `@agilant/toga-blox/dist/components/Talos/index.js` resolves (package `exports` allows `./dist/*`); importing the package barrel pulls **every** blox component's CSS into the consumer bundle.
 - **⚠ `dist/` must be REBUILT for a blox change to reach any browser — and a green consumer typecheck does NOT prove it did.** The consumer's `tsc` resolves blox **types from source** through the `node_modules` symlink, but Vite loads the package at **runtime from `dist/`**. So `tsc --noEmit` can be perfectly clean while the browser runs stale blox code. Every blox change needs `npm run build` in the blox repo **plus** `rm -rf node_modules/.vite` in the consumer. See [Local-linking blox into a consumer app §4](../workflows/local-link-into-a-consumer-app.md). **Correction:** an earlier note claimed blox `npm run build` fails on a missing `tsconfig.app.json` — it does not; it builds fine.
@@ -198,4 +198,4 @@ So the launcher was **already painted under the backdrop and its clicks were alr
 
 - [TableRecordModal](table-record-modal.md) — sets the `data-blox-modal-open` marker the launcher dims on.
 - [Dynamic npm Publish Pipeline](../workflows/dynamic-publish-pipeline.md) and [Local-linking blox](../workflows/local-link-into-a-consumer-app.md) — how consumers get Talos.
-- [toga25-supply Talos integration](../../toga25-supply/features/talos-integration.md), [AI-BDR landing chat drawer](../../ai-bdr/features/landing-chat-drawer.md), [Talos chat frontend](../../talos/features/chat-frontend.md).
+- [toga25-supply Talos integration](../../toga25-supply/features/talos-integration.md), [AI-BDR landing chat drawer](../../../../standalone/apps/bdr/features/landing-chat-drawer.md), [Talos chat frontend](../../../../standalone/apps/talos/features/chat-frontend.md).

package/knowledge/2.0/apps/toga-blox/workflows/local-link-into-a-consumer-app.md

@@ -18,7 +18,7 @@ related:
   - ./dynamic-publish-pipeline.md
   - ../features/talos-assistant.md
   - ../architecture.md
-  - ../../ai-bdr/features/landing-chat-drawer.md
+  - ../../../../standalone/apps/bdr/features/landing-chat-drawer.md
   - ../../toga25-supply/features/talos-integration.md
 ---
 
@@ -26,7 +26,7 @@ How to develop an unpublished blox against a real consumer on Windows (incl. a R
 
 ## Summary
 
-How to develop a **blox component that is not published yet** against a real consuming app, on Windows, including the hard case: a **React 19 / Next.js (Turbopack)** host consuming a blox whose peers cap React at `^18`. Proven 2026-08-27 wiring blox `Talos` into `ai-bdr` ([landing-chat-drawer.md](../../ai-bdr/features/landing-chat-drawer.md)).
+How to develop a **blox component that is not published yet** against a real consuming app, on Windows, including the hard case: a **React 19 / Next.js (Turbopack)** host consuming a blox whose peers cap React at `^18`. Proven 2026-08-27 wiring blox `Talos` into `ai-bdr` ([landing-chat-drawer.md](../../../../standalone/apps/bdr/features/landing-chat-drawer.md)).
 
 Use it for the inner loop only. **It is not a deploy path** — see [Dynamic npm Publish Pipeline](./dynamic-publish-pipeline.md) for how the consumer eventually gets the code.
 
@@ -204,4 +204,4 @@ So `tsc` can be completely clean while the browser runs **stale** blox code —
 - [dynamic-publish-pipeline](./dynamic-publish-pipeline.md) — how the consumer gets blox for real (not this dev loop).
 - [Talos assistant](../features/talos-assistant.md) — the component this loop was proven on.
 - [architecture](../architecture.md) — library overview.
-- [AI-BDR landing chat drawer](../../ai-bdr/features/landing-chat-drawer.md), [toga25-supply Talos integration](../../toga25-supply/features/talos-integration.md).
+- [AI-BDR landing chat drawer](../../../../standalone/apps/bdr/features/landing-chat-drawer.md), [toga25-supply Talos integration](../../toga25-supply/features/talos-integration.md).

package/knowledge/2.0/apps/toga25-supply/INDEX.md

@@ -17,7 +17,7 @@
 | [Side navigation & default route — an empty nav renders a BLANK PAGE and gets reported as "cannot log in"](features/side-navigation-and-default-route.md) | 2.5 nav/routes/landing are 100% backend-driven by the `navigation` surface; a user with zero visible nav elements gets a BLANK page reported as "can't log in" — |
 | [SSO redirect & public-vs-user session gating (useAuthenticationFlow)](features/sso-redirect-and-session-gating.md) | How 2.5 decides on every navigation whether to bounce an anonymous visitor to their client's SSO IdP vs the local `/login` (useAuthenticationFlow); open for SSO |
 | [Surface Frontend (DB-driven UI consumption, src/surface/)](features/surface-frontend.md) | The FE consumer of the platform-wide Surface layer (DB-driven UI config via `GET /v2/surfaces/meta`) — hooks, registries, adapter seams, migration recipe, and t |
-| [Talos Integration (AppLayout host, live Aegra streaming, LangGraph→blox mapper)](features/talos-integration.md) | toga25-supply hosts the shared blox Talos assistant — AppLayout host, live Aegra/LangGraph streaming in a supply-owned provider, surface-gated per tenant (NYCHH |
+| [Talos Integration in TOGa Supply](features/talos-integration.md) | TOGa Supply embeds a surface-gated blox Talos panel using the current Supply user's bearer; open for assistant discovery, streaming, frontend lifecycle gaps or |
 | [Toolbar button chrome (ToolbarButton) — and why blox BaseButton can't be resized from the theme](features/toolbar-button-chrome.md) | toga25-supply's own ToolbarButton chrome (pill/chip, both pinned 27px) for page-toolbar buttons, because blox `BaseButton` sizes are un-themeable literals; open |
 | [Transfer Orders page (TableView → Core surfaces → React page + record modal)](features/transfer-orders-page.md) | The Transfer Orders vertical — list + record modal + Create modal, built surface-first from nothing (NYCHH-only today); the cleanest worked example of the curre |
 | [AWS Amplify Multi-Environment Deployment](workflows/amplify-deployment.md) | How toga25-supply deploys to all its Amplify environments from one shared `amplify.yml` — branch `_<mode>` → Vite `--mode`; open when adding an environment or d |

package/knowledge/2.0/apps/toga25-supply/features/talos-integration.md

@@ -1,187 +1,85 @@
 ---
-title: Talos Integration (AppLayout host, live Aegra streaming, LangGraph→blox mapper)
+title: Talos Integration in TOGa Supply
 framework: "2.0"
 repo: toga25-supply
 project: TOGa 2.5 Supply
 client: shared
 type: feature
 status: active
-updated: 2026-09-16
-owners: [apeterson]
+updated: 2026-10-08
+owners: [apeterson, akhokhani]
 files:
   - toga25-supply/src/api/talos.ts
   - toga25-supply/src/hooks/useTalosSession.ts
   - toga25-supply/src/hooks/useTalosThreads.ts
   - toga25-supply/src/hooks/useTalosSurface.ts
-  - toga25-supply/src/App.tsx
   - toga25-supply/src/providers/TalosStreamProvider.tsx
   - toga25-supply/src/utils/talosMessages.ts
-  - toga25-supply/src/utils/talosMessages.test.ts
   - toga25-supply/src/layout/AppLayout/AppLayout.tsx
-  - toga25-supply/src/components/Header/Header.tsx
-  - toga25-supply/src/index.css
+  - toga25-supply/.env.production
 related:
   - ../../toga-blox/features/talos-assistant.md
-  - ../../toga-blox/features/table-record-modal.md
-  - ./surface-frontend.md
   - ../../dbchanges2/features/surface-layer-schema.md
-  - ../../../clients/nychh/features/talos-assistant-enablement.md
-  - ../../toga-blox/workflows/local-link-into-a-consumer-app.md
-  - ../../talos/features/chat-frontend.md
-  - ../../talos/features/aegra-api.md
-  - ../../../standards/frontend.md
+  - ../../../../standalone/apps/talos-backend/workflows/toga-supply-client-onboarding.md
+  - ../../../../standalone/apps/talos/features/chat-frontend.md
+  - ../../../../clients/nychh/features/talos-assistant-enablement.md
+  - ../../../../clients/compass-usa/features/talos-supply-assistant.md
 ---
 
-toga25-supply hosts the shared blox Talos assistant — AppLayout host, live Aegra/LangGraph streaming in a supply-owned provider, surface-gated per tenant (NYCHH only today); open for Talos wiring, gating, or Aegra debugging.
+TOGa Supply embeds a surface-gated blox Talos panel using the current Supply user's bearer; open for assistant discovery, streaming, frontend lifecycle gaps or tenant enablement.
 
 ## Summary
 
-toga25-supply is the first host of the shared blox [Talos assistant](../../toga-blox/features/talos-assistant.md). `AppLayout` is the host: the launcher renders in the header, the panel at app root.
-
-**⚠ Corrected 2026-09-02 — Talos is NO LONGER enabled for all clients.** It is now gated by the **surface layer**, seeded **OFF for every client**, and opted in per tenant. **NYCHH is the only tenant granted it today.** The suggestion chips come from the same surface, not from blox.
-
-As of **2026-09-01 the live Aegra/LangGraph streaming backend is BUILT** (TRUE-80693). `onSend` is no longer a stub — the panel runs in blox's new **controlled mode** and the stream lives in a supply-owned provider. All LangChain code is **app-local**; blox stays pure UI.
+`AppLayout` hosts the launcher and controlled panel; one Supply-owned `TalosStreamProvider` wraps LangGraph SDK `useStream`. Blox is UI only. The shared `talos-assistant` Surface is seeded off and granted per tenant; a provisioned backend assistant does not establish that a customer's panel is enabled.
 
 ## How it works
 
-```
-src/api/talos.ts                        — sole VITE_TALOS_API read chokepoint + assistants lookup
-src/hooks/useTalosSurface.ts            — DB feature gate + suggestion chips (surface bundle)
-src/hooks/useTalosSession.ts            — resolves token + assistantId → a `connection` state
-src/hooks/useTalosThreads.ts            — Aegra thread list → blox TalosThreadSummary[] + title polling
-src/providers/TalosStreamProvider.tsx   — the single useStream() call, exposed via own context
-src/utils/talosMessages.ts              — LangGraph Message[] → blox TalosMessageModel[]
-src/layout/AppLayout/AppLayout.tsx      — host: provider wrapper + controlled panel props
-```
-
-### `src/api/talos.ts` — the env chokepoint
-
-The only place `VITE_TALOS_API` is read. Exposes `getTalosApiUrl()`, `isTalosConfigured()`, `fetchTalosAssistants()`, `pickTalosAssistant()`.
-
-`fetchTalosAssistants()` **logs the HTTP status before throwing** — `401` = token/issuer mismatch, `403` = tenant not provisioned. Without that log React Query reduces both to a silent boolean and you debug blind.
-
-**Do NOT send a body or a `Content-Type` on the `GET /assistants` call.** `fetch` forbids a body on GET, and a JSON content-type only adds a CORS preflight. (The curl example circulated internally used `--data '{}'`; do not copy it into the browser client.)
-
-### `assistantId` is PER-TENANT — fetch it, never pin it in env
-
-`GET {aegra}/assistants?limit=100&offset=0` with the bearer returns:
-
-```jsonc
-{ "assistants": [ { "assistant_id": "…", "name": "…", "graph_id": "…",
-    "metadata": { "customer_facing": true, "read_only": true, "tenant_slug": "…" } } ],
-  "total": 1 }
-```
-
-NYCHH returns exactly one — name `TOGa Supply`, `graph_id: deep_agent`, `read_only: true`, `customer_facing: true`.
-
-**Selection rule (`pickTalosAssistant`):** prefer `metadata.customer_facing === true`; else fall back to the full list; take the first; `console.warn` when there is more than one candidate. Never silently pick, never throw.
-
-**Design consequence:** `useStream` takes **both** the token and `assistantId` as *values*, so there are now **two** async prerequisites that must resolve above the hook. That is why the panel's `connection` prop is load-bearing rather than polish.
-
-### `useTalosSession()` — the gate
-
-Returns `connection: "disabled" | "connecting" | "ready" | "error"` plus the token and assistantId.
-- **Gate on `isAuthenticated` (the `"user"` key), NOT on the presence of an accessToken.** blox `fetchPublicToken` leaves a **public** token in storage for logged-out visitors, so "there is a token" is not "there is a user."
-- **Key the assistants query by USER UUID, never by the token.** Supply persists the React Query cache to `localStorage` (`supply-chain-query-cache`), so a token in a query key gets written to disk.
-- `staleTime: Infinity`, `gcTime: Infinity`, `retry: false`.
-- **Anti-pattern — do not gate with `useSuspenseQuery`/Suspense.** It suspends the subtree, so the panel **unmounts and remounts** — flashing and dropping whatever the user typed in the composer. Keep it a plain query and gate on the returned `connection`.
-
-### `TalosStreamProvider` — outer gate + inner session
-
-One `useStream` call from `@langchain/langgraph-sdk/react`, exposed through a supply-owned context. `threadId` lives in the URL via react-router `useSearchParams` (**never `nuqs`** — Next-only).
-
-It is deliberately **split in two**: an outer gate component and an inner `TalosStreamSession`. Hooks cannot be called conditionally, and the stream must not mount until the token *and* the assistantId are real values — so the gate renders the session only once both resolve.
+### Source, environment and request boundary
 
-Why a provider at all: `useStream` is stateful (two call sites = two conversations), it must sit **above the router outlet** so navigation does not kill an in-flight response, and it is the app↔blox boundary where LangGraph shapes are translated.
+Use the locally saved `origin/_production` source, not a stale working clone. The October 7 survey inspected Supply commit `bb35181`: React 18.3, Vite 7, LangGraph SDK `^1.10.0` and blox `1.1.4-production.196`. `.env.production` is tracked and sets `VITE_TALOS_API=https://api.togaiq.com`; non-production env files target `https://api.beta.togaiq.com`. There is a production backend.
 
-### `AppLayout` — split into `AppLayoutInner`
+`src/api/talos.ts` is the only env read chokepoint. Browser requests go directly to Aegra using the current Supply access token and real browser Origin; there is no token exchange or Next proxy. Aegra's TogaHub auth resolves the exact JWT client slug into the Talos control plane; `/v2/users/me` supplies persona-linked AI-model slug grants. The platform MCP receives that same user's bearer and trusted frontend routing context; api2 remains the business-record/field authorization authority.
 
-`AppLayout` now wraps `AppLayoutInner` in `TalosStreamProvider`, **because a component cannot consume a provider it renders**. The panel is driven by the controlled props (`messages` / `isStreaming` / `onSubmit` / `onStop` / `connection`); the adapter now carries only navigation side effects (`onOpenRecord` → `/sales-orders?sales-orders=<id>`, `onAction("approvals")` → `/sales-orders`, `onDockWidthChange` → reserve layout width).
+`useTalosSession` requires `isAuthenticated`, not merely a token (logged-out public tokens exist). Assistant query keys use user UUID, never the token: React Query cache is persisted locally. Query state produces disabled/connecting/ready/error; keep it a non-Suspense query so reconnect does not unmount the composer.
 
-**Maps the app-only `"disabled"` state to `"error"`** when passing `connection` down — blox has no `"disabled"` value.
+`GET /assistants?limit=100&offset=0` discovers the client assistant. Prefer `metadata.customer_facing===true`, otherwise use the list, choose first, and warn on multiple candidates. Never pin an assistant ID in env. GET carries no body or unnecessary JSON Content-Type. HTTP status logs distinguish rejected identity/tenant/entitlement from absent configuration.
 
-### `useTalosSurface()` — the DB feature gate (2026-09-02)
+### Surface and provider lifecycle
 
-Reads the Core surface `talos-assistant` (type `DRAWER`, not record-bound) and returns **whether Talos is on for this tenant** plus the suggestion chips. **TWO elements** in the bundle:
+`useTalosSurface` finds a `featureEnabled` element in the `talos-assistant` bundle; missing/loading/failed/no-marker means disabled. Its `suggestion` element owns the whole ordered chip row. `config.items` has label/prompt/icon; prompt falls back to label. Invalid items are skipped, while a visible empty row is permitted.
 
-| Element | Meaning |
-|---|---|
-| `config.role = 'featureEnabled'`, `config.key = 'feature'` | **master switch** — its `isVisible` turns the launcher, the panel and the Aegra stream on or off |
-| `config.role = 'suggestion'`, `config.key = 'suggestions'` | **the whole chip row in ONE element.** `config.items` is an ordered array of `{label, prompt, icon}` (`icon` = a literal FontAwesome name). Its `isVisible` switches the entire row. |
+The outer provider gate waits for enabled Surface, authenticated token and resolved assistant. The inner session owns the single `useStream` above the router outlet, so navigation does not recreate the conversation. Thread ID is URL state via React Router `useSearchParams`, not Next-only nuqs. `talosMessages` maps LangGraph to blox models; tool-call/HITL payloads are not rendered as general user chat.
 
-**⚠ SUPERSEDED: chips were briefly ONE ELEMENT PER CHIP** — four `config.role='suggestion'` elements with `labelMessageId` / `config.prompt` / `iconToken` / `sortOrder` and keys `pendingApproval`, `blockedApprovals`, `pendingFulfillment`, `statusCounts`. **That shape is gone.** There is no `config.prompt` and no per-chip `sortOrder`; **order is array order.** Reason: per-chip elements meant a tenant could only switch on wording **Core** had already seeded, so rephrasing a tenant's own business question needed a Core migration. One element lets a tenant author its own set in a single `CONFIG` override. Costs (labels not translatable; whole sets scope, not chips): [surface-layer-schema](../../dbchanges2/features/surface-layer-schema.md).
+The inspected implementation shows final-answer outcomes and suppresses final-answer content after failed/cancelled runs. It supports thread list/title polling and metadata writes with `If-Match`. That code alone does not satisfy all backend recovery requirements.
 
-**The FE FAILS CLOSED — deliberately, not defensively.** Missing bundle, failed fetch, absent marker element, or *still loading* all resolve to **disabled**. Talos answers questions from a tenant's live supply-chain data through an external AI service, so an unresolved gate must never read as "on".
+### Backend contract versus inspected gaps
 
-`TalosStreamProvider` gained an **`enabled`** prop off the back of this: a tenant without Talos never even calls `GET /assistants`.
+The intended embedded-chat contract requires explicit server cancellation, persisted `/transcript` reload, run-ID rejoin with replay cursor, meaningful terminal-state handling, feedback and refreshed bearer retry. In the inspected production provider, Stop calls only `stream.stop()`; the dedicated cancellation helper is not wired. The saved Compass handoff also identified missing persisted transcript reload and run-ID rejoin. Treat these as release gaps, not finished behavior.
 
-#### `readItems` is TOLERANT on purpose — one bad item costs one chip
+The backend assistant is `toga-supply` on `deep_agent`, with user-bearer `toga-platform` tools and tenant-specific persona/BLP. Compass and NYCHH backend deployments existed in the dated survey; neither tenant's current production Surface grants were read from Core/Client DB during that survey. See their client pages for exact evidence.
 
-`config.items` is a **hand-edited JSON column that differs per tenant**, so `readItems` **skips a malformed item instead of throwing**. One bad entry in a tenant's `CONFIG` override must cost them **one chip** — not every chip, and not a render crash. **14 tests** cover: `null`, a bare string, a number, a missing `label`, a whitespace-only `label`, a non-string `icon`, `items` absent, `items` not an array, and **array order being the authored order**.
+### Hosting evidence
 
-🚨 **The gate the FE cannot defend against:** a tenant `CONFIG` override that omits `"role":"suggestion"` **wipes** it (a `CONFIG` override replaces `config` wholesale — verified at `_underscore Model/Core/Surface.php:839`, no merge). The element then no longer matches, the chips **silently disappear**, and there is no error anywhere. That is a seeding rule, not a parser fix. Seeding side, the marker-element choice, and the prod-cluster procedure: [surface-layer-schema](../../dbchanges2/features/surface-layer-schema.md).
+As of 2026-10-07, production Supply used Amplify app `d3v9s1i7kge4ll` and its CloudFront distribution. `*.beta.togasupply.com` used a different Amplify app/repo, `toga2-supply`, not this production 2.5 source. Confirm hostname→build mapping before reproducing a version-specific failure.
 
-### `useTalosThreads()` — the history drawer's data (2026-09-02)
-
-Maps Aegra `AegraThread` rows → blox `TalosThreadSummary[]` (`{ id, title, updated }`); the provider exposes `threads` / `threadsStatus` / `activeThreadId` / `selectThread` and `AppLayout` passes them straight into the panel's data-driven drawer props.
-
-**Refresh fires on two edges, and both are needed:**
-1. **On `threadId` assignment.** The server creates the thread at the **START** of the first run, so without this the *currently active* chat is missing from its own drawer.
-2. **On the falling edge of a run** (streaming → idle).
-
-**Opening an existing thread needs NO transcript parser.** `@langchain/langgraph-sdk`'s **`setThreadId`** (`dist/ui/orchestrator.js`, `#fetchHistoryForThread`) **clears the stream and fetches that thread's history itself** when `fetchStateHistory: true`. So selecting a history row is just "set the threadId", and the existing LangGraph→blox mapper handles the rest. This took `fromTalosTranscript()` **off the critical path for history**. It is still needed for the durable message ids that feedback requires — do not delete it.
-
-**Auto-title polling — `titlePollInterval`.** Aegra names a thread **asynchronously after the first run** and emits **no SSE event**, so the only way to see a title is to ask again. `titlePollInterval` feeds React Query's `refetchInterval`: **4s while any RECENT thread is unnamed, `false` otherwise.** Two deliberate deviations from the handoff doc §8.10:
-- The doc says poll *"while the first run is active"*. **The title lands AFTER the run**, so polling only during it usually misses it entirely.
-- Added a **2-minute age window** the doc does not specify. A thread whose first message was never text-bearing **never gets named at all**, so "stop when the title arrives" on its own polls every 4s for the whole session.
-
-7 tests cover the stop condition, including a server clock running ahead of the client.
-
-> Server-side, auto-naming is currently **off or broken** for NYCHH's assistant, so this polling has nothing to find yet — see [aegra-api](../../talos/features/aegra-api.md). Supply reads `metadata.thread_name` correctly and cannot fix it.
-
-### `talosMessages.ts` — the LangGraph → blox mapper
-
-- Handles LangGraph's **string-OR-content-blocks** union for `content`.
-- Filters to `human` / `ai` only — `system`, `tool` and `remove` are dropped.
-- An **empty AI turn becomes `{ thinking: true }`** (the stream opening). An empty human turn is dropped.
-- Streamed text goes to **`message`** (light markdown), *not* `paras`: during a stream the text is partial, and splitting it into finished paragraphs reflows the bubble on every token.
-- **Deliberate loss:** tool calls and HITL interrupts have **no representation** in `TalosMessageModel` and are dropped. Surfacing them needs new blox props — not a mapper change.
-
-13 unit tests in `src/utils/talosMessages.test.ts`.
-
-### The `talos` query namespace is EXCLUDED from the persisted cache (2026-09-02)
-
-`App.tsx`'s `PersistQueryClientProvider` used to persist **every** query to `localStorage`. A `shouldDehydrateQuery` filter now excludes the **`talos`** namespace alongside `surface-meta` / `surface-meta-group`: chat titles and assistant ids are **user/session data that must not outlive the session on disk**. The surface half of the same fix is what makes a Talos grant (or a **revoke**) actually reach a running browser — see [surface-frontend](./surface-frontend.md).
-
-### Environment config
-
-`VITE_TALOS_API` is set to `https://api.beta.togaiq.com` in **7 non-prod env files**. It is left **commented out in `.env` (production)** because no live production Aegra host is confirmed — `isTalosConfigured()` therefore returns false there and `connection` resolves to `"disabled"`.
-
-### Package choice: `@langchain/langgraph-sdk/react`, app-local
-
-- **`useStream` from `@langchain/langgraph-sdk/react`** — the same import talos uses — **not `@langchain/react`.** The decision stands: the path is proven in-house so `talos/src/providers/Stream.tsx` lifts nearly verbatim, and `@langchain/react`'s extra surface is largely **media hooks/players** blox has no components to render.
-- **⚠ Correction (verified 2026-09-01): the old "one fewer peer" reason was WRONG.** `@langchain/langgraph-sdk@1.10.0` declares `@langchain/core: ^1.1.48` as a **required (non-optional) peer** itself; `react`/`react-dom` are the optional ones. Symptom when it is missing: Vite/esbuild dep-optimize **hard-fails** with `Could not resolve "@langchain/core/messages"` and the dev server exits. Install `@langchain/core` alongside the SDK.
-- **Take only `useStream` for v1.** Add `useToolCalls` when tool activity is actually shown, and the headless interrupt helpers for human-in-the-loop — that is when `@langchain/react` earns its keep.
-- **It stays in supply, never in blox.** Per [frontend.md §21](../../../standards/frontend.md) this is n=1, and per §13(b) a blox peer would force `@langchain/core` on all five consumer apps.
+## Gotchas
 
-**Still open:** mid-stream 401 behavior — silent refresh and resume vs. surface an error and let the user resend. Decide deliberately rather than discovering it in QA.
+- Surface visibility, active `/users/me` slug grant, control-plane tenant, assistant capability assignments and MCP data ACL are independent gates. Do not repair a denied user with an admin/service token.
+- A backend assistant with successful historical runs is not proof of a production panel launch. Read the actual Surface bundle as the intended user.
+- SDK Stop closes browser streaming; it is not a verified server-run cancellation in the inspected implementation.
+- The mapper hides tools/HITL. Keep this customer-facing read-only assistant within its declared boundary; richer decision UI requires deliberate contract work.
+- Stable query keys must not include credentials. Caches and logs can persist long after a token changes.
+- Code comments calling Aegra “TOGa IQ” are legacy naming. Talos is the backend agent platform; TOGa IQ is the separate management UI.
 
-### Aegra-side status (was "three blockers")
+## Related
 
-- **(a) CORS — NARROWED, still open.** The old worry that Aegra's `Domains` gate needs a spoofed `Origin` (talos pins `Origin: http://talos` server-side, and `Origin` is a forbidden header) is **moot for supply**: supply's real browser origin *is* the registered hostname, so the browser sets it correctly for free. What **is** still a blocker: **Aegra returns CORS headers on the OPTIONS preflight only, not on the actual response** — verified 2026-09-01 from both `http://nychh.dev.sandbox.togasupply:5173` and `https://nychh.togasupply.com` (`access-control-allow-origin: *` on the preflight, **no** `access-control-*` header on the real GET). The browser will block every real response until this is fixed. Reported to `akhokhani`.
-- **(b) Tenant provisioning is explicit, no auto-create.** Unchanged — an unregistered org gets a clean **403 on every call even with a valid token**.
-- **(c) Token audience — RESOLVED, and it is the good outcome.** Aegra `AUTH_TYPE=togahub` **accepts a supply-issued api2 JWT.** Proven by a working `GET /assistants` whose bearer carried `aud` = a **supply** hostname (`https://nychh.togasupply.com`), `iss`/`sub` = `http://api.beta.togahub.com`, `id.client.slug: Nychh`, 3600s life. Answer from `akhokhani`. **Consequence: no token exchange, no `/auth/delegator` handoff, no auth-exchange layer.** Supply passes the token it already has straight into `defaultHeaders`.
+- [talos assistant](../../toga-blox/features/talos-assistant.md)
+- [surface layer schema](../../dbchanges2/features/surface-layer-schema.md)
+- [toga supply client onboarding](../../../../standalone/apps/talos-backend/workflows/toga-supply-client-onboarding.md)
+- [chat frontend](../../../../standalone/apps/talos/features/chat-frontend.md)
+- [talos assistant enablement](../../../../clients/nychh/features/talos-assistant-enablement.md)
+- [talos supply assistant](../../../../clients/compass-usa/features/talos-supply-assistant.md)
 
-## Gotchas
+## Change history
 
-- **React Query sets `isError` on a FAILED BACKGROUND REFETCH while KEEPING the data.** This printed **"Couldn't load your chats." over a full, usable chat list.** The status line was `isError ? "error" : …`, so **one** failed 4s title poll flipped a good list into an error state. **Fix:** a pure `threadsStatusFrom` helper that returns `"error"` **only when there is no data to show**. Pair it with the blox-side fix (drawer notes gated on an empty row list) — both halves were needed. Any hook with a `refetchInterval` inherits this trap: `isError` is not "the data is unusable", it is "the last fetch failed".
-- **user turns use `text`, Talos turns use `message`/`paras`.** blox `TalosMessage` renders **`message.text` for a `role: "user"` turn** but **`message.message` / `message.paras` for a `role: "talos"` turn**. The mapper originally put user turns in `message`, which renders an **empty bubble**. Caught only by writing the tests.
-  ```ts
-  // CORRECT
-  { role: "user",  text }       // user turn
-  { role: "talos", message }    // assistant turn
-  ```
-  Anyone writing to `TalosMessageModel` by hand will hit this.
-- **The launcher greys out under a modal via a ROOT ATTRIBUTE, and keyboard access is still open.** blox `TableRecordModal` sets **`data-blox-modal-open` on `document.documentElement`** while open, and blox dims the launcher under `:root[data-blox-modal-open]`. It applies to **every** `TableRecordModal` in supply automatically — do not add a per-page dim. Why an attribute and not a prop: the launcher renders in the **header**, outside the modal's subtree, so no shared React state reaches it (same boundary-crossing trick as `--talos-dock-width`). Mechanism + the counter it needs: [TableRecordModal](../../toga-blox/features/table-record-modal.md). 🚨 **Known gap, unfixed:** the launcher is still **keyboard-reachable** behind a modal. `pointer-events: none` stops the mouse only. Supply's header receives `inert` for the **side nav** but **not** for modals — closing this is a **supply** change that affects every header control, so it was flagged rather than patched.
-- **Reserve dock width on the OUTER wrapper, not just `<main>`.** The docked panel is `position: fixed`, full-height on the right. The blox `Header` is in **normal document flow** (NOT fixed). Reserving the dock width only on `<main>` leaves the header's right edge sitting **under** the panel, hiding the launcher + avatar. **Fix:** apply the reserved dock width (`marginRight` = the `onDockWidthChange` value) to the **outer app wrapper**.
-- **`npm install` silently un-links blox and Talos disappears.** Any `npm install` in this repo replaces the `node_modules/@agilant/toga-blox` **symlink** with the registry version, which contains no Talos. The symptom looks nothing like a dependency problem. Full symptom + relink recipe: [Local-linking blox into a consumer app §1a](../../toga-blox/workflows/local-link-into-a-consumer-app.md).
+- 2026-10-08 — Reconciled local source and dated 2026-10-07 research into the canonical KB (akhokhani).

package/knowledge/2.0/apps/worker2/INDEX.md

@@ -60,6 +60,7 @@
 | [Stranded WorkerJobs (deploy-window job loss) and how to repair them](features/stranded-worker-jobs.md) | Jobs queued while worker2 restarts are dropped silently with no retry and no alert; open when a cron "never ran" or a webhook vanished. |
 | [Talos Meeting-Notes Integration & Token Auto-Refresh (consumer)](features/talos-meeting-notes-integration.md) | Consumer side of Talos: the endpoints, DevCore assistant, and JWT auto-refresh a worker2 dev tool needs to query the meeting-notes corpus; open when a tool must |
 | [Talos Pricing Automation (worker2 Cron — Usage Import, AWS Actuals, Margins, Profiles)](features/talos-pricing-automation.md) | Monthly worker2 crons that turn raw usage + AWS cost into a measured platform unit price, per-client cost, and margins; open for the pricing pipeline, its AWS c |
+| [Pricing & COGS Model (Talos Pricing Calculator)](features/talos-pricing-cogs-model.md) | Cost basis + pricing methodology for selling Talos chat and voice; open to re-price Talos or understand COGS drivers, token rates, and the band/org-fee model. |
 | [Talos Transcript Ingestion Pipeline (worker2 → AWS Bedrock KBs)](features/talos-transcript-ingestion.md) | `_Worker_Team_Transcripts` cron pipeline ingesting Teams VTT transcripts into Talos Bedrock KBs + emailing organizer recaps; open for Export/Process/Retry/SyncK |
 | [Team Sprint Management & Reporting](features/team-sprint-management.md) | `_Worker_Team_Sprint` engine for TOGA's biweekly dev-sprint process + reporting (warnings, lock, daily dashboard, sprint-end scoring, AI release notes, 7 PM In |
 | [Teams Meeting Transcript Export](features/teams-transcript-export.md) | How `Team/Transcripts/Export` polls Microsoft Graph for Teams transcripts; open for the Graph organizer-resolution, GUID-vs-UPN, and Entra-permission details (t |

package/knowledge/2.0/apps/worker2/features/callback-scheduling.md

@@ -14,9 +14,9 @@ files:
   - worker2/Controller/Index.php
 related:
   - vapi-webhook-handler.md
-  - ../../ai-bdr/features/vapi-integration.md
-  - ../../ai-bdr/features/call-orchestration.md
-  - ../../ai-bdr/workflows/safe-call-loop-testing.md
+  - ../../../../standalone/apps/ai-bdr/features/vapi-integration.md
+  - ../../../../standalone/apps/ai-bdr/features/call-orchestration.md
+  - ../../../../standalone/apps/ai-bdr/workflows/safe-call-loop-testing.md
 ---
 
 The AI-BDR callback path (prospect says "call me back" → dialer redials) and its production defects; open when working on Vapi callback capture/parse/schedule or the callback root cause.
@@ -43,14 +43,14 @@ Callback path spans the webhook (`Worker/Vapi.php` — capture, persist, parse,
 
 ### End to end
 
-1. **Capture.** Requested time arrives at `message.artifact.structuredOutputs.<schema-uuid>.result.callbackTime` (`Worker/Vapi.php:99` → `:109–122` → `:144`) — **not** in `analysis`, which is always empty. The published schema documents ISO 8601 ([vapi-integration.md](../../ai-bdr/features/vapi-integration.md)), but in practice the field is **unconstrained free text**; the file's own sample payload contains `"callbackTime": "later"` (`Worker/Vapi.php:45`).
+1. **Capture.** Requested time arrives at `message.artifact.structuredOutputs.<schema-uuid>.result.callbackTime` (`Worker/Vapi.php:99` → `:109–122` → `:144`) — **not** in `analysis`, which is always empty. The published schema documents ISO 8601 ([vapi-integration.md](../../../../standalone/apps/ai-bdr/features/vapi-integration.md)), but in practice the field is **unconstrained free text**; the file's own sample payload contains `"callbackTime": "later"` (`Worker/Vapi.php:45`).
 2. **Persist (raw).** `updateContactAttempt` stores the string with **no conversion** (`:213–215`) alongside `c_callOutcome='CALLBACK_SCHEDULED'` / `c_actionToTake='SCHEDULE_CALLBACK'` (map at `:161–173`, applied `:189–194`), via `PUT /contact-attempts/{uuid}` (`:256–266`). `c_callbackTime` is `VARCHAR(255)` (dbchanges2 `_modules/Ai/Bdr/2026-03-04 bdrCustomfields.sql:84`), so a malformed value **survives persistence intact** and only dies at the act step.
 3. **Act.** `executePostCallActions` dispatches `SCHEDULE_CALLBACK` → `actionScheduleCallback`, which resolves the campaign timezone (`:400`), calls `self::convertDateTime($callbackTime, …)` (`:401`), and on success writes `Contacts.dtNextContactRequested` (`:414`) and `contactCallTypeId = CALLBACK` (`:417`) via `PUT /contacts/{uuid}` (`:420–430`).
 4. **Redial.** The dialer treats a due `dtNextContactRequested` as the privileged path: it **bypasses call windows, weekly cadence, call-window exception dates and prior DNC** (`Ai/Bdr/Vapi.php:238–303`, `:314–331`). Callbacks also sort first in the per-run queue (`:334`).
 
 ### Recommended fix direction (decided 2026-07-30 — priority order)
 
-1. **DURABLE FIX — Vapi assistant config side.** Constrain `callbackTime` in the structured-output schema to **strict ISO 8601 with offset**, and prompt the assistant to resolve relative speech to an absolute timestamp before emitting. ISO parses cleanly today, so this removes the whole class of failure. See [vapi-integration.md](../../ai-bdr/features/vapi-integration.md).
+1. **DURABLE FIX — Vapi assistant config side.** Constrain `callbackTime` in the structured-output schema to **strict ISO 8601 with offset**, and prompt the assistant to resolve relative speech to an absolute timestamp before emitting. ISO parses cleanly today, so this removes the whole class of failure. See [vapi-integration.md](../../../../standalone/apps/ai-bdr/features/vapi-integration.md).
 2. **`Worker/Vapi.php:71–79` and `:135`** — add `if (($message->type ?? '') !== 'end-of-call-report') return 'ok';` and stop force-stamping `dtEnded` from `NOW()` (set it only from a real `endedAt`). Independent of the callback bug and arguably **more urgent** — see the concurrency gotcha in [vapi-webhook-handler.md](vapi-webhook-handler.md).
 3. **`:609–635` and `:403–406`** — defensive normalization (strip leading `in `/`at `/`around `/`sometime `, map spelled-out numbers) and, on failure, **schedule a flagged sane default rather than `return`** — never silently drop a promised callback.
 4. **`:401–414`** — clamp the parsed time: if `<= now`, push to the next valid slot; snap into the campaign's call windows; reject beyond ~90 days.
@@ -174,6 +174,6 @@ A prospect who asks for a callback **on their 3rd attempt** gets `dtNextContactR
 ## Related
 
 - [vapi-webhook-handler](vapi-webhook-handler.md)
-- [vapi-integration](../../ai-bdr/features/vapi-integration.md)
-- [call-orchestration](../../ai-bdr/features/call-orchestration.md)
-- [safe-call-loop-testing](../../ai-bdr/workflows/safe-call-loop-testing.md)
+- [vapi-integration](../../../../standalone/apps/ai-bdr/features/vapi-integration.md)
+- [call-orchestration](../../../../standalone/apps/ai-bdr/features/call-orchestration.md)
+- [safe-call-loop-testing](../../../../standalone/apps/ai-bdr/workflows/safe-call-loop-testing.md)

package/knowledge/2.0/apps/worker2/features/clickup-task-fluffer-talos.md

@@ -20,7 +20,7 @@ related:
   - ./talos-transcript-ingestion.md
   - ./creating-worker-actions.md
   - ../architecture.md
-  - ../../talos/features/talos-agent.md
+  - ../../../../standalone/apps/talos-backend/features/talos-agent.md
 ---
 
 Async worker that "fluffs" a new ClickUp task by running its name+description through the Talos `dev-core` KB agent and writing the report back; open for the fluffer flow, the distinct Talos agent endpoint/key, or the open prod 403.
@@ -99,4 +99,4 @@ Both are separate from the existing `[talos]` `api_key`/`endpoint` (used by Tran
 - [Talos transcript ingestion](./talos-transcript-ingestion.md)
 - [Creating Worker Actions](./creating-worker-actions.md)
 - [Worker (worker2) Architecture](../architecture.md)
-- [Talos agent](../../talos/features/talos-agent.md)
+- [Talos agent](../../../../standalone/apps/talos-backend/features/talos-agent.md)

package/knowledge/2.0/apps/worker2/features/talos-meeting-notes-integration.md

@@ -11,7 +11,7 @@ owners: [dfranks]
 files:
   - .claude/skills/plan-ticket/scripts/talos.js
 related:
-  - ../../talos/architecture.md
+  - ../../../../standalone/apps/talos-backend/architecture.md
   - ./clickup-richtext-api.md
   - ../workflows/ticket-to-pseudocode-planning.md
 ---
@@ -20,7 +20,7 @@ Consumer side of Talos: the endpoints, DevCore assistant, and JWT auto-refresh a
 
 ## Summary
 
-How a **dev tool / agent consumes Talos** to query the team meeting-notes corpus programmatically. Talos itself (the LangGraph Agent Protocol server) is documented under [`2.0/apps/talos/`](../../talos/architecture.md); this doc is the **consumer side**.
+How a **dev tool / agent consumes Talos** to query the team meeting-notes corpus programmatically. Talos itself (the LangGraph Agent Protocol server) is documented under [`standalone/apps/talos-backend/architecture.md`](../../../../standalone/apps/talos-backend/architecture.md); this doc is the **consumer side**.
 
 **Key distinction:** the `[talos]` key in worker2 config only reaches the **stateless** `/api/ai/generate` and `/api/ai/chat` endpoints, which have **no meeting-notes access**. Meeting-notes access comes only from **assistants wired to knowledge bases**, reached through the Agent Protocol `/threads` + `/threads/{id}/runs/wait` surface with a **user Bearer JWT**.
 
@@ -81,6 +81,6 @@ The consumer tool stores its token pair at **`~/.talos/credentials.json`** (user
 
 ## Related
 
-- [Talos architecture](../../talos/architecture.md)
+- [Talos architecture](../../../../standalone/apps/talos-backend/architecture.md)
 - [ClickUp rich-text API](./clickup-richtext-api.md)
 - [Ticket-to-pseudocode planning workflow](../workflows/ticket-to-pseudocode-planning.md)

package/knowledge/2.0/apps/worker2/features/talos-pricing-automation.md

@@ -21,9 +21,9 @@ files:
 related:
   - ./creating-worker-actions.md
   - ../architecture.md
-  - ../../talos/architecture.md
-  - ../../talos/features/pricing-cogs-model.md
-  - ../../../1.0/apps/tools/features/talos-pricing-ui.md
+  - ../../../../standalone/apps/talos-backend/architecture.md
+  - talos-pricing-cogs-model.md
+  - ../../../../1.0/apps/tools/features/talos-pricing-ui.md
 ---
 
 Monthly worker2 crons that turn raw usage + AWS cost into a measured platform unit price, per-client cost, and margins; open for the pricing pipeline, its AWS cost query, or the `Talos*` Team-DB schema.
@@ -64,7 +64,7 @@ Untagged spend is attributed to the **defaultTenant**, and every row records `is
 
 Computes `clientMonthlyCost = client tokens × unitCostPer1kTokens`, margin vs. the client's band, the consecutive out-of-band streak, and emits a `TalosFeeRecommendations` row only once the streak ≥ the client's `consecutiveMonths`. **Internal tenants are excluded from fee recommendations** (see below). The old per-client "calibration factor" is gone.
 
-**Per-interaction clients (`pricingModel = PER_INTERACTION`, 2026-09-17):** revenue = `orgFee + interactions × perInteractionFee` (no bands); everything else — the streak, the org-fee recommendation — is identical. This **recurring** margin is the only input to the recommendation/streak. The **blended** margin (`blendedMarginPct`) additionally includes `amortizedFeeMonthly` (the amortized contract fees) and is written for **display only** — never fed to the streak. Design/rationale in [pricing-cogs-model](../../talos/features/pricing-cogs-model.md).
+**Per-interaction clients (`pricingModel = PER_INTERACTION`, 2026-09-17):** revenue = `orgFee + interactions × perInteractionFee` (no bands); everything else — the streak, the org-fee recommendation — is identical. This **recurring** margin is the only input to the recommendation/streak. The **blended** margin (`blendedMarginPct`) additionally includes `amortizedFeeMonthly` (the amortized contract fees) and is written for **display only** — never fed to the streak. Design/rationale in [pricing-cogs-model](talos-pricing-cogs-model.md).
 
 ### 4. RecomputeProfiles (09:00)
 
@@ -111,6 +111,6 @@ The existing `[talos]` config group gains **`usage_url`** and **`usage_api_key`*
 
 - [Creating worker actions](./creating-worker-actions.md)
 - [Worker (worker2) architecture](../architecture.md)
-- [Talos architecture](../../talos/architecture.md)
-- [Talos pricing COGS model](../../talos/features/pricing-cogs-model.md)
-- [Talos pricing UI (tools)](../../../1.0/apps/tools/features/talos-pricing-ui.md)
+- [Talos architecture](../../../../standalone/apps/talos-backend/architecture.md)
+- [Talos pricing COGS model](talos-pricing-cogs-model.md)
+- [Talos pricing UI (tools)](../../../../1.0/apps/tools/features/talos-pricing-ui.md)