diff --git a/docs/dataflow-coverage.md b/docs/dataflow-coverage.md
index 022de27c..901264d4 100644
--- a/docs/dataflow-coverage.md
+++ b/docs/dataflow-coverage.md
@@ -47,6 +47,18 @@ The work came from the data-flow experience audit (October 2026). The audit trac
- State: a zustand store (a `store` row) holds what its function returns and what its `set` receives; a selector reads the field it names, and calling a member read through the hook (`setCount` after `useStore((s) => s.setCount)`) runs the store's function, never the selector it is written beside. a setter's value reaches its getter's reads (#182); a function handed to a setter (`setCount((c) => c + step)`, a store path's `setState("todos", (list) => […])`) stands for what it returns, and its parameter is the current state; a Solid store path (`setStore("left", "title", v)`) writes `left.title` alone, so a sibling field never sees it, an updater there (inline or named) receives that field, not the whole store, and a predicate in the path (`(item) => item.id === id`) is a test, never stored; a store or signal's initial value seeds it field by field, so `createStore({ left, right })` keeps `left` and `right` apart; a state-pair row marked `lazy` (React's `useState`) calls a function initial value, and Solid's `createSignal` holds it; a `useReducer` dispatch binds the reducer's action, and the reducer's return is the state; state-pair factory and getter calls are framework-modelled (#188), modelled in the collector (#196).
- Ends: a binding every read of which decides a branch (`if (forgotten)` in nested branches, `ready && `, ``, through `!` or an accessor's call) ends at its binding, decides a branch. A parameter or local binding the program never reads (`_level`, `const { isSuccess } = useQuery()`) ends there, named "never read": the checker proves the value goes nowhere, so the trace is `ok`, not a dead end; the value-flow ledger draws every in-place end (a branch test, a write, never read) as a terminal row, as the graph does. A service write or read is a method on a client's contract, or a function-valued field a declaration file declares (a generated client's); a callback prop the project declares (`props.save(x)`) is never one. A thrown value ends at the throw (#207). An exported value that nothing in the program reads ends at its export, the module's API, not at a dead end; a call through a callback an export nothing in the program calls was handed (`onSeen(id)`, `props.onPick(x)` in a component nothing renders) ends there too, at the module's API, never a library method on a parameter (`items.forEach`); one the program reads, by name or through `import * as`, walks on to that read. A function handed to a timer (`setTimeout`, `requestAnimationFrame`) ends there: the callback runs later. A function handed to a page element as a handler (``) is not an escape: the page drops what it returns. A trace whose anchor has nothing beyond it names a stop at the anchor. A context `Provider`'s `value` goes to its readers, never into the page. A request ends a forward trace where it is sent, and a backward trace lands at the request a response came from: neither stops at the response body. A promise chain the code drops (`load().then(show);`) ends there. A function library code hands to a callback (`resolve` in `new Promise((resolve) => …)`) hands its argument back to the call that supplied it: the value reaches that call's value, and the call through it ends there, handed back. A DOM call the code makes for its effect (`el.classList.add(x);`, `target.dispatchEvent(e);`, `() => input?.focus()`) is a page write: any member TypeScript's DOM lib declares, whose value the code drops. A DOM call whose value the code uses (`root.querySelector(sel)`) carries on. An arrow whose body is an optional void call returns nothing. `response.json()` on a `fetch` result reads the body; only a server's response is an HTTP response. An emitter's `emit("name", x)` binds each listener its receiver registered with `on("name", fn)`, and `yield* list` hands on `list`'s elements. A return that is a call (`yield items.slice(…)`, an arrow's `() => f(x)`) is judged by its own function's call sites, never the call's, so an exported generator no code iterates ends at its return. A class's annotated uses fold by file (#208). A `when` on Solid's `Show` or `Match` ends as "decides a branch", not a DOM attribute.
+## Symbol view of an anchor trace
+
+The symbol view is the default drawing of an anchor trace. `?model=steps` (the Steps toggle in `FlowTools.tsx`) keeps the old one-box-per-step view.
+
+- One box per symbol. Boxes merge by scope and checker symbol id, which the anchor-graph contract carries as `symbol` and `field`, never by label. A node with no symbol keeps its own box. A merged edge keeps every original proof. On visual-notes' DocView this cut 557 boxes to 225.
+- The anchor-graph contract is the one seam: `engine/src/srcly/contracts/anchor-graph.ts` feeds `client/src/viz/deps/flow/anchorAdapter.ts`. Each fact the view needs (symbol, operation, `symbolKind`, frame `role`, links) is an additive field on it.
+- Value flow is the bold main line, with its operation named on the edge. Calls, reads and writes are thin links that can be toggled.
+- `symbolKind` (state, transform, origin, end, plain) carries the strong color. Frames stay neutral and name their role: component, hook, helper, handler, fetcher, server or unknown. Roles live in `engine/src/srcly/scope-roles.ts`, never in `engine/src/analysis`. Unknown is never drawn as a component.
+- The first screen is the main path from origins to ends. Side branches are stubs that grow with `+` and `−` or the depth control (1, 2, 3, all) in `depth.ts`. Depth is in the URL and the view-state registry.
+- Plain TypeScript modules with no routes or components are an equal target.
+- Tune the view on `engine/examples/symbol-graph/`. Use a large app such as visual-notes to check scale and find engine errors.
+
## Python type source
A Python root uses the same engine contracts as TypeScript. `engine/python/srcly_pyfacts.py` provides the evidence. The engine runs it once per generation.
diff --git a/docs/initiative-history.json b/docs/initiative-history.json
index 2a7060d3..9e0f3ccc 100644
--- a/docs/initiative-history.json
+++ b/docs/initiative-history.json
@@ -184,5 +184,21 @@
"Not built: a model passed to a serializer (model_validate, jsonable_encoder) as a use of its class. Nothing tracks it.",
"Not built for Python: following values across methods through instance state (self.x = ...). Partly tracked as pgm task-writers-of-a-model-field."
]
+ },
+ {
+ "title": "Show the logic, not only the components, in TS/TSX data flow",
+ "slug": "dataflow-logic-visibility",
+ "outcome": "completed",
+ "retired_at": "2026-11-12",
+ "summary": "Anchor traces in TS/TSX data flow now draw one box per symbol, name what happens along each edge, color symbols by kind, give frames a role, open on the main path and grow by depth, and link functions to what they call, read and write. Reconciled against origin/main 0bfb05ba.",
+ "reason": "All five milestones merged on 2026-10-29 (#268, #272, #274, #278, #279).",
+ "documentation": [
+ "docs/dataflow-coverage.md"
+ ],
+ "revision": "0bfb05ba",
+ "follow_ups": [
+ "Run audit round 3 on visual-notes against the symbol view. Nothing tracks it.",
+ "Below the cut line and not built: a types overlay, the symbol view in the Components lens, Python frame roles. Hub summaries are tracked as pgm task-hub-summaries-for-size-limited-slices."
+ ]
}
]
diff --git a/docs/intent/dataflow-logic-visibility/implementation-plan.md b/docs/intent/dataflow-logic-visibility/implementation-plan.md
deleted file mode 100644
index b5114475..00000000
--- a/docs/intent/dataflow-logic-visibility/implementation-plan.md
+++ /dev/null
@@ -1,127 +0,0 @@
----
-title: "Show the logic, not only the components — implementation plan"
-slug: "dataflow-logic-visibility"
-phase: plan
-status: draft
-last_updated: "2026-10-28"
----
-
-# Show the logic — implementation plan
-
-## Plan at a glance
-
-The order is: merge first, then edges and color, then frames and the server band, then the main path, then the call and read links.
-
-The merge comes first because it is the cheapest change that can alter the plan. Engine elements already carry the checker's `symbolId`. The anchor contract does not send it. When the contract sends it, the client can merge `doc`, `doc()`, `doc.error` and the `when` read into one box. That one change shows whether a 25-symbol first screen can be built on symbols in the client. It also shows whether the main path must be picked in the engine. Building frames, colors and links before that answer could put the work on the wrong layer.
-
-Each milestone adds one new kind of fact to the anchor contract: symbol, operation and symbol kind, frame role, path rank, links. Each fact has a client consumer and a small fixture. A switch keeps today's box model (`model=steps`) until the symbol view passes the visual-notes check. Any milestone can stop with the old view still working.
-
-The UX is tuned on a new small example app, which keeps the loop under a second. visual-notes is the scale check at the end of each milestone, with the October audit's known answers. Only one engine runs at a time on visual-notes (see `AGENTS.md`).
-
-Pushback: do not build the "+N stub" folding before the merge numbers exist. The flood may be mostly duplicate boxes, and in that case a simple rank is enough.
-
-## Implementation strategy
-
-- **First proof:** one box per symbol on the small example and on `/docs/[id]`, with box counts before and after.
-- **Primary seam:** the anchor-graph contract (`engine/src/srcly/contracts/anchor-graph.ts`) → `anchorAdapter.ts`. Every new fact crosses here, and the client never guesses a fact.
-- **Fast local loop:** `cd engine && pnpm vitest run test/srcly/anchor-graph.test.ts && cd ../client && pnpm vitest run src/viz/deps/flow`
-- **Local dependencies:** a new example app `engine/examples/symbol-graph/` with a route, a resource, a `Match`, a store, a `"use server"` query, a database client stub, a `.map`, and one plain TS module. It is opened in the browser with `?repo=/engine/examples/symbol-graph&view=graph`. Contract drift is caught by `contracts-current.test.ts` and the fake-engine snapshots.
-- **Scale confirmation:** visual-notes `/docs/[id]` and anchors T2–T4, run one at a time under `--max-old-space-size=4096`.
-- **Rollout and rollback:** a `model=symbols|steps` URL parameter and a toolbar switch. The default becomes `symbols` at the end of M4. Each contract field is additive, so an old client ignores it.
-
-## Milestone 1: One box per symbol
-
-This is the merge, and nothing else. Colors, frames and folding stay as they are.
-
-- **Change — send the symbol**
- - Add `symbol: string | null` (and `field: string | null` for a member read) to `anchorNodeSchema`. Fill them in `addElement` from `element.symbolId`. Then run `pnpm emit-contracts`, update the Pydantic mirror and refresh the fake-engine snapshots.
- - Add an engine test on `symbol-graph`: the binding `doc`, the call `doc()`, `doc.error` and the `when` read share one symbol, and the two `x`s in `same-name-negative.tsx` do not.
-- **Spike — the symbol on accessor calls and field reads**
- - Decision required: does `doc()` carry `doc`'s symbol, or the call's symbol?
- - Evidence to gather: dump `symbolId` for each node of the `/docs/[id]` trace.
- - Fallback: in the engine op, resolve the callee identifier and the receiver of a member read to their declaration symbol.
-- **Change — merge in the adapter**
- - Replace `passThroughs` with a merge by `(scope, symbol)`. A merged box lists its reads. A field read becomes the field name on its out-edge. A node with no symbol keeps its own box. Each merged edge keeps every original proof (#153).
- - Add `model=symbols|steps` to `navigation/location.ts` and a switch in `FlowTools.tsx`.
- - Add adapter tests: the `doc` case, a field read, the same name in two frames, and a node with no symbol.
-
-### Desired end state
-
-- `/docs/[id]` shows one `doc` box with out-edges labelled `when`, `error` and `prop doc`.
-- The box counts before and after for `/docs/[id]`, T2, T3 and T4 are recorded in the initiative map.
-- The decision rule is applied: a cut of a third or more means M4 ranks in the client; less means M4 ranks in the engine.
-
-## Milestone 2: Edges that say what happens, colored symbol kinds
-
-- **Change — operations on edges**
- - The engine sends `operation` on each edge, from the relation's or element's `operationKind` and from the kind traits (`map-result`, `parse`, `serialization`, `await`, `component-prop`, `function-return`).
- - The adapter folds each unnamed intermediate (call result, literal, prop read) into the edge between two symbols, and labels the edge with its operations. A click on the edge shows each folded step and its proof.
-- **Change — symbol kinds**
- - Add a `symbolKind` trait in `kinds.ts`: `state` (signal, store, resource, memo, context), `transform`, `origin`, `end` or `plain`. The anchor node carries it.
- - In `marks.ts`, the strong color goes to the symbol kind. `value` and `component` stop sharing the grey token. Frames become neutral. Update `FlowLegend.tsx` to match.
-
-### Desired end state
-
-- On the small example, `items` → `.map` → `rows` reads as two symbols and one labelled edge. The store and the resource are drawn as state.
-- The survey gate does not move, or its baseline is committed with the reason.
-
-## Milestone 3: Frames with a role, and the server band
-
-- **Change — frame role**
- - The engine sends `role` on `scope` nodes: `component`, `hook`, `helper`, `handler`, `fetcher`, `server` or `unknown`. The role comes from facts the engine already has: a component binding, a DOM event or `on*` relation, a resource fetcher argument, a `query`/`action`/`"use server"` wrapper, and a `use` name prefix for hooks. This code goes in `engine/src/srcly`, never in `engine/src/analysis`.
- - A frame shows a role word on a neutral tint. `unknown` is never shown as `component`.
-- **Change — server band and folded frames**
- - Frames with the `server` role, and the database ends below them, go into a server band. The layout gets a band region next to `in`/`center`/`out` in `layout.ts`.
- - A folded frame shows its symbols in and out (`id → DocDetail`) instead of a step count.
-
-### Desired end state
-
-- On `/docs/[id]`, `fetchDoc` and `prisma.doc.findUnique` sit in the server band. `DocView` reads `component`, and the `createResource` callback reads `fetcher`.
-
-## Milestone 4: Main path first, then grow
-
-- **Change — rank and first screen**
- - Rank each symbol by whether it is on a proven path from an origin to an end. This runs in the client or the engine, as M1 decided. The first screen shows the top path, about 25 symbols. Each side branch becomes a `+N` stub.
-- **Change — growth controls**
- - `+` and `−` add or remove one hop around the view. A depth slider (1, 2, 3, all) sits in the toolbar. Enter opens the selected stub. The depth goes into the URL and the view-state registry, so Back restores it.
-- **Change — default**
- - `model=symbols` becomes the default. `steps` stays as the switch.
-
-### Desired end state
-
-- `/docs/[id]` opens on about 25 symbols, from `URL: useParams()` through the server band to `DocumentViewer`.
-- Two `+` presses reach the round 2 audit's answer to T1.
-- `pnpm e2e` covers the depth control by `data-*` attributes.
-
-## Milestone 5: Calls and reads as thin links, plain TS
-
-- **Change — links**
- - The anchor op adds `links`: calls, callers and state reads or writes for each drawn frame. The links come from the facts behind `function-effects`, and the walk does not change. The canvas draws them as thin grey lines, with one toggle.
-- **Change — plain TS check**
- - Anchor a symbol in `engine/src/srcly/ops/anchor-graph.ts` and one in a plain utility of srcly's client (`client/src/utils/`). Check that each reads with no route, no Components lens, and no server band unless the trace crosses one.
-
-### Desired end state
-
-- Spidering out from a store in the small example shows its writers and readers as thin links next to the bold value flow.
-
-## Cross-cutting verification
-
-- After M4 and after M5, run audit round 3 on visual-notes with the October rubric. Record it in `docs/research/tsx-dataflow-audit-2026-10/`. The target is Focus 3 or higher on T1 and T3.
-- Take dark-mode and light-mode screenshots for each PR that changes the UI (`pnpm shot … --dark`).
-
-## Open decisions and spikes
-
-1. **Where the main path is ranked.** Output: client or engine. Evidence: the M1 counts. Fallback: an engine rank by proven origin-to-end paths.
-2. **The symbol for accessor calls.** Output: the existing `symbolId`, or resolution in the op. Evidence: the M1 spike dump. Fallback: resolve in `anchor-graph.ts`.
-
-## Below the cut line
-
-- Types overlay (the shape of the value at each step).
-- Symbol view for the Components lens and for route-graph ops.
-- Python frames with roles (the facts come from the Python helper; follow later).
-- Hub summaries across generations (`task-hub-summaries-for-size-limited-slices` stays its own task).
-- New walk semantics, such as calls through function-typed params.
-
-## Overlap with open work
-
-#210 (`assumed`) changes `anchor-graph.ts`, `kinds.ts`, the anchor contract, `anchorAdapter.ts`, `FlowLegend.tsx`, `FlowInspector.tsx` and `DependencyCanvas.tsx`. #211 changes `kinds.ts`. Land #210 and #211 before M1 starts, or build M1 on #210's branch. Both have backlog tasks to land them.
diff --git a/docs/intent/dataflow-logic-visibility/initiative-map.json b/docs/intent/dataflow-logic-visibility/initiative-map.json
deleted file mode 100644
index ec31239a..00000000
--- a/docs/intent/dataflow-logic-visibility/initiative-map.json
+++ /dev/null
@@ -1,487 +0,0 @@
-{
- "schema_version": "0.5.0",
- "initiative": {
- "title": "Show the logic, not only the components, in TS/TSX data flow",
- "slug": "dataflow-logic-visibility",
- "scale": "capability",
- "current_phase": "execution",
- "last_updated": "2026-10-28",
- "summary": "Make non-component code (helpers, hooks, handlers, server queries, transforms, state) visible in the TSX Data flow graph.",
- "status": "active",
- "next_action": "Merge the stack with gh stack merge; then run audit round 3 on visual-notes and retire"
- },
- "artifacts": {
- "intent_brief": {
- "path": "intent-brief.md",
- "status": "current",
- "updated_at": "2026-10-28"
- },
- "brainstorm": {
- "path": "brainstorm.md",
- "status": "not-requested",
- "updated_at": null
- },
- "shape_brief": {
- "path": "shape-brief.md",
- "status": "current",
- "updated_at": "2026-10-28"
- },
- "implementation_plan": {
- "path": "implementation-plan.md",
- "status": "current",
- "updated_at": "2026-10-28"
- }
- },
- "sources": [
- {
- "id": "S1",
- "type": "transcript",
- "location": "conversation 2026-10-28",
- "anchor": "see non TSX stuff in the data flow graph",
- "authority": "user"
- },
- {
- "id": "S2",
- "type": "repo",
- "location": "client/src/viz/deps/flow/anchorAdapter.ts",
- "anchor": "ANCHOR_FOLD_BUDGET, foldedBoxes symbolKind value",
- "authority": "repo"
- },
- {
- "id": "S3",
- "type": "repo",
- "location": "client/src/viz/deps/flow/marks.ts",
- "anchor": "value and component share --plc-tertiary",
- "authority": "repo"
- },
- {
- "id": "S4",
- "type": "repo",
- "location": "engine/src/srcly/ops/anchor-graph.ts",
- "anchor": "scopeFor: scope has name and file only; classify()",
- "authority": "repo"
- },
- {
- "id": "S5",
- "type": "doc",
- "location": "docs/research/tsx-dataflow-audit-2026-10/README.md",
- "anchor": "Round 2, Where it stands: Focus 7/16",
- "authority": "repo"
- },
- {
- "id": "S6",
- "type": "doc",
- "location": "https://github.com/byronwall/srcly/pull/256",
- "anchor": "Pro research prompt: dependency analysis and investigation UI",
- "authority": "inferred"
- },
- {
- "id": "S7",
- "type": "doc",
- "location": "https://github.com/byronwall/srcly/pull/153",
- "anchor": "projection witness preservation",
- "authority": "inferred"
- },
- {
- "id": "S8",
- "type": "repo",
- "location": "pgm/data srcly-backlog",
- "anchor": "task-route-data-traces-flood-the-canvas, task-hub-summaries-for-size-limited-slices",
- "authority": "repo"
- },
- {
- "id": "qa-round-1",
- "type": "qa",
- "location": "conversation",
- "anchor": "round 1",
- "authority": "user"
- },
- {
- "id": "qa-round-2",
- "type": "qa",
- "location": "conversation",
- "anchor": "round 2",
- "authority": "user"
- },
- {
- "id": "qa-round-3",
- "type": "qa",
- "location": "conversation",
- "anchor": "round 3",
- "authority": "user"
- },
- {
- "id": "S9",
- "type": "qa",
- "location": "conversation 2026-10-28",
- "anchor": "three boxes for doc: condense into a single box",
- "authority": "user"
- },
- {
- "id": "S10",
- "type": "repo",
- "location": "engine/src/analysis/program-evidence-compact-facts.ts",
- "anchor": "symbolId: string | null",
- "authority": "repo"
- },
- {
- "id": "qa-shape",
- "type": "qa",
- "location": "conversation",
- "anchor": "take both proposals",
- "authority": "user"
- }
- ],
- "intent": {
- "claims": [
- {
- "id": "C1",
- "statement": "Byron wants non-component code visible in the TS/TSX data-flow graph.",
- "kind": "goal",
- "sources": [
- "S1"
- ],
- "confidence": "high"
- },
- {
- "id": "C2",
- "statement": "The graph reads as all components because folded functions use the grey value mark, which matches the component grey, and frames carry no role.",
- "kind": "finding",
- "sources": [
- "S2",
- "S3",
- "S4"
- ],
- "confidence": "high"
- },
- {
- "id": "C3",
- "statement": "The goal is to see how data is reshaped, stored and handed across the server line, with components as one stop.",
- "kind": "hypothesis",
- "sources": [
- "S1"
- ],
- "confidence": "high"
- },
- {
- "id": "C4",
- "statement": "Folding and merging must keep the witness for each claim.",
- "kind": "constraint",
- "sources": [
- "S7"
- ],
- "confidence": "medium"
- },
- {
- "id": "C5",
- "statement": "Two equal jobs: follow a full route, and spider out from a known symbol to what it touches or depends on.",
- "kind": "goal",
- "sources": [
- "qa-round-1"
- ],
- "confidence": "high"
- },
- {
- "id": "C6",
- "statement": "Emphasize the symbols involved and how they relate and depend on each other.",
- "kind": "goal",
- "sources": [
- "qa-round-1"
- ],
- "confidence": "high"
- },
- {
- "id": "C7",
- "statement": "Must show transforms, state containers and the server boundary; plain helpers matter less.",
- "kind": "priority",
- "sources": [
- "qa-round-1"
- ],
- "confidence": "high"
- },
- {
- "id": "C8",
- "statement": "Engine changes are allowed when the client lacks the facts.",
- "kind": "decision",
- "sources": [
- "qa-round-1"
- ],
- "confidence": "high"
- },
- {
- "id": "C9",
- "statement": "Proposal (consider-contrast): symbol graph with named bindings as boxes, unnamed steps as labelled edges, functions as role-tinted frames, server as a band.",
- "kind": "hypothesis",
- "sources": [
- "qa-round-1"
- ],
- "confidence": "high"
- },
- {
- "id": "C10",
- "statement": "Spidering covers data plus calls and state reads/writes and callers, not only value flow.",
- "kind": "decision",
- "sources": [
- "qa-round-2"
- ],
- "confidence": "high"
- },
- {
- "id": "C11",
- "statement": "Start with a reasonable first screen; grow by keyboard shortcut or depth slider, never by many single clicks.",
- "kind": "decision",
- "sources": [
- "qa-round-2"
- ],
- "confidence": "high"
- },
- {
- "id": "C12",
- "statement": "Components are frames like hooks or helpers in the Data lens; the Components lens stays for the render tree.",
- "kind": "decision",
- "sources": [
- "qa-round-2"
- ],
- "confidence": "high"
- },
- {
- "id": "C13",
- "statement": "A route first screen reaches through the server boundary to the database.",
- "kind": "decision",
- "sources": [
- "qa-round-2"
- ],
- "confidence": "high"
- },
- {
- "id": "C14",
- "statement": "Value flow is the bold main line with its operation labelled; calls and reads/writes are thin grey links that can be toggled.",
- "kind": "decision",
- "sources": [
- "qa-round-3"
- ],
- "confidence": "high"
- },
- {
- "id": "C15",
- "statement": "First screen is the main path from origins to ends, about 25 symbols, with side branches as +N stubs that grow by key or slider.",
- "kind": "decision",
- "sources": [
- "qa-round-3"
- ],
- "confidence": "high"
- },
- {
- "id": "C16",
- "statement": "Plain TS modules with no routes or components are an equal target.",
- "kind": "decision",
- "sources": [
- "qa-round-3"
- ],
- "confidence": "high"
- },
- {
- "id": "C17",
- "statement": "Dial in the UX on a small reasonable example; use larger repos to build out the engine and find errors.",
- "kind": "decision",
- "sources": [
- "qa-round-3"
- ],
- "confidence": "high"
- },
- {
- "id": "C18",
- "statement": "All reads of one binding (doc, doc(), doc() in Match when) are one symbol box; the pattern recurs widely.",
- "kind": "goal",
- "sources": [
- "S9"
- ],
- "confidence": "high"
- },
- {
- "id": "C19",
- "statement": "Field reads merge into their symbol with the field name on the edge; symbol kind carries the strong color, frames neutral with a role word.",
- "kind": "decision",
- "sources": [
- "qa-shape"
- ],
- "confidence": "high"
- }
- ],
- "themes": [],
- "conflicts": [],
- "human_questions": []
- },
- "brainstorm": {
- "status": "not-requested",
- "requested_count": 7,
- "options": [],
- "selected_option": null
- },
- "shape": {
- "status": "current",
- "appetite": "One initiative: replace the Data lens box model and folding; walk, routes and Components lens unchanged.",
- "problem": "The Data lens draws functions as grey component-like boxes and repeats one symbol across several boxes, hiding transforms, state and the server hop.",
- "recommended_shape": "Symbol graph: one box per checker symbol per frame, bold labelled value-flow edges, thin toggled call/read links, role frames, server band, main path first with +/- and depth slider.",
- "success_moves": [
- "Merge by symbolId sent through the contract, never by label",
- "Prove the merge on a small fixture before visual-notes",
- "Keep the old box model behind a view switch until the visual-notes check passes",
- "Reuse function-effects facts for call and read links; no new walk semantics",
- "Keep each merged edge opening to every original proof (#153)"
- ],
- "first_proof": {
- "question": "Does merging by checker symbol cut the box count enough to build a main-path first screen on symbols?",
- "artifact": "anchor contract symbol field + anchorAdapter merge, run on a new small fixture and on visual-notes /docs/[id]",
- "decision_rule": "A cut of a third or more on /docs/[id], with no wrong merges, means build on symbols; otherwise move main-path selection into the engine first."
- },
- "rabbit_holes": [
- "Main-path selection needs a ranking the engine may not have",
- "symbolId may be null or the callee symbol for accessor calls",
- "Role detection for frames could drift into srcly-specific analysis edits"
- ],
- "open_decisions": []
- },
- "implementation": {
- "status": "active",
- "repository_snapshot": {
- "root": null,
- "revision": null,
- "worktree_status": null,
- "subsystems": [],
- "test_commands": [],
- "deployment_surfaces": []
- },
- "dependencies": [
- {
- "name": "anchor-graph contract",
- "first_proof_role": "real",
- "seam": "engine/src/srcly/contracts/anchor-graph.ts -> anchorAdapter.ts",
- "proof_strategy": "engine tests on a small example app, fake-engine snapshots, visual-notes scale check"
- }
- ],
- "fast_local_loop": {
- "command": "cd engine && pnpm vitest run test/srcly/anchor-graph.test.ts && cd ../client && pnpm vitest run src/viz/deps/flow"
- },
- "milestones": [
- {
- "id": "m1",
- "title": "One box per symbol",
- "outcome": "/docs/[id] shows one doc box with labelled out-edges; box counts before and after are recorded.",
- "dependencies_introduced": [
- "anchor contract symbol and field"
- ],
- "proof": "engine anchor-graph test on symbol-graph example; adapter tests; visual-notes counts",
- "valid_intermediate_state": "model=steps remains available; new contract fields are additive",
- "rollback": "switch to model=steps; drop the adapter merge",
- "status": "completed",
- "evidence": [
- "PR #268; visual-notes DocView 557 -> 225 boxes (60% cut); client/src/viz/deps/flow/symbolMerge.test.ts"
- ]
- },
- {
- "id": "m2",
- "title": "Edges that say what happens, colored symbol kinds",
- "outcome": "Unnamed intermediates fold into labelled edges; state, transform, origin and end symbols are colored apart.",
- "dependencies_introduced": [
- "edge operation",
- "symbolKind trait"
- ],
- "proof": "adapter tests; survey gate; screenshots",
- "valid_intermediate_state": "frames unchanged; legend matches marks",
- "rollback": "model=steps",
- "status": "completed",
- "evidence": [
- "PR #272; symbolView.ts, state mark"
- ]
- },
- {
- "id": "m3",
- "title": "Frames with a role, and the server band",
- "outcome": "Frames read component, hook, helper, handler, fetcher, server or unknown; server frames and DB ends sit in a band.",
- "dependencies_introduced": [
- "scope role"
- ],
- "proof": "engine test on symbol-graph roles; /docs/[id] screenshot",
- "valid_intermediate_state": "unknown role renders neutral",
- "rollback": "model=steps",
- "status": "completed",
- "evidence": [
- "PR #274; engine/src/srcly/scope-roles.ts"
- ]
- },
- {
- "id": "m4",
- "title": "Main path first, then grow",
- "outcome": "/docs/[id] opens on about 25 symbols; +/- and a depth slider grow it; symbols becomes the default.",
- "dependencies_introduced": [
- "path rank (client or engine per m1)"
- ],
- "proof": "adapter tests; pnpm e2e depth control; audit round 3",
- "valid_intermediate_state": "steps model stays behind the switch",
- "rollback": "default back to model=steps",
- "status": "completed",
- "evidence": [
- "PR #278; depth.ts, e2e Symbol view on symbol-graph"
- ]
- },
- {
- "id": "m5",
- "title": "Calls and reads as thin links, plain TS",
- "outcome": "Calls, callers and state reads/writes draw as toggled thin links; plain TS anchors read well.",
- "dependencies_introduced": [
- "anchor links from function-effects facts"
- ],
- "proof": "engine test; plain TS anchors in engine and client utils",
- "valid_intermediate_state": "links toggle off by default if noisy",
- "rollback": "hide links toggle",
- "status": "completed",
- "evidence": [
- "PR #279; links, plain TS anchor on client edgeLabel"
- ]
- }
- ],
- "cut_line": [
- "Types overlay",
- "Symbol view for Components lens",
- "Python frame roles",
- "Cross-generation hub summaries",
- "New walk semantics"
- ],
- "open_decisions": [
- "Where the main path is ranked (client or engine), decided by m1 counts",
- "Symbol identity for accessor calls"
- ]
- },
- "history": [
- {
- "date": "2026-10-28",
- "event": "Initiative created from Byron's request; intent brief drafted from repo, audit and open PRs."
- },
- {
- "date": "2026-10-28",
- "event": "Question round 1 recorded; consider-contrast picked a symbol-graph hybrid as the working proposal."
- },
- {
- "date": "2026-10-28",
- "event": "Question round 2 recorded."
- },
- {
- "date": "2026-10-28",
- "event": "Question round 3 recorded; intent brief rewritten as shape-ready."
- },
- {
- "date": "2026-10-28",
- "event": "Symbol merging added to intent from Byron screenshot; shape brief drafted."
- },
- {
- "date": "2026-10-28",
- "event": "Both shape choices taken; implementation plan written with five milestones."
- },
- {
- "date": "2026-10-28",
- "event": "M1-M5 implemented as a gh stack (#265, #268, #272, #274, #278, #279), rebased on main after #210, #211 and #273 landed. Main path ranks in the client (M1 rule met). Server band shipped as a frame tone, not a layout band."
- }
- ],
- "closure": null
-}
diff --git a/docs/intent/dataflow-logic-visibility/intent-brief.md b/docs/intent/dataflow-logic-visibility/intent-brief.md
deleted file mode 100644
index 207e3ce4..00000000
--- a/docs/intent/dataflow-logic-visibility/intent-brief.md
+++ /dev/null
@@ -1,52 +0,0 @@
-# Intent brief: show the logic, not only the components
-
-Status: shape-ready after three question rounds (2026-10-28).
-
-## My read
-
-The TS/TSX Data flow graph should show how data moves through the code: where it enters, which symbols carry it, where it is reshaped, where it is stored, where it crosses to the server, and where it ends. Today the graph "looks like everything is a TSX component", so the internal logic is hidden. Components stay on the path, but only as one kind of stop.
-
-There are two equal jobs:
-
-1. **Follow a full route.** Open a page route and see its data from the URL, through the server query and the database read, back to the values that the page renders.
-2. **Spider out from a symbol.** Start at a known symbol and branch out to what it touches and depends on. This includes value flow, the functions it calls, the state it reads or writes, and its callers.
-
-Plain TS (servers, CLIs, the engine itself, with no routes or components) is an equal target.
-
-Why the graph reads as components today:
-
-- A trace over 40 boxes folds each function except the anchor's own into one box (`anchorAdapter.ts`). `/docs/[id]` opens as 74 grey boxes named after functions.
-- A folded box uses the `value` mark. That mark has the same grey as `component`.
-- The engine's `scope` frame carries only a name and a file. It has no role.
-- Edge relation kinds (`argument`, `function-return`, `component-prop`, `http-bridge`) are in the contract, but the canvas draws them only as solid or dashed. Standard-library transforms are proof notes, not visible steps.
-- Calls, callers and state reads come from a separate op (`function-effects`). The view shows them only in the inspector rail.
-
-## What matters most
-
-- **Symbols are the boxes.** Named bindings are the boxes: parameters, `const`s, signals, stores, resources, memos and returns. An unnamed intermediate (a call result, a prop read, a literal) folds into the edge between two symbols.
-- **One symbol, one box.** All reads of the same binding are one box. In `/docs/[id]` today, `doc` (the binding), `doc()` (the accessor call) and `doc()` in `` are three boxes. They must be one `doc` box, with its uses listed on it or on its out-edges. Byron sees this pattern in many places.
-- **Value flow is the bold main line.** Its edge names the operation: `.map`, `await`, `JSON.parse`, `pack`, prop handoff. Calls and state reads or writes are thin grey links that you can turn on and off.
-- **Functions are light frames with a role.** The roles are component, hook, helper, event handler, resource fetcher and server query. A component is one frame among the others, and its props are symbols on the edges into it.
-- **Three things must be visible:** transforms, state containers and the server boundary (a separate band, with the database on its far side). Plain helpers matter less.
-- **The first screen is the main path.** It is about 25 symbols, from origins to ends. Side branches fold into `+N` stubs.
-- **Growth must be easy.** You grow the view with a keyboard shortcut or a depth slider. Many single clicks are not acceptable.
-- **A folded function still names its symbols.** It shows the symbols that go in and come out (`id → DocDetail`). It is never one anonymous grey box.
-
-## Boundaries
-
-- Engine changes are in scope when the client does not have the facts (function role, transform operation, call and state links). A new kind gets one entry in `kinds.ts`. A change that moves the survey gate commits its baseline.
-- Do not edit `engine/src/analysis` to recognize code that only srcly uses.
-- "Not analyzed" is never 0. A role that the engine cannot decide is "unknown". It must not fall back to "component".
-- Folding must keep the witness for each claim (#153). An edge that folds intermediates must still open to its proof.
-- The Components lens stays as the render-tree view.
-
-## What seems settled
-
-- **Proof approach.** Dial in the UX on a small example first: a fixture with a route, a store, a server query, a transform and some plain TS logic. Use larger repos (visual-notes, srcly's own client and server) to build out the engine and to find errors. The `/docs/[id]` answers from the October audit are the known-good check at scale.
-- **#256 (draft Pro prompt)** surveys how IDEs and analysis tools show value, call and effect dependencies at scale. Its gap map should inform the shape. Its rule holds: UI filtering cannot repair facts the engine does not have.
-- **#210 (`assumed` outcome) and #211 (library rows)** change edge colors and end words. The row model is a natural home for named transform operations.
-- **Backlog to absorb or replace:** `task-route-data-traces-flood-the-canvas`, `task-hub-summaries-for-size-limited-slices`, `task-show-and-match-when-is-not-a-dom-attribute`, `task-calls-through-function-typed-params-and-props`.
-
-## Next step after confirmation
-
-Shape the symbol-graph view. Define what a "symbol" is in the contract, how the main path is chosen, how the grow key and the slider work, and which fixture proves the first screen.
diff --git a/docs/intent/dataflow-logic-visibility/shape-brief.md b/docs/intent/dataflow-logic-visibility/shape-brief.md
deleted file mode 100644
index cb7ad064..00000000
--- a/docs/intent/dataflow-logic-visibility/shape-brief.md
+++ /dev/null
@@ -1,75 +0,0 @@
----
-title: "Show the logic, not only the components — shape brief"
-slug: "dataflow-logic-visibility"
-phase: shape
-status: draft
-last_updated: "2026-10-28"
----
-
-# Show the logic — what we are adding
-
-**Outcome:** A Data flow trace reads as symbols and how they depend on each other. Transforms, state and the server hop are visible, and components are one kind of frame among others.
-
-**Primary flow:** Open a route or anchor a symbol → see the main path (about 25 symbols) → grow it with a key or a slider → open any edge to its proof.
-
-## Feature scope
-
-```text
-Symbol graph (the Data lens, for routes and anchors, TS and TSX)
-├── PLANNED ADDITIONS
-│ ├── One box per symbol
-│ │ ├── Every read of one binding merges into its box: doc, doc(), doc.error, doc() in
-│ │ ├── Uses are listed on the box ("read 4×: when, prop doc, Title") and on its out-edges
-│ │ └── An unnamed intermediate (call result, literal, prop read) folds into the edge it sits on
-│ ├── Edges that say what happens
-│ │ ├── Value flow: bold, labelled with its operation (.map, await, JSON.parse, prop, return)
-│ │ └── Calls, callers and state reads/writes: thin grey links, one toggle
-│ ├── Symbol kinds you can tell apart
-│ │ └── State (signal, store, resource, memo, context) · transform result · origin · end
-│ ├── Function frames with a role
-│ │ ├── component · hook · helper · event handler · resource fetcher · server query · unknown
-│ │ └── A folded frame shows its symbols in and out (id → DocDetail), never one grey box
-│ ├── Server band
-│ │ └── Server-side frames and the database sit in their own band; the hop is drawn across it
-│ └── Main path first, then grow
-│ ├── First screen: the strongest origin-to-end path, about 25 symbols; side branches as +N stubs
-│ └── Grow with + / − or a depth slider (1, 2, 3, all); one stub opens with Enter
-└── LATER POSSIBILITIES
- └── A types overlay (the shape of the value at each step)
-```
-
-## Behavior
-
-| Situation | Expected result |
-| --- | --- |
-| Three reads of `doc` in one function | One `doc` box. Its out-edges are labelled `when`, `prop doc` and `Title`. |
-| A binding's reads in two functions | One box in each frame, linked by the hand-off edge. A symbol never spans frames. |
-| The engine sends no symbol for a node | The node stays its own box, as today. It is never merged by its label. |
-| An edge folds three intermediates | The edge label names the operation. A click shows every step and its proof, as today. |
-| A function's role is not known | The frame reads `unknown` with a neutral tint, not `component`. |
-| Plain TS module, no routes | Anchor mode works the same. There is no Components lens and no server band unless the trace crosses one. |
-| Grow past the walk limit | The slider stops at the engine's limit and names the stop, as frontiers do today. |
-
-## Examples
-
-`/docs/[id]`, first screen (*proposal*, wording not final):
-
-```text
-[URL] params.id ──id──▶ fetchDoc(id) ┆ server ┆ ──findUnique──▶ [DB read] doc row
- ◀──return DocDetail── doc (resource) ──when──▶ d ──prop doc──▶ DocumentViewer
- └──title──▶ +6 more
-```
-
-## Decisions and boundaries
-
-**Appetite:** One initiative. The work replaces the Data lens's box model and folding. Route discovery, the walk itself and the Components lens stay as they are.
-
-**Key decision:** Merge by the checker's symbol, not by label. Engine elements already carry `symbolId` (`program-evidence-compact-facts.ts`). The anchor contract does not send it yet. Sending it lets the client merge exactly. The label rule in `passThroughs` cannot merge `doc` and `doc()`, and it would wrongly merge two different `x`s. The roles, the operation names and the call links are also new contract fields. The client must not guess them.
-
-**Boundary:** No new walk semantics in this initiative. Calls and reads come from facts the engine already has (`function-effects`, relations). A merge must keep every witness: the edges of a merged box still open to each original proof (#153). The old box model stays behind a view switch until the symbol view passes the visual-notes check.
-
-**First proof:** **Try:** add `symbol` to the anchor contract, then merge by it in `anchorAdapter.ts`. Run this on a new small fixture (a route, a resource, a `Match`, a store, a server query, a `.map`, and one plain TS helper) and on `/docs/[id]`. **Observe:** the number of boxes before and after. Check that `doc`, `doc()` and the `when` read become one box, and that no two different bindings merge. **Decide:** if merging alone cuts `/docs/[id]` by a third or more, the main-path first screen can be built on symbols. If it does not, the main-path selection has to be done in the engine, and that comes before the frame and edge work.
-
-**Settled choices (2026-10-28):** A field read (`doc.error`) merges into `doc`, and its edge carries the field name. The symbol kind carries the strong color, and frames stay neutral with a role word.
-
-Intent: [intent-brief.md](intent-brief.md).
diff --git a/site/src/docs/pages/design.md b/site/src/docs/pages/design.md
index be0cea9b..b7de60b8 100644
--- a/site/src/docs/pages/design.md
+++ b/site/src/docs/pages/design.md
@@ -34,7 +34,7 @@ This prevents missing analysis from looking like improvement. Rankings can use d
## Use the right relationship for the question
-The TypeScript and TSX dependency graph shows imports. Data flow traces modeled values, function effects, or supported routes. An import edge does not prove value movement; a data-flow edge is not a general runtime trace. The [integration brief](https://github.com/byronwall/srcly/blob/main/docs/intent/tsx-dataflow-integration/intent-brief.md) and [logic visibility brief](https://github.com/byronwall/srcly/blob/main/docs/intent/dataflow-logic-visibility/intent-brief.md) explain the boundary.
+The TypeScript and TSX dependency graph shows imports. Data flow traces modeled values, function effects, or supported routes. An import edge does not prove value movement; a data-flow edge is not a general runtime trace. The [integration brief](https://github.com/byronwall/srcly/blob/main/docs/intent/tsx-dataflow-integration/intent-brief.md) and [logic visibility brief](https://github.com/byronwall/srcly/blob/0bfb05ba/docs/intent/dataflow-logic-visibility/intent-brief.md) explain the boundary.
Symbols mode shows which functions, classes, fields, or types reference a symbol. Edges carry a resolution label: exact, typed, or name-based. This gives more detail than imports without claiming value flow. See the [symbol graph brief](https://github.com/byronwall/srcly/blob/0bfb05ba/docs/intent/symbol-graph/intent-brief.md).