Replace the elicitation path with a plain Petrinaut-to-Flue chat.

apps/brunch-agent/.oxlintrc.json

@@ -26,7 +26,7 @@
         "paths": [
           {
             "name": "@hashintel/petrinaut",
-            "message": "The Brunch server must remain independent of Petrinaut implementations."
+            "message": "The Brunch server must remain independent of Petrinaut UI. Import catalogs from @hashintel/petrinaut-core, not the editor package."
           }
         ],
         "patterns": [
@@ -35,8 +35,12 @@
             "message": "The Brunch application must not depend on unpublished HASH packages."
           },
           {
-            "group": ["@hashintel/petrinaut/*", "@hashintel/petrinaut-*"],
-            "message": "The Brunch server must remain independent of Petrinaut implementations."
+            "group": [
+              "@hashintel/petrinaut/*",
+              "@hashintel/petrinaut-cli",
+              "@hashintel/petrinaut-cli/*"
+            ],
+            "message": "The Brunch server must remain independent of Petrinaut UI and CLI. Import catalogs from @hashintel/petrinaut-core, not the editor package."
           }
         ]
       }

apps/brunch-agent/AGENTS.md

@@ -2,10 +2,12 @@
 
 This application belongs to the Brunch context rooted at
 `../../libs/@hashintel/brunch-agent/`. Read that context's `AGENTS.md` and current `MISSION.md`
-before changing this application. Consult `CONTEXT.md` or historical design documents only when a
+before changing this application. If `MISSION.next.md` exists, it is a scratchpad for later
+concerns and is not execution authority. Consult `CONTEXT.md` or historical design documents only when a
 concrete question requires them; ADRs and specs are hypotheses, not implementation obligations.
 HASH root guidance takes precedence.
 
 The application composes the Flue runtime, HTTP routes, and the Brunch packages required by the
-current mission. It must remain independent of Petrinaut implementation packages;
-`apps/petrinaut-website` meets it through the AI SDK/HTTP transport.
+current mission. It must remain independent of Petrinaut UI (`@hashintel/petrinaut`); it may import
+published catalogs from `@hashintel/petrinaut-core` (for example user-guide page ids the panel
+already executes). `apps/petrinaut-website` meets the editor through the AI SDK/HTTP transport.

apps/brunch-agent/README.md

@@ -1,17 +1,43 @@
 # Brunch agent application
 
-## Run the process-model panel locally
+## Run the Petrinaut panel locally
 
 From the repository root, make `ANTHROPIC_API_KEY` available in the environment and run:
 
 ```sh
 yarn dev:brunch
 ```
 
-The command starts the Brunch server at `http://127.0.0.1:4321` and the real Petrinaut website at
-`http://127.0.0.1:4915`. The website proxies `/api/chat` to Brunch, where the panel runs the SDCPN
-process-model elicitor.
+The first step builds the Petrinaut libraries the panel imports (`dist/` and design-system
+codegen). Then it starts the Brunch server at `http://127.0.0.1:4321` and the real Petrinaut
+website at `http://127.0.0.1:4915`. The website proxies `/api/chat` to Brunch. The panel talks to one plain
+Flue chat agent: streamed text and reasoning, one server `ping` tool, and the existing Petrinaut
+`readPetrinautDoc` client tool. There is no elicitation, capture, or `brunch_ask` on this path.
 
-Conversations persist in `apps/brunch-agent/.data-wipe-me/conversations.db`. Owned target documents
-persist as per-document JSON files under `apps/brunch-agent/.data-wipe-me/target-documents/`.
-`BRUNCH_DEV_DB_PATH` and `BRUNCH_DEV_TARGET_DOCUMENT_DIR` override those local paths.
+Conversations persist in `apps/brunch-agent/.data-wipe-me/conversations.db`. `BRUNCH_DEV_DB_PATH`
+overrides that local path. Flue history is the conversation log; the browser may cache messages
+but reload hydrates from `GET /api/chat?id=`.
+
+The mounted Flue URL `/agents/chat/:id` requires the same principal and conversation identity (`x-brunch-principal` and `x-brunch-conversation`) as `/api/chat`; the path id is the hash of those, not a bearer token.
+
+Print a human-readable transcript of one conversation from that same Flue history (server already
+running):
+
+```sh
+yarn workspace @apps/brunch-agent transcript -- --principal <key> --id <conversationId>
+```
+
+## Voice dock
+
+A second input modality joins the same chat door. It is not a voice route and does not own
+provider audio or session state.
+
+|                       |                                                                                                                                                 |
+| --------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- |
+| URL                   | `POST /api/chat` (and `GET /api/chat?id=` to hydrate)                                                                                           |
+| Identity              | `x-brunch-principal` plus body `id` (the conversation id). The server hashes those into the Flue instance id.                                   |
+| Initial turn          | JSON `{ id, trigger: "submit-message", messages }` whose last user text part is the utterance.                                                  |
+| Client-tool follow-up | Same POST, with `messageId` of the assistant message and completed client-tool parts (`providerExecuted` not true). Correlated by `toolCallId`. |
+| Response              | AI SDK UI-message stream (SSE).                                                                                                                 |
+
+`OPTIONS /api/chat` is the CORS preflight for that same contract.

apps/brunch-agent/index.html

@@ -3,7 +3,7 @@
   <head>
     <meta charset="utf-8" />
     <meta name="viewport" content="width=device-width, initial-scale=1" />
-    <title>Elicitation harness — dev app</title>
+    <title>Brunch Flue chat</title>
   </head>
   <body>
     <div id="root"></div>

apps/brunch-agent/package.json

@@ -2,29 +2,26 @@
   "name": "@apps/brunch-agent",
   "version": "0.0.0-private",
   "private": true,
-  "description": "Remote Brunch server, local development loop, target gallery, and diagnostic probe surface.",
+  "description": "Remote Brunch server, local development loop, and Petrinaut /api/chat door.",
   "license": "AGPL-3.0",
   "type": "module",
   "scripts": {
-    "baseline:harness": "node --experimental-strip-types ../../libs/@hashintel/brunch-agent/evaluations/protocols/process-model-elicitation/baseline/harness-run.ts",
     "build": "vite build && vite build --config vite.client.config.ts",
     "dev": "vite dev",
     "fix:eslint": "oxlint --fix --type-aware --type-check --report-unused-disable-directives-severity=error .",
     "lint:eslint": "oxlint --type-aware --type-check --report-unused-disable-directives-severity=error .",
     "lint:tsc": "tsgo --noEmit",
     "petrinaut:dev": "vite dev --config petrinaut-local.vite.config.ts",
-    "test:unit": "vitest run --config vitest.config.ts"
+    "test:unit": "vitest run --config vitest.config.ts",
+    "transcript": "node --experimental-strip-types src/transcript-cli.ts"
   },
   "dependencies": {
     "@flue/opentelemetry": "2.0.3",
     "@flue/react": "2.0.3",
     "@flue/runtime": "2.0.3",
     "@flue/sdk": "2.0.3",
-    "@hashintel/brunch-agent": "workspace:*",
-    "@hashintel/brunch-agent-binding-flue": "workspace:*",
-    "@hashintel/brunch-agent-plugin-gherkin": "workspace:*",
-    "@hashintel/brunch-agent-plugin-sdcpn": "workspace:*",
     "@hashintel/brunch-agent-transport-aisdk": "workspace:*",
+    "@hashintel/petrinaut-core": "workspace:*",
     "@opentelemetry/api": "1.9.1",
     "hono": "4.13.2",
     "react": "19.2.6",
@@ -34,8 +31,6 @@
   "devDependencies": {
     "@earendil-works/pi-ai": "0.83.0",
     "@flue/vite": "2.0.3",
-    "@opentelemetry/sdk-trace-base": "2.9.0",
-    "@opentelemetry/sdk-trace-node": "2.9.0",
     "@types/node": "22.18.13",
     "@types/react": "19.2.14",
     "@types/react-dom": "19.2.3",

apps/brunch-agent/src/agent-ownership.ts

@@ -0,0 +1,36 @@
+/** Hono middleware: the mounted Flue route enforces the same ownership rule as `/api/chat`. */
+
+import { BRUNCH_PRINCIPAL_HEADER } from "@hashintel/brunch-agent-transport-aisdk/headers";
+
+import { ownsFlueInstance } from "./conversation-identity.ts";
+import { BRUNCH_CONVERSATION_HEADER } from "./conversation-payload.ts";
+
+import type { MiddlewareHandler } from "hono";
+
+export const agentOwnershipGuard = (mountPrefix: string): MiddlewareHandler => {
+  return async (context, next) => {
+    const principalKey = context.req.header(BRUNCH_PRINCIPAL_HEADER)?.trim();
+    const conversationId = context.req
+      .header(BRUNCH_CONVERSATION_HEADER)
+      ?.trim();
+    if (
+      principalKey === undefined ||
+      principalKey.length === 0 ||
+      conversationId === undefined ||
+      conversationId.length === 0
+    ) {
+      return context.json({ error: "unauthorized" }, 401);
+    }
+    const instanceId = context.req.path
+      .slice(mountPrefix.length)
+      .split("/")
+      .find((segment) => segment.length > 0);
+    if (
+      instanceId === undefined ||
+      !ownsFlueInstance(principalKey, conversationId, instanceId)
+    ) {
+      return context.json({ error: "forbidden" }, 403);
+    }
+    return next();
+  };
+};

apps/brunch-agent/src/agents/chat-agent.ts

@@ -0,0 +1,32 @@
+"use agent";
+/**
+ * One plain Flue chat agent for the Petrinaut panel throughline.
+ *
+ * No elicitation, capture, or plugin. The model can call a server-side ping
+ * and a browser-executed Petrinaut doc reader; Flue history is the session log.
+ */
+
+import { useModel, useTool } from "@flue/runtime";
+
+import { ping } from "../tools/ping.ts";
+import { readPetrinautDoc } from "../tools/read-petrinaut-doc.ts";
+
+export const CHAT_MODEL_ID =
+  process.env["BRUNCH_CHAT_MODEL"] || "claude-haiku-4-5";
+
+export function ChatAgent() {
+  useModel(`anthropic/${CHAT_MODEL_ID}`);
+  useTool(ping);
+  useTool(readPetrinautDoc);
+  return [
+    "You are a concise assistant inside the Petrinaut editor.",
+    "Call ping when you need to confirm the server tool path.",
+    "When the user asks how Petrinaut's UI works, call readPetrinautDoc.",
+    "A client-tool-result signal is JSON [{ toolCallId, toolName, output }]. Treat output as the browser's result for that call and continue helping the user.",
+  ].join("\n");
+}
+
+/**
+ * Pinned, and never to be edited: conversation storage keys on this literal.
+ */
+ChatAgent.agentName = "brunch-chat-agent";