chore: rebrand HexaUI / HexUI -> HexKit + README mark/gallery

CONTRACT.md

@@ -1,6 +1,6 @@
-# HexaUI Developer Contract (v1)
+# HexKit Developer Contract (v1)
 
-> The contract between a **developer's agent backend** and the **HexaUI proxy**.
+> The contract between a **developer's agent backend** and the **HexKit proxy**.
 > It imposes only the **shape of the API** (five endpoints) and that each stream
 > frame is **tagged with the framework that produced it**. You do *not* rewrite
 > your agent's events: you forward your framework's **native** events, tagged,
@@ -15,7 +15,7 @@
 ## 1. Roles
 
 ```
-browser ──HTTP/SSE──▶ HexaUI proxy ──HTTP/SSE──▶ your agent backend
+browser ──HTTP/SSE──▶ HexKit proxy ──HTTP/SSE──▶ your agent backend
                        (this platform)            (you implement this)
 ```
 
@@ -65,16 +65,16 @@ previous revision; see those sections at the end.)
 
 - `run_id` — opaque id the proxy assigns; accept it on `cancel`.
 - `input.messages` — the chat transcript.
-- **Provider API keys are *not* in the context.** HexUI does not store or forward
+- **Provider API keys are *not* in the context.** HexKit does not store or forward
   them — your backend reads its own provider keys (OpenAI, Google, …) from its
   own environment. The platform never holds your model credentials.
 - `context.files` — files the user attached to the conversation (persist across
   turns; forwarded every run). `content` is the decoded text for text mimes,
   `null` for binary (fetch by `id` is post-v1). Inline them into the prompt /
   provider content blocks as your framework needs.
-- `context.user` — caller identity. Always exactly three keys: `id` (the HexUI
+- `context.user` — caller identity. Always exactly three keys: `id` (the HexKit
   user uuid), `name` (display name or `null`), and `role` (free-text string or
-  `null`). HexUI does not interpret `role`; it's there so policy-aware runtimes
+  `null`). HexKit does not interpret `role`; it's there so policy-aware runtimes
   (hexgate, etc.) can scope per-call decisions to the calling user.
   **NEVER** includes email, password hash, or any internal identifier. An agent
   backend that doesn't use this can ignore the block.

Makefile

@@ -1,4 +1,4 @@
-# HexaUI dev Makefile — see QUICKSTART.md for prerequisites (uv + Node 18+).
+# HexKit dev Makefile — see QUICKSTART.md for prerequisites (uv + Node 18+).
 #
 # Common flow on a fresh clone:
 #

QUICKSTART.md

@@ -22,7 +22,7 @@ Without API keys the agents reply with an echo placeholder. That's fine for a fi
 
 ## One-time setup
 
-From the repo root (`HexUI/`):
+From the repo root (`HexKit/`):
 
 ```bash
 make setup
@@ -58,7 +58,7 @@ To enable real LLM replies (instead of echo), prepend the env var:
 AGENT_ENABLE_LLM=1 bash demo/scripts/run-backends.sh
 ```
 
-Provider keys live in the **agent backend's** environment, not the app — set `OPENAI_API_KEY` (Probe + healthcare/devops) and/or `GOOGLE_API_KEY` (Orbit) in your shell or in `demo/agent-server/.env` before starting the backends. HexUI never holds your model keys.
+Provider keys live in the **agent backend's** environment, not the app — set `OPENAI_API_KEY` (Probe + healthcare/devops) and/or `GOOGLE_API_KEY` (Orbit) in your shell or in `demo/agent-server/.env` before starting the backends. HexKit never holds your model keys.
 
 #### Terminal 2 — frontend
 
@@ -88,7 +88,7 @@ Open <http://localhost:8873>.
    startup when `PLATFORM_DEMO_USERS_FILE` is set (the `make dev` launcher sets
    it for you). Or sign up a fresh account at **/signup**.
 3. The chat shell loads with an empty greeting.
-4. Click the agent picker in the top bar and choose **Probe** (or any of the six demo agents — Probe, Orbit, Atlas, Forge, Healthcare, DevOps).
+4. Click the agent picker in the top bar and choose **Probe** (or any of the eight demo agents — Probe, Orbit, Atlas, Forge, Healthcare, DevOps, ITSM, HR).
 5. Type a message and send.
 6. With `AGENT_ENABLE_LLM=1` and a provider key in the agent backend's env → a real LLM reply.
    Without → an echo of your message (this confirms the full pipeline works).

README.md

@@ -1,21 +1,23 @@
 <div align="center">
 
-# HexUI
+<img src="assets/hexkit-mark.svg" width="72" height="72" alt="HexKit">
+
+# HexKit
 
 </div>
 
 A **UI/UX-first multi-agent chat platform**. Developers bring their own
-streaming agent backend (any framework); HexaUI provides the chat experience —
+streaming agent backend (any framework); HexKit provides the chat experience —
 a configurable, YAML-driven UI, conversation history, folders, file attachments
 — and a thin proxy that normalizes any framework's event stream into one schema
 the UI renders.
 
 <p align="center">
-  <img src="assets/hero.png" alt="HexaUI — the DevOps agent: a YAML-driven dashboard (service metrics + table) above a streaming chat, themed by the active agent's accent color." width="100%">
+  <img src="assets/hero.png" alt="HexKit — the DevOps agent: a YAML-driven dashboard (service metrics + table) above a streaming chat, themed by the active agent's accent color." width="100%">
 </p>
 
 <p align="center">
-  <a href="https://github.com/HexamindOrganisation/HexUI/actions/workflows/ci.yml"><img src="https://github.com/HexamindOrganisation/HexUI/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
+  <a href="https://github.com/HexamindOrganisation/hexkit/actions/workflows/ci.yml"><img src="https://github.com/HexamindOrganisation/hexkit/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
   <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT"></a>
   <img src="https://img.shields.io/badge/python-3.11+-3776AB?logo=python&logoColor=white" alt="Python 3.11+">
   <img src="https://img.shields.io/badge/node-18+-339933?logo=node.js&logoColor=white" alt="Node 18+">
@@ -67,10 +69,10 @@ events; the **proxy translates** and the **UI renders from YAML**. See
 | Path | Purpose |
 |---|---|
 | [custom-UI/](custom-UI/) | The product's heart: a React + TS library that renders a configurable agent UI from YAML (`<AgentUI>` + 12 built-in widgets). Theme bridge, streaming chat, the actions/`data_source` system. |
-| [front-app/](front-app/) | The HexaUI shell that consumes `custom-UI` and talks to the proxy. |
+| [front-app/](front-app/) | The HexKit shell that consumes `custom-UI` and talks to the proxy. |
 | [proxy-server/](proxy-server/) | The platform backend (FastAPI): JWT auth, conversations, folders, files, and the per-framework translators that normalize native events into the hexa SSE schema. Import package stays `platform_backend`. |
 | [packages/hexa-events/](packages/hexa-events/) | The internal event schema package consumed by the proxy (a local path dependency). |
-| [demo/](demo/) | The runnable reference backends: [`agent-server/`](demo/agent-server/) (a contract-conformant developer backend with 6 sample agents), [`hexgate-agent/`](demo/hexgate-agent/) (a standalone hexgate-wrapped backend), [`starter-agent/`](demo/starter-agent/) (a minimal **copy-me** backend — the whole contract in one file), and [`scripts/`](demo/scripts/) (run + smoke checks, incl. the `verify_backend.py` conformance CLI). |
+| [demo/](demo/) | The runnable reference backends: [`agent-server/`](demo/agent-server/) (a contract-conformant developer backend with 8 sample agents — Probe, Orbit, Atlas, Forge, Healthcare, DevOps, ITSM, HR), [`hexgate-agent/`](demo/hexgate-agent/) (a standalone hexgate-wrapped backend), [`starter-agent/`](demo/starter-agent/) (a minimal **copy-me** backend — the whole contract in one file), and [`scripts/`](demo/scripts/) (run + smoke checks, incl. the `verify_backend.py` conformance CLI). |
 | [legacy/](legacy/) | The dropped unified-runtime backend (`backend-runtime`), kept for reference. Not part of the live product. |
 | [CONTRACT.md](CONTRACT.md) | The developer contract — the one document an integrator reads. |
 
@@ -103,8 +105,8 @@ To get real model replies rather than the deterministic echo/canned fallback,
 put your provider keys in the **agent backend's environment** (`OPENAI_API_KEY`
 for Probe + the healthcare/devops agents, `GOOGLE_API_KEY` for Orbit) and start
 it with `AGENT_ENABLE_LLM=1` — see [`demo/agent-server/.env.sample`](demo/agent-server/.env.sample).
-HexUI never holds your model keys. The **Settings** page carries only your
-display name and a free-text `role`; if you point HexUI at a `hexgate`-wrapped
+HexKit never holds your model keys. The **Settings** page carries only your
+display name and a free-text `role`; if you point HexKit at a `hexgate`-wrapped
 agent, that role is forwarded to the agent and drives hexgate's per-tool policy
 + audit pipeline.
 
@@ -122,10 +124,161 @@ The bundled agents demonstrate the contract end to end:
 | **Forge** | `openai-agents` | the OpenAI Agents translator (canned native events) |
 | **Healthcare** | `openai-agents` (OpenAI) | a real clinical-assistant agent; HexGate-gated when `HEXGATE_KEY` is set, scoping per-tool policy to the caller's `context.user` role |
 | **DevOps** | `google-adk` (OpenAI via LiteLLM) | a real infra-assistant agent; HexGate-gated when `HEXGATE_KEY` is set, scoping per-tool policy to the caller's `context.user` role |
+| **ITSM** | `langchain` (deepagents) | a change-request assistant with a live lifecycle dashboard (refresh button → funnel metrics + change table updates as the agent's tools run) |
+| **HR** | `langchain` (deepagents) | an internal HR assistant; demonstrates stateful per-user data (`hr_state.py`) and role-gated tools when HexGate is wired |
 | **Hexgate Guard** | `hexgate` | a hexgate-wrapped agent that opens `User(user_id, role)` per run and emits audit decisions to the hexgate cloud (separate backend at [`demo/hexgate-agent/`](demo/hexgate-agent/)) |
 
 ---
 
+## Components
+
+Every agent UI is built out of these widgets. You compose them in a `ui.yaml` —
+`type` picks the widget, `position` + `size` lay it out on the grid,
+`data_source` wires it to one of your `/actions/{name}` endpoints. No React,
+no CSS.
+
+<table>
+<tr>
+<td width="50%">
+
+**`ai-response`** — streaming chat surface
+
+The flagship widget. Renders the assistant's turn as it streams; embeds tool calls, markdown, code blocks, files inline.
+
+```yaml
+- name: transcript
+  type: ai-response
+  position: { horizontal: left, vertical: high }
+  size: { width: 8, height: 520 }
+  empty_text: "Ask something…"
+  thinking_indicator: dots
+```
+
+</td>
+<td width="50%">
+
+**`ai-chat-input`** — composer with attachments
+
+Text input + file attach + keyboard send. Forwards the composed turn through `POST /conversations/{id}/messages`.
+
+```yaml
+- name: chat-input
+  type: ai-chat-input
+  position: { horizontal: left, vertical: low }
+  size: { width: 12, height: auto }
+  placeholder: "Message Hexgate Guard…"
+  rows: 2
+```
+
+</td>
+</tr>
+<tr>
+<td>
+
+**`metrics`** — KPI strip
+
+A row of labeled stat tiles. Pulls its values from a `data_source` action; auto-refreshes when a button-group action lists it in `refresh:`.
+
+```yaml
+- name: kpis
+  type: metrics
+  size: { width: 12, height: auto }
+  columns: 4
+  data_source: { action: change_summary }
+  metrics:
+    - { id: new, label: "New", format: number }
+```
+
+</td>
+<td>
+
+**`table`** — scrollable data grid
+
+Header row + virtualized rows. Sourced from an action that returns rows; can be refreshed by a button.
+
+```yaml
+- name: changes
+  type: table
+  size: { width: 12, height: auto }
+  data_source: { action: change_table }
+  has_header: true
+  empty_text: "No changes yet."
+```
+
+</td>
+</tr>
+<tr>
+<td>
+
+**`tool-calls`** — live tool log
+
+Shows each tool the agent calls in real time, with arguments, status, and result. Default landing surface for any unaddressed tool call.
+
+```yaml
+- name: tools
+  type: tool-calls
+  position: { horizontal: right, vertical: high }
+  size: { width: 4, height: 520 }
+  empty_text: "Tool calls will appear here."
+```
+
+</td>
+<td>
+
+**`markdown`** — rich text + code
+
+Static markdown sourced from a string, or live markdown that the agent updates via a data source (e.g. a runbook the agent edits).
+
+```yaml
+- name: runbook
+  type: markdown
+  size: { width: 8, height: auto }
+  data_source: { action: get_runbook }
+```
+
+</td>
+</tr>
+<tr>
+<td>
+
+**`form`** — structured input
+
+Labeled fields the user submits as one payload to an action. Each field declares its type; submit fires `POST /actions/{name}`.
+
+```yaml
+- name: deploy
+  type: form
+  size: { width: 6, height: auto }
+  action: deploy_service
+  fields:
+    - { name: env, label: Environment, type: text }
+    - { name: replicas, label: Replicas, type: number }
+```
+
+</td>
+<td>
+
+**`button-group`** — actions row
+
+A row of buttons, each tied to an action. Optional `refresh:` list re-pulls the named widgets after the action completes — the basis for the lifecycle dashboards in ITSM / DevOps.
+
+```yaml
+- name: ops
+  type: button-group
+  size: { width: 12, height: auto }
+  buttons:
+    - { label: "Refresh", action: refresh_changes,
+        refresh: [change-metrics, change-table] }
+```
+
+</td>
+</tr>
+</table>
+
+Four more widgets ship for layout and inputs: `dropdown`, `page-header`, `page-footer`, `spacer`. See [`custom-UI/src/registry/builtin.ts`](custom-UI/src/registry/builtin.ts) for the canonical registry and [demo/agent-server/src/agent_server/ui/](demo/agent-server/src/agent_server/ui/) for complete `ui.yaml` examples (the ITSM and DevOps agents stitch most of the widgets together into a live ops dashboard).
+
+---
+
 ## The two things a developer configures
 
 1. **`ui.yaml`** — which widgets, where, and the accent color. Placed/served by

custom-UI/src/lib/context-card.tsx

@@ -3,7 +3,7 @@ import { useAgentUIContext } from "../runtime/context.js";
 
 /**
  * Frame for a display widget (table / markdown) that can toggle its content
- * into the conversation's model context. Ported from the HexaUI "Context Toggle"
+ * into the conversation's model context. Ported from the HexKit "Context Toggle"
  * design: a header (icon + caption + the pill switch) over the widget body; when
  * on, the card lights to the agent accent with a one-shot scan/glow.
  *

custom-UI/src/registry/builtin.ts

@@ -68,7 +68,7 @@ export const builtinWidgets: AnyWidgetDefinition[] = [
     schema: AiChatInputWidgetSchema,
     component: AiChatInputWidgetComponent,
     chromeless: true,
-    // HexaUI: the composer is constant chrome pinned to the bottom of the chat
+    // HexKit: the composer is constant chrome pinned to the bottom of the chat
     // area (the transcript scrolls above it), not a widget in the content flow.
     slot: "footer",
   }),

custom-UI/src/schema/page.ts

@@ -27,7 +27,7 @@ export const PageSchema = {
     layout_type: { enum: ["grid", "flex"] },
     /**
      * The active agent's signature color — the ONE variable that recolors the
-     * whole page (HexaUI's core principle). Bridges to shadcn `--primary` /
+     * whole page (HexKit's core principle). Bridges to shadcn `--primary` /
      * `--ring` (and the legacy `--accent` agent usage). Takes precedence over
      * `theme.accent`. Hex (`#RGB`, `#RRGGBB`, `#RRGGBBAA`).
      */

custom-UI/src/shadcn.css

@@ -1,14 +1,14 @@
 /*
- * agent-ui shadcn base layer — themed to HexaUI.
+ * agent-ui shadcn base layer — themed to HexKit.
  *
  * Consumer usage (from your app's CSS entry):
  *   @import "agent-ui/shadcn.css";
  *
  * Pair with the Tailwind preset:
  *   presets: [require("agent-ui/tailwind-preset")]
  *
- * The shadcn HSL tokens are mapped onto the HexaUI neutral scale (Spec §2), so
- * every native widget inherits the HexaUI look automatically. The ONE color in
+ * The shadcn HSL tokens are mapped onto the HexKit neutral scale (Spec §2), so
+ * every native widget inherits the HexKit look automatically. The ONE color in
  * the product is the active agent's hue: `page.main_color` → `--primary` /
  * `--ring` (set inline at the page root by the theme bridge) and a raw hex
  * `--accent-color` for links / carets / status dots. Chrome stays monochrome —
@@ -20,7 +20,7 @@
 
 @layer base {
   :root {
-    /* ----- HexaUI light neutral scale (Spec §2) ----- */
+    /* ----- HexKit light neutral scale (Spec §2) ----- */
     --background: 0 0% 100%; /* --bg #ffffff */
     --foreground: 216 14% 11%; /* --text #191c21 */
 
@@ -52,7 +52,7 @@
 
     --radius: 0.6875rem; /* --r-md 11px */
 
-    /* ----- HexaUI raw tokens (for widget chrome that needs them) ----- */
+    /* ----- HexKit raw tokens (for widget chrome that needs them) ----- */
     --hx-bg: #ffffff;
     --hx-bg-2: #f5f6f7;
     --hx-surface: #ffffff;
@@ -78,7 +78,7 @@
   }
 
   .dark {
-    /* ----- HexaUI dark neutral scale (Spec §2) ----- */
+    /* ----- HexKit dark neutral scale (Spec §2) ----- */
     --background: 240 2% 12%; /* --bg #1e1e1f */
     --foreground: 220 9% 93%; /* --text #ecedef */
 
@@ -106,7 +106,7 @@
     --input: 240 3% 22%;
     --ring: 174 43% 47%;
 
-    /* ----- HexaUI raw tokens (dark) ----- */
+    /* ----- HexKit raw tokens (dark) ----- */
     --hx-bg: #1e1e1f;
     --hx-bg-2: #181819;
     --hx-surface: #282829;

custom-UI/src/styles.css

@@ -281,7 +281,7 @@
 }
 
 /* ============================================================
- * HexaUI chat — metaball thinking loader, streaming caret, status shimmer.
+ * HexKit chat — metaball thinking loader, streaming caret, status shimmer.
  * All tint from the agent color via --accent-color (set by the theme bridge),
  * falling back to the shadcn primary.
  * ============================================================ */
@@ -386,7 +386,7 @@
 
 /* ============================================================
    File handling — composer attach popover + attached pills.
-   Ported from the HexaUI design handoff; tokens scoped to `.hxf`
+   Ported from the HexKit design handoff; tokens scoped to `.hxf`
    and mapped onto the shadcn palette + the active agent accent.
    The agent stays the only color; glyphs are neutral.
    ============================================================ */