hecatehq · chicoxyzzy · May 5, 2026 · May 5, 2026 · May 5, 2026 · May 5, 2026
diff --git a/README.md b/README.md
@@ -78,26 +78,22 @@ Pinned image tags, binary tarballs (linux/darwin × amd64/arm64), checksums, com
 
 ### Add a provider
 
-On first boot, Chats is already available, but model chat needs at least one configured provider. Open **Providers**, click **Add provider**, pick a preset, and save the minimal setup:
+On first boot, Chats is already available. If Hecate detects a local runtime such as Ollama or LM Studio, the model chat setup can be one click: choose **Add detected providers** and Hecate adds the detected local endpoints with the preset defaults.
+
+![Chats first-run state — detected local providers and one-click Add detected providers setup](docs/screenshots/chat-empty.png)
+
+You can still configure providers manually from **Providers → Add provider**:
 
 - Cloud providers need an API key.
 - Local providers need a running local server URL, usually the preset default.
 - Custom OpenAI-compatible endpoints can be added from the same modal when the preset catalog is not enough.
 
-![Empty Providers tab on first boot — Add provider CTA](docs/screenshots/providers-empty.png)
-
-![Add provider modal on the Cloud tab — preset catalog](docs/screenshots/providers-presets.png)
-
-![Providers table populated with three providers — Health, Endpoint, Credentials, Models](docs/screenshots/providers.png)
-
 After a provider is saved, Hecate discovers models and the Chats model picker becomes routable. The full preset catalog, env bootstrapping, custom-endpoint walk-through, and credential rotation live in [`docs/providers.md`](docs/providers.md).
 
 ### Talk to it
 
 Chats is the primary day-to-day surface. It explains missing setup before you send a request, then lets you choose between model traffic and local coding-agent sessions.
 
-![Chats first-run state — no configured providers or available external agents yet](docs/screenshots/chat-empty.png)
-
 ![Chats workspace talking to a local Ollama llama3.1:8b model with sessions sidebar and inline runtime metadata](docs/screenshots/chat.png)
 
 ![Chats workspace with an external-agent file-write approval waiting for operator review](docs/screenshots/chat-agent-approval.png)
@@ -158,6 +154,12 @@ The embedded UI is a runtime console for the operator.
 
 ![Observability view — request ledger and route-report drilldown](docs/screenshots/observe.png)
 
+![Empty Providers tab — Add provider CTA](docs/screenshots/providers-empty.png)
+
+![Add provider modal — local preset catalog with detected runtime status](docs/screenshots/providers-presets.png)
+
+![Providers workspace — configured cloud and local providers with health, endpoint, credentials, and models](docs/screenshots/providers.png)
+
 ![Tasks workspace — task list with run state and approval queue](docs/screenshots/tasks.png)
 
 ![Costs workspace — balance card and usage table](docs/screenshots/costs.png)

diff --git a/docs-ai/skills/ui/SKILL.md b/docs-ai/skills/ui/SKILL.md
@@ -50,6 +50,29 @@ Every screen should answer the following quickly:
 
 When choosing between "pretty" and "operationally clear," choose clarity.
 
+## Accessibility baseline
+
+Accessibility is part of the design pass, not a cleanup pass. Every UI change
+should preserve keyboard, screen-reader, focus, contrast, and motion ergonomics
+unless there is an explicit product reason and a documented follow-up.
+
+- Use semantic HTML first: buttons for actions, anchors for navigation, labels
+  for form controls, tables for real tables, headings that reflect structure.
+- Every interactive control needs a visible focus state and a keyboard path.
+  Dropdowns, dialogs, popovers, and slideovers must support Escape / outside
+  dismissal where appropriate and return focus to the trigger.
+- Icon-only controls need accessible names. Status chips and color-coded states
+  need text, not color alone.
+- Modal and dialog work must include focus management, `aria-modal`, labelled
+  titles, and non-trapping escape paths.
+- Keep contrast readable in the dark operator theme. Muted text can be quiet,
+  but it still needs to be legible against panels and borders.
+- Respect reduced-motion expectations. Motion should orient; avoid relying on
+  animation to convey state.
+- Add or update tests for accessibility-sensitive behavior when changing UI
+  primitives: role/name queries, focus movement, disabled states, and keyboard
+  interactions are preferred over brittle DOM selectors.
+
 ## Information architecture
 
 Organize the UI around operator jobs, not components:

diff --git a/docs/providers.md b/docs/providers.md
@@ -28,7 +28,7 @@ The Providers tab starts empty:
 
 Click **Add provider** to open the modal:
 
-![Add provider modal — Cloud preset catalog](screenshots/providers-presets.png)
+![Add provider modal — Local preset catalog with detected runtime status](screenshots/providers-presets.png)
 
 
 1. Pick **Cloud** or **Local** at the top.
@@ -39,6 +39,23 @@ Click **Add provider** to open the modal:
    - **API Key** is shown for cloud and custom-cloud providers; stored encrypted at rest with `GATEWAY_CONTROL_PLANE_SECRET_KEY`.
 4. Click **Add provider**.
 
+The Local tab also runs a lightweight discovery check before you choose a
+preset. Hecate checks whether the expected command is on `PATH` (`ollama`,
+`lms`, `llama-server`, `local-ai` / `localai`) and probes each unique local
+HTTP endpoint once. The preset cards then show:
+
+- **Running** — the local HTTP API responded; model count is shown when the
+  provider returned one.
+- **Installed** — the command is available, but the HTTP server is not running
+  yet.
+- **Not detected** — no command on `PATH` and no response from the default
+  endpoint.
+
+`llamacpp` and `localai` share `127.0.0.1:8080` by default, so Hecate sends one
+HTTP request and reuses that result for both cards. The signal is advisory:
+adding the provider still uses the configured endpoint URL, and routing health
+continues to come from the normal `/admin/providers` probes.
+
 A provider you add is immediately routable. There is no separate enable/disable toggle — to take a provider out of rotation, delete it.
 
 ![Providers table populated with three providers — Health, Endpoint, Credentials, Models columns](screenshots/providers.png)
@@ -114,6 +131,9 @@ normalized assistant text, model, and token usage are.
 Every UI action maps to a control-plane endpoint:
 
 - `POST /admin/control-plane/providers` — add a provider. Body `{name, kind, protocol, base_url?, api_key?, custom_name?, preset_id?}`.
+- `GET /admin/control-plane/providers/local-discovery` — probe local presets
+  for command presence and default endpoint availability. Used by the Add
+  provider modal before a provider is created.
 - `DELETE /admin/control-plane/providers/{id}` — remove it.
 - `PATCH /admin/control-plane/providers/{id}` — partial update; accepts `base_url`, `name`, and `custom_name`.
 - `PUT /admin/control-plane/providers/{id}/api-key` — set the API key (empty `key` clears it).

diff --git a/docs/runtime-api.md b/docs/runtime-api.md
@@ -349,6 +349,48 @@ GET /v1/provider-presets
 
 The list is built from `config.BuiltInProviders()` — see [`docs/providers.md`](providers.md) for the full catalog and OpenAI-compatible custom-endpoint flow.
 
+### `GET /admin/control-plane/providers/local-discovery`
+
+Advisory discovery for the Providers tab's **Add provider → Local** catalog.
+The gateway checks whether the expected provider command is on `PATH` and
+probes each unique default local endpoint once. Shared endpoints, such as the
+`llama.cpp` / `LocalAI` default `127.0.0.1:8080/v1`, are only called once and
+then reused for every matching preset card.
+
+```json
+GET /admin/control-plane/providers/local-discovery
+→ 200
+{
+  "object": "local_provider_discovery",
+  "data": [
+    {
+      "preset_id": "ollama",
+      "name": "Ollama",
+      "base_url": "http://127.0.0.1:11434/v1",
+      "probe_url": "http://127.0.0.1:11434/api/tags",
+      "status": "running",
+      "command": "ollama",
+      "command_available": true,
+      "command_path": "/opt/homebrew/bin/ollama",
+      "http_available": true,
+      "model_count": 2,
+      "models": ["llama3.1:8b", "qwen2.5:7b"]
+    }
+  ]
+}
+```
+
+`status` is one of:
+
+- `running` — the HTTP probe returned 2xx.
+- `installed` — the command is present on `PATH`, but the default HTTP
+  endpoint did not respond.
+- `not_detected` — neither the command nor the default HTTP endpoint was found.
+
+This endpoint does not create or mutate provider records. It is a UX helper for
+the picker; routing readiness still comes from `GET /admin/providers` after the
+operator adds a provider.
+
 ### `GET /v1/agent-adapters`
 
 External coding-agent adapter catalog. This is the first discovery surface for

diff --git a/docs/screenshots/chat-agent-approval.png b/docs/screenshots/chat-agent-approval.png
diff --git a/docs/screenshots/chat-empty.png b/docs/screenshots/chat-empty.png
diff --git a/docs/screenshots/chat.png b/docs/screenshots/chat.png
diff --git a/docs/screenshots/observe.png b/docs/screenshots/observe.png
diff --git a/docs/screenshots/providers-presets.png b/docs/screenshots/providers-presets.png
diff --git a/docs/screenshots/providers.png b/docs/screenshots/providers.png
diff --git a/docs/screenshots/settings-external-agents.png b/docs/screenshots/settings-external-agents.png
diff --git a/docs/screenshots/tasks.png b/docs/screenshots/tasks.png