opencode-plugboard 0.20.0 → 0.21.0

package/README.md

@@ -43,7 +43,7 @@ export PLUGBOARD_KEY_MY_APP=choose-a-long-random-string
 {
   "plugins": [
     {
-      "package": "opencode-plugboard@0.20.0",
+      "package": "opencode-plugboard@0.21.0",
       "options": {
         "keys": { "my-app": "{env:PLUGBOARD_KEY_MY_APP}" },
         "models": ["anthropic/claude-sonnet-4-6", "github-copilot/*", "opencode-go/glm-5.3-flash"],
@@ -64,7 +64,7 @@ Keep the version pinned. A bare `"opencode-plugboard"` means `latest`, so a cach
 opencode api get /api/agent
 ```
 
-For an always-on server, run that from the service unit. See [warm-up](https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/warm-up.md).
+For an always-on server, run that from the service unit. See [warm-up][warm-up].
 
 **4. Call it.** The listener is on `127.0.0.1:4010`. This prints `pong`:
 
@@ -107,7 +107,7 @@ Give every tool its own key. Traffic, history and usage are then separated by ap
 }
 ```
 
-Callers send the key as `Authorization: Bearer <key>` (or `x-api-key` on the Anthropic routes). They never need the OpenCode server password. Removing a key takes effect on the running server at once.
+Callers send the key as `Authorization: Bearer <key>` (or `x-api-key` on the Anthropic routes). They never need the OpenCode server password. Removing a key takes effect on the running server at once; see [reloads and restarts][configuration-reload].
 
 #### Model allowlist
 
@@ -390,7 +390,7 @@ A server shared by a Discord bot, n8n, Open WebUI and a coding agent, with most
 {
   "plugins": [
     {
-      "package": "opencode-plugboard@0.20.0",
+      "package": "opencode-plugboard@0.21.0",
       "options": {
         // Reachable from the LAN. Non-loopback hosts refuse to start without keys.
         "listen": { "host": "10.0.1.50", "port": 4010 },
@@ -468,8 +468,8 @@ The protocol-level results are in the [compatibility matrix][compatibility]. It
 
 ## Known gaps
 
-- **Responses is stateless.** Every request carries the whole conversation, and nothing is stored. Clients that default to `store: true` fail on a second turn ([#147](https://github.com/jjjermiah/opencode-plugboard/issues/147)). Set `store: false`. For n8n, turn "Use Responses API" off.
-- **Stop sequences** are removed from the answer only on `/v1/messages` ([#145](https://github.com/jjjermiah/opencode-plugboard/issues/145)). Other endpoints return the matched text.
+- **Responses keeps no conversation.** Every request carries the whole conversation, and `previous_response_id` and `conversation` are rejected. The output items of recent answers are held in memory for an hour, so the `item_reference`s that the Vercel AI SDK sends back resolve; a restart clears them ([#147](https://github.com/jjjermiah/opencode-plugboard/issues/147)). For n8n, turn "Use Responses API" off.
+- **Stop sequences** are removed from the answer when they end it, on the Anthropic, OpenAI and Gemini routes. The host reports only that the turn stopped, so a sequence the provider ended on without returning it is not removed.
 - **Anthropic structured outputs** (`output_config.format`) are refused with a 400. Use `json_schema` on the OpenAI or Gemini endpoints.
 - **Gemini `candidateCount` above 1** is refused. The OpenAI endpoints serve `n`.
 - **Gemini model ids with a `/`** sent by the AI SDK's Google provider or LangChain's `google-genai` lack the `models/` prefix and get a 401. The `@google/genai` SDK works.
@@ -485,7 +485,7 @@ Open work is tracked in the [v0.1 milestone](https://github.com/jjjermiah/openco
 ## Security
 
 - **Bind address.** The listener binds `127.0.0.1` by default. A non-loopback `listen.host` refuses to start without `keys`. On a loopback bind, a `Host` header that is not `localhost`, `127.0.0.1` or `[::1]` gets a 403.
-- **Keys cross the network in clear text.** The endpoint speaks plain HTTP. For remote access, allow only known addresses in the firewall, tunnel over SSH, or put a TLS-terminating proxy in front. See [reaching plugboard from another host or container][reach].
+- **Keys cross the network in clear text.** The endpoint speaks plain HTTP. For remote access, allow only known addresses in the firewall, tunnel over SSH, or put a TLS-terminating proxy in front. See [reaching plugboard from another host or container][reach] and [remote access][remote-access].
 - **Keep keys out of the config file.** Use `{env:VAR}` and keep the values in the environment or a secret store. Every app needs its own key.
 - **`adminKey` reads usage only.** It cannot call the model routes, and it must differ from every app key.
 - **Keyless mode is loopback only.** Without `keys`, the one app is `default`, and only local callers can reach it.
@@ -510,28 +510,32 @@ The plugin swaps the caller's prompt, messages and tools in for OpenCode's own.
 
 MIT
 
-[configuration]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/configuration.md
-[configuration-apps]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/configuration.md#apps-and-keys
-[configuration-aliases]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/configuration.md#aliases-and-fallbacks
-[configuration-limits]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/configuration.md#concurrency-and-rate-limits
-[configuration-usage]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/configuration.md#usage-statistics
-[responses]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/responses.md
-[messages]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/messages.md
-[gemini]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/gemini.md
-[completions]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/completions.md
-[history]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/history.md
-[warm-up]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/warm-up.md
-[architecture]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/architecture.md
-[development]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/development.md
-[compatibility]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/compatibility.md
+Not affiliated with or endorsed by OpenAI, Anthropic, Google, GitHub or Microsoft. Product names are trademarks of their owners.
+
+[configuration]: https://jjjermiah.github.io/opencode-plugboard/configuration
+[configuration-apps]: https://jjjermiah.github.io/opencode-plugboard/configuration#apps-and-keys
+[configuration-aliases]: https://jjjermiah.github.io/opencode-plugboard/configuration#aliases-and-fallbacks
+[configuration-limits]: https://jjjermiah.github.io/opencode-plugboard/configuration#concurrency-and-rate-limits
+[configuration-usage]: https://jjjermiah.github.io/opencode-plugboard/configuration#usage-statistics
+[configuration-reload]: https://jjjermiah.github.io/opencode-plugboard/configuration#reloads-and-restarts
+[responses]: https://jjjermiah.github.io/opencode-plugboard/responses
+[messages]: https://jjjermiah.github.io/opencode-plugboard/messages
+[gemini]: https://jjjermiah.github.io/opencode-plugboard/gemini
+[completions]: https://jjjermiah.github.io/opencode-plugboard/completions
+[history]: https://jjjermiah.github.io/opencode-plugboard/history
+[warm-up]: https://jjjermiah.github.io/opencode-plugboard/warm-up
+[architecture]: https://jjjermiah.github.io/opencode-plugboard/architecture
+[development]: https://jjjermiah.github.io/opencode-plugboard/development
+[compatibility]: https://jjjermiah.github.io/opencode-plugboard/compatibility
 [release-candidate]: https://github.com/jjjermiah/opencode-plugboard/blob/main/compat/results/release-candidate.md
-[reach]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/any-app.md#reaching-plugboard-from-another-host-or-container
-[client-openai-python]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/openai-python.md
-[client-openai-js]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/openai-js.md
-[client-vercel-ai-sdk]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/vercel-ai-sdk.md
-[client-langchain-python]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/langchain-python.md
-[client-langchain-js]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/langchain-js.md
-[client-pydantic-ai]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/pydantic-ai.md
-[client-open-webui]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/open-webui.md
-[client-n8n]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/n8n.md
-[client-any-app]: https://github.com/jjjermiah/opencode-plugboard/blob/main/docs/clients/any-app.md
+[reach]: https://jjjermiah.github.io/opencode-plugboard/clients/any-app#reaching-plugboard-from-another-host-or-container
+[remote-access]: https://jjjermiah.github.io/opencode-plugboard/remote-access
+[client-openai-python]: https://jjjermiah.github.io/opencode-plugboard/clients/openai-python
+[client-openai-js]: https://jjjermiah.github.io/opencode-plugboard/clients/openai-js
+[client-vercel-ai-sdk]: https://jjjermiah.github.io/opencode-plugboard/clients/vercel-ai-sdk
+[client-langchain-python]: https://jjjermiah.github.io/opencode-plugboard/clients/langchain-python
+[client-langchain-js]: https://jjjermiah.github.io/opencode-plugboard/clients/langchain-js
+[client-pydantic-ai]: https://jjjermiah.github.io/opencode-plugboard/clients/pydantic-ai
+[client-open-webui]: https://jjjermiah.github.io/opencode-plugboard/clients/open-webui
+[client-n8n]: https://jjjermiah.github.io/opencode-plugboard/clients/n8n
+[client-any-app]: https://jjjermiah.github.io/opencode-plugboard/clients/any-app

package/dist/tui.js

@@ -0,0 +1,334 @@
+// Generated by .github/scripts/build-tui.ts from src/tui.tsx. Do not edit.
+import { effect as _$effect } from "@opentui/solid";
+import { createComponent as _$createComponent } from "@opentui/solid";
+import { insert as _$insert } from "@opentui/solid";
+import { memo as _$memo } from "@opentui/solid";
+import { createTextNode as _$createTextNode } from "@opentui/solid";
+import { insertNode as _$insertNode } from "@opentui/solid";
+import { setProp as _$setProp } from "@opentui/solid";
+import { createElement as _$createElement } from "@opentui/solid";
+/** @jsxImportSource @opentui/solid */
+/**
+ * The `/plugboard` usage page in OpenCode's terminal UI (package entry `./tui`). OpenCode loads it
+ * for every server that runs plugboard with the TUI feature: it reads the usage report over the
+ * server's own RPC (`context.client.rpc(UsageRpc).get`, src/rpc.ts), so it sees every app there.
+ * Nothing is served over HTTP, and nothing here runs on the server.
+ *
+ * Page keys: t, w, m, q pick the range (today, 7, 30, 90 UTC days); r reloads; esc goes back.
+ * Setup only registers the page and the command (no awaits). Shaping and formats: src/tui-format.ts.
+ */
+
+import { Plugin } from "@opencode/plugin/tui";
+import { TextAttributes } from "@opentui/core";
+import { createResource, createSignal, For, Show } from "solid-js";
+import { UsageRpc } from "../src/rpc.js";
+import { APP_COLUMNS, appRows, appValues, barCells, formatLine, MODEL_COLUMNS, modelRows, modelValues, presetRange, RANGE_PRESETS, totalItems, unavailableMessage } from "../src/tui-format.js";
+const PRESET_LABEL = preset => RANGE_PRESETS.find(item => item.preset === preset)?.label ?? preset;
+function UsagePage(props) {
+  const theme = props.context.theme;
+  const [query, setQuery] = createSignal({
+    preset: "month",
+    reload: 0
+  });
+  const [report] = createResource(query, async current => {
+    const range = presetRange(current.preset, Date.now());
+    try {
+      return await props.context.client.rpc(UsageRpc).get({
+        from: range.from,
+        to: range.to
+      });
+    } catch (error) {
+      props.context.ui.toast.show({
+        variant: "error",
+        title: "Plugboard",
+        message: unavailableMessage(error)
+      });
+      throw error;
+    }
+  });
+  props.context.keymap.layer(() => ({
+    commands: [{
+      bind: "escape",
+      title: "back",
+      run: props.onClose
+    }, ...RANGE_PRESETS.map(item => ({
+      bind: item.key,
+      title: `range ${item.label}`,
+      run: () => setQuery(current => ({
+        preset: item.preset,
+        reload: current.reload
+      }))
+    })), {
+      bind: "r",
+      title: "refresh",
+      run: () => setQuery(current => ({
+        ...current,
+        reload: current.reload + 1
+      }))
+    }]
+  }));
+  return (() => {
+    var _el$ = _$createElement("box"),
+      _el$2 = _$createElement("scrollbox"),
+      _el$3 = _$createElement("box"),
+      _el$4 = _$createElement("text"),
+      _el$6 = _$createElement("text"),
+      _el$7 = _$createTextNode(` · t today · w 7 days · m 30 days · q 90 days · r refresh · esc back`);
+    _$insertNode(_el$, _el$2);
+    _$setProp(_el$, "width", "100%");
+    _$setProp(_el$, "height", "100%");
+    _$insertNode(_el$2, _el$3);
+    _$setProp(_el$2, "flexGrow", 1);
+    _$insertNode(_el$3, _el$4);
+    _$insertNode(_el$3, _el$6);
+    _$setProp(_el$3, "flexDirection", "column");
+    _$setProp(_el$3, "gap", 1);
+    _$setProp(_el$3, "paddingLeft", 2);
+    _$setProp(_el$3, "paddingRight", 2);
+    _$setProp(_el$3, "paddingTop", 1);
+    _$setProp(_el$3, "paddingBottom", 1);
+    _$insertNode(_el$4, _$createTextNode(`Plugboard usage`));
+    _$insertNode(_el$6, _el$7);
+    _$insert(_el$6, (() => {
+      var _c$ = _$memo(() => !!report());
+      return () => _c$() ? `${report().from} to ${report().to} (UTC)` : PRESET_LABEL(query().preset);
+    })(), _el$7);
+    _$insert(_el$3, _$createComponent(Show, {
+      get when() {
+        return !report.error;
+      },
+      get fallback() {
+        return (() => {
+          var _el$9 = _$createElement("text");
+          _$insert(_el$9, () => unavailableMessage(report.error));
+          _$effect(_$p => _$setProp(_el$9, "fg", theme.text.feedback.error.base, _$p));
+          return _el$9;
+        })();
+      },
+      get children() {
+        return _$createComponent(Show, {
+          get when() {
+            return report();
+          },
+          get fallback() {
+            return (() => {
+              var _el$0 = _$createElement("text");
+              _$insertNode(_el$0, _$createTextNode(`Loading usage…`));
+              _$effect(_$p => _$setProp(_el$0, "fg", theme.text.muted, _$p));
+              return _el$0;
+            })();
+          },
+          get children() {
+            return _$createComponent(UsageBody, {
+              get report() {
+                return report();
+              },
+              theme: theme
+            });
+          }
+        });
+      }
+    }), null);
+    _$effect(_p$ => {
+      var _v$ = theme.background.base,
+        _v$2 = theme.text.base,
+        _v$3 = TextAttributes.BOLD,
+        _v$4 = theme.text.muted;
+      _v$ !== _p$.e && (_p$.e = _$setProp(_el$, "backgroundColor", _v$, _p$.e));
+      _v$2 !== _p$.t && (_p$.t = _$setProp(_el$4, "fg", _v$2, _p$.t));
+      _v$3 !== _p$.a && (_p$.a = _$setProp(_el$4, "attributes", _v$3, _p$.a));
+      _v$4 !== _p$.o && (_p$.o = _$setProp(_el$6, "fg", _v$4, _p$.o));
+      return _p$;
+    }, {
+      e: undefined,
+      t: undefined,
+      a: undefined,
+      o: undefined
+    });
+    return _el$;
+  })();
+}
+function UsageBody(props) {
+  const apps = () => appRows(props.report);
+  const models = () => modelRows(props.report, 10);
+  return (() => {
+    var _el$10 = _$createElement("box"),
+      _el$11 = _$createElement("box"),
+      _el$12 = _$createElement("text"),
+      _el$15 = _$createElement("text");
+    _$insertNode(_el$10, _el$11);
+    _$insertNode(_el$10, _el$12);
+    _$insertNode(_el$10, _el$15);
+    _$setProp(_el$10, "flexDirection", "column");
+    _$setProp(_el$10, "gap", 1);
+    _$setProp(_el$11, "flexDirection", "row");
+    _$setProp(_el$11, "flexWrap", "wrap");
+    _$setProp(_el$11, "gap", 2);
+    _$insert(_el$11, _$createComponent(For, {
+      get each() {
+        return totalItems(props.report.totals);
+      },
+      children: item => (() => {
+        var _el$18 = _$createElement("box"),
+          _el$19 = _$createElement("text"),
+          _el$20 = _$createElement("text");
+        _$insertNode(_el$18, _el$19);
+        _$insertNode(_el$18, _el$20);
+        _$setProp(_el$18, "flexDirection", "row");
+        _$setProp(_el$18, "gap", 1);
+        _$insert(_el$19, () => item.label);
+        _$insert(_el$20, () => item.value);
+        _$effect(_p$ => {
+          var _v$9 = props.theme.text.muted,
+            _v$0 = props.theme.text.base,
+            _v$1 = TextAttributes.BOLD;
+          _v$9 !== _p$.e && (_p$.e = _$setProp(_el$19, "fg", _v$9, _p$.e));
+          _v$0 !== _p$.t && (_p$.t = _$setProp(_el$20, "fg", _v$0, _p$.t));
+          _v$1 !== _p$.a && (_p$.a = _$setProp(_el$20, "attributes", _v$1, _p$.a));
+          return _p$;
+        }, {
+          e: undefined,
+          t: undefined,
+          a: undefined
+        });
+        return _el$18;
+      })()
+    }));
+    _$insertNode(_el$12, _$createTextNode(`Apps`));
+    _$insert(_el$10, _$createComponent(Show, {
+      get when() {
+        return apps().length > 0;
+      },
+      get fallback() {
+        return (() => {
+          var _el$21 = _$createElement("text");
+          _$insertNode(_el$21, _$createTextNode(`No requests in this range.`));
+          _$effect(_$p => _$setProp(_el$21, "fg", props.theme.text.muted, _$p));
+          return _el$21;
+        })();
+      },
+      get children() {
+        return [(() => {
+          var _el$14 = _$createElement("text");
+          _$insert(_el$14, () => formatLine(APP_COLUMNS.map(column => column.title), APP_COLUMNS));
+          _$effect(_$p => _$setProp(_el$14, "fg", props.theme.text.muted, _$p));
+          return _el$14;
+        })(), _$createComponent(For, {
+          get each() {
+            return apps();
+          },
+          children: row => (() => {
+            var _el$23 = _$createElement("box"),
+              _el$24 = _$createElement("text"),
+              _el$25 = _$createElement("text");
+            _$insertNode(_el$23, _el$24);
+            _$insertNode(_el$23, _el$25);
+            _$setProp(_el$23, "flexDirection", "row");
+            _$insert(_el$24, () => formatLine(appValues(row), APP_COLUMNS));
+            _$insert(_el$25, () => `  ${barCells(row.share, 24)}`);
+            _$effect(_p$ => {
+              var _v$10 = props.theme.text.base,
+                _v$11 = props.theme.categorical[0][200];
+              _v$10 !== _p$.e && (_p$.e = _$setProp(_el$24, "fg", _v$10, _p$.e));
+              _v$11 !== _p$.t && (_p$.t = _$setProp(_el$25, "fg", _v$11, _p$.t));
+              return _p$;
+            }, {
+              e: undefined,
+              t: undefined
+            });
+            return _el$23;
+          })()
+        })];
+      }
+    }), _el$15);
+    _$insertNode(_el$15, _$createTextNode(`Models`));
+    _$insert(_el$10, _$createComponent(Show, {
+      get when() {
+        return models().length > 0;
+      },
+      get fallback() {
+        return (() => {
+          var _el$26 = _$createElement("text");
+          _$insertNode(_el$26, _$createTextNode(`No model calls in this range.`));
+          _$effect(_$p => _$setProp(_el$26, "fg", props.theme.text.muted, _$p));
+          return _el$26;
+        })();
+      },
+      get children() {
+        return [(() => {
+          var _el$17 = _$createElement("text");
+          _$insert(_el$17, () => formatLine(MODEL_COLUMNS.map(column => column.title), MODEL_COLUMNS));
+          _$effect(_$p => _$setProp(_el$17, "fg", props.theme.text.muted, _$p));
+          return _el$17;
+        })(), _$createComponent(For, {
+          get each() {
+            return models();
+          },
+          children: row => (() => {
+            var _el$28 = _$createElement("text");
+            _$insert(_el$28, () => formatLine(modelValues(row), MODEL_COLUMNS));
+            _$effect(_$p => _$setProp(_el$28, "fg", props.theme.text.base, _$p));
+            return _el$28;
+          })()
+        })];
+      }
+    }), null);
+    _$effect(_p$ => {
+      var _v$5 = props.theme.text.muted,
+        _v$6 = TextAttributes.BOLD,
+        _v$7 = props.theme.text.muted,
+        _v$8 = TextAttributes.BOLD;
+      _v$5 !== _p$.e && (_p$.e = _$setProp(_el$12, "fg", _v$5, _p$.e));
+      _v$6 !== _p$.t && (_p$.t = _$setProp(_el$12, "attributes", _v$6, _p$.t));
+      _v$7 !== _p$.a && (_p$.a = _$setProp(_el$15, "fg", _v$7, _p$.a));
+      _v$8 !== _p$.o && (_p$.o = _$setProp(_el$15, "attributes", _v$8, _p$.o));
+      return _p$;
+    }, {
+      e: undefined,
+      t: undefined,
+      a: undefined,
+      o: undefined
+    });
+    return _el$10;
+  })();
+}
+export default Plugin.define({
+  id: "opencode-plugboard.tui",
+  setup(context) {
+    const [previous, setPrevious] = createSignal({
+      ...context.ui.router.current()
+    });
+    context.ui.router.register({
+      name: "plugboard.usage",
+      render: () => _$createComponent(UsagePage, {
+        context: context,
+        onClose: () => context.ui.router.navigate(previous())
+      })
+    });
+    context.keymap.layer(() => ({
+      mode: "global",
+      commands: [{
+        id: "plugboard.usage.open",
+        title: "Plugboard usage",
+        group: "Plugboard",
+        slash: {
+          name: "plugboard"
+        },
+        palette: true,
+        run() {
+          const current = context.ui.router.current();
+          if (current.type === "plugin" && current.name === "plugboard.usage") return;
+          // The router exposes a mutable store; keep the route before navigating.
+          setPrevious({
+            ...current
+          });
+          context.ui.dialog.clear();
+          context.ui.router.navigate({
+            type: "plugin",
+            name: "plugboard.usage"
+          });
+        }
+      }]
+    }));
+  }
+});

package/index.ts

@@ -27,7 +27,7 @@ import { withFallbacks } from "./src/fallback.js"
 import { withBudgets } from "./src/budgets.js"
 import { withLimits } from "./src/limits.js"
 import { createOpenAiRoute } from "./src/openai/route.js"
-import { ensureBackend, ensurePump, loadServingCatalog, processBudgets, processKeys, processLimits, processMetrics, processUsage, registerGeneration, sessionStore, withResponseCache } from "./src/runtime.js"
+import { ensureBackend, ensurePump, loadServingCatalog, processBudgets, processKeys, processLimits, processMetrics, processUsage, registerGeneration, setConfigError, sessionStore, withResponseCache } from "./src/runtime.js"
 import { createSessionCaller } from "./src/session/caller.js"
 import { installSessionHooks } from "./src/session/context.js"
 import { createHistoryPort, createSessionPort, createStoragePort, loadCatalog } from "./src/session/host.js"
@@ -43,10 +43,16 @@ export default Plugin.define({
       try {
         options = parseOptions(ctx.options)
       } catch (error) {
-        // Invalid options: stay inert and say why, instead of failing the Location's plugin load.
-        logError("options.invalid", { error: errorMessage(error) })
-        return
+        // Invalid options fail this setup, so the host lists the plugin as `failed` with the message
+        // (opencode packages/core/src/plugin.ts: a failed setup becomes slot.error, then the state).
+        // Nothing is registered, so the Location stays inert. The same message goes to the error log,
+        // and to `/v1/status` and `/readyz` while a listener from an earlier good generation serves (#159).
+        const message = errorMessage(error)
+        setConfigError(message)
+        logError("options.invalid", { error: message })
+        return yield* Effect.die(new Error(`plugboard options are invalid: ${message}`))
       }
+      setConfigError(undefined)
 
       // ---- every Location: join the shared listener (sync registration only).
       const sessions = sessionStore()
@@ -82,12 +88,13 @@ export default Plugin.define({
         // Usage is outermost, so it records the answer the client gets (src/usage.ts).
         // Caps are outside the fallbacks, so one request takes one slot even when it falls back (src/limits.ts).
         // Aliases written as a list retry the next target before the first output (src/fallback.ts).
-        // The response cache sits outside the fallbacks, so a hit never reaches a provider (src/cache.ts).
+        // The response cache is outside budgets and limits: a hit spends nothing, takes no slot and counts toward no
+        // rate window, and never reaches a provider (src/cache.ts). Only a miss is budgeted and limited.
         caller: withUsage(
-          // Budgets sit outside the limiter, so a budget refusal never takes a slot (src/budgets.ts).
-          withBudgets(
-            withLimits(
-              withResponseCache(
+          withResponseCache(
+            // Budgets sit outside the limiter, so a budget refusal never takes a slot (src/budgets.ts).
+            withBudgets(
+              withLimits(
                 withFallbacks(createSessionCaller(callerDeps), {
                   options,
                   loadCatalog: servingCatalog,
@@ -96,26 +103,26 @@ export default Plugin.define({
                     usage?.fallback(app)
                   },
                 }),
-                options,
+                () => limits,
+                processLimits(),
+                {
+                  onLimited: (app, reason) => {
+                    if (options.metrics) processMetrics().limited(app, reason)
+                    usage?.limited(app)
+                  },
+                },
               ),
-              () => limits,
-              processLimits(),
+              options.budgets,
+              usage,
+              processBudgets(),
               {
-                onLimited: (app, reason) => {
-                  if (options.metrics) processMetrics().limited(app, reason)
+                onRefused: (app) => {
+                  if (options.metrics) processMetrics().limited(app, "budget")
                   usage?.limited(app)
                 },
               },
             ),
-            options.budgets,
-            usage,
-            processBudgets(),
-            {
-              onRefused: (app) => {
-                if (options.metrics) processMetrics().limited(app, "budget")
-                usage?.limited(app)
-              },
-            },
+            options,
           ),
           usage,
         ),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-plugboard",
-  "version": "0.20.0",
+  "version": "0.21.0",
   "type": "module",
   "license": "MIT",
   "description": "Use the ChatGPT, Claude, Copilot and OpenCode Go logins in your OpenCode server from any app: an OpenCode V2 plugin serving OpenAI, Anthropic and Gemini compatible APIs.",
@@ -18,9 +18,9 @@
   "exports": {
     ".": { "types": "./index.ts", "default": "./index.ts" },
     "./rpc": { "types": "./rpc.ts", "default": "./rpc.ts" },
-    "./tui": { "types": "./tui.ts", "default": "./tui.ts" }
+    "./tui": { "types": "./tui.ts", "default": "./dist/tui.js" }
   },
-  "files": ["index.ts", "rpc.ts", "tui.ts", "src/", "README.md", "LICENSE"],
+  "files": ["index.ts", "rpc.ts", "tui.ts", "src/", "dist/", "README.md", "LICENSE"],
   "engines": {
     "bun": ">=1.4.2"
   },
@@ -31,6 +31,9 @@
     "mutations": "bun .github/scripts/mutation-check.ts",
     "flake-hunt": "bun .github/scripts/flake-hunt.ts",
     "coverage-manifest": "bun .github/scripts/coverage-manifest.ts",
+    "live-matrix": "bun test/live/matrix.ts",
+    "build:tui": "bun .github/scripts/build-tui.ts",
+    "prepack": "bun run build:tui",
     "smoke:pack": "bun .github/scripts/smoke-pack.ts",
     "docs:dev": "vitepress dev docs",
     "docs:build": "vitepress build docs",
@@ -38,6 +41,7 @@
   },
   "dependencies": {
     "@prometheus-io/client": "0.16.1",
+    "@types/escape-html": "1.0.4",
     "escape-html": "1.0.3",
     "ipaddr.js": "2.5.0",
     "lru-cache": "11.5.3",
@@ -52,6 +56,8 @@
     "@ai-sdk/openai": "4.0.91",
     "@ai-sdk/openai-compatible": "3.0.67",
     "@anthropic-ai/sdk": "0.133.0",
+    "@babel/core": "7.29.7",
+    "@babel/preset-typescript": "7.27.1",
     "@commitlint/cli": "21.2.3",
     "@commitlint/config-conventional": "21.2.3",
     "@google/genai": "2.28.0",
@@ -62,12 +68,12 @@
     "@opentui/core": "0.5.17",
     "@opentui/solid": "0.5.17",
     "@types/bun": "1.4.2",
-    "@types/escape-html": "^1.0.4",
     "@types/lcov-parse": "1.0.2",
     "@types/node": "^26.6.2",
     "@types/semver": "7.8.0",
     "ai": "7.0.137",
     "ajv": "8.20.0",
+    "babel-preset-solid": "1.9.12",
     "conventional-commits-parser": "7.1.3",
     "effect": "4.0.0-rc.112",
     "eventsource-parser": "4.1.1",

package/src/anthropic/response.ts

@@ -114,6 +114,8 @@ export async function runMessage(
   emit: (event: MessagesStreamEvent) => void,
 ): Promise<MessageRun> {
   const id = newId("msg")
+  // Only a non-streamed body needs the aggregate: a stream is already on the wire, so nothing grows (#153).
+  const aggregate = !exchange.stream
   const content: Block[] = []
   /** Blocks opened so far: the next block's `index`. */
   let nextIndex = 0
@@ -155,7 +157,7 @@ export async function runMessage(
     emit({ type: "content_block_stop", index: open.index })
     // The signature is always empty: the host does not expose the provider's thinking signature (#146), so a
     // replayed thinking block cannot carry the real one. Clients that replay thinking are told by the docs.
-    content.push(open.kind === "text" ? { type: "text", text: open.text } : { type: "thinking", thinking: open.text, signature: "" })
+    if (aggregate) content.push(open.kind === "text" ? { type: "text", text: open.text } : { type: "thinking", thinking: open.text, signature: "" })
     open = undefined
   }
   const releaseHeld = () => delta("text", holdback.release())
@@ -173,7 +175,7 @@ export async function runMessage(
       })
       ping()
     }
-    open.text += text
+    if (aggregate) open.text += text
     emit({
       type: "content_block_delta",
       index: open.index,
@@ -192,7 +194,7 @@ export async function runMessage(
     ping()
     emit({ type: "content_block_delta", index, delta: { type: "input_json_delta", partial_json: JSON.stringify(input) } })
     emit({ type: "content_block_stop", index })
-    content.push({ type: "tool_use", id: toolUseId, name: call.name, input })
+    if (aggregate) content.push({ type: "tool_use", id: toolUseId, name: call.name, input })
     calls++
   }