Merge pull request 'fix: reasoning effor mismatch' (#9) from fix/fix-reasoning-effort into main

Reviewed-on: #9

README.md

@@ -12,6 +12,12 @@ Run the installer without sudo:
 curl -fsSL https://git.krtss.de/StanPonomarev/stanp-opencode/raw/branch/main/install.sh | bash
 ```
 
+Pass `--playwright` to let the plugin automatically configure a Playwright MCP server:
+
+```sh
+curl -fsSL https://git.krtss.de/StanPonomarev/stanp-opencode/raw/branch/main/install.sh | bash -s -- --playwright
+```
+
 Rerun the command to upgrade to the latest published release.
 
 ## Install from a local checkout
@@ -22,6 +28,8 @@ This requires Bash, jq, and Bun 1.3.14+:
 bash /path/to/stanp-opencode/install-local.sh
 ```
 
+Add `--playwright` to this command to enable automatic Playwright MCP configuration.
+
 The script installs dependencies and builds the checkout before installing it.
 
 After either installation, restart `opencode2` or run `opencode2 service restart` for an existing service. Connect OpenAI and select `coordinate`. The required models must be available for the workers to run.
@@ -36,7 +44,7 @@ For an installed plugin, use `./plugins/stanp-opencode` as the plugin package pa
 
 ## Playwright
 
-The plugin prefers a local Playwright MCP installation. If ordinary Playwright is installed, it can download and run the MCP package with `npx`; otherwise Docker is required for the automatic fallback. Automatically configured servers allow five minutes for startup because the first launch may need to download a package or container image. Docker reuses its cached image on later launches. Existing `playwright` configuration, including its timeout, is preserved.
+When installed with `--playwright`, the plugin prefers a local Playwright MCP installation. If ordinary Playwright is installed, it can download and run the MCP package with `npx`; otherwise Docker is required for the automatic fallback. Automatically configured servers allow five minutes for startup because the first launch may need to download a package or container image. Docker reuses its cached image on later launches. Existing `playwright` configuration, including its timeout, is preserved.
 
 If `opencode mcp list` reports `MCP error -32001: Request timed out`, prefetch the Docker fallback with `docker pull mcr.microsoft.com/playwright/mcp`, or install `@playwright/mcp` in the project so the plugin can launch it directly. Then reconnect the server from `/mcps` or restart OpenCode. A custom `playwright` entry under `mcp.servers` can also set its own `timeout.startup` in milliseconds.
 

install-local.sh

@@ -1,6 +1,15 @@
 #!/usr/bin/env bash
 set -euo pipefail
 
+playwright=false
+if (($#)); then
+    if [[ "$#" -ne 1 || "$1" != "--playwright" ]]; then
+        printf 'Usage: %s [--playwright]\n' "$0" >&2
+        exit 2
+    fi
+    playwright=true
+fi
+
 for command in bun jq; do
     command -v "$command" >/dev/null || { printf 'Required command not found: %s\n' "$command" >&2; exit 1; }
 done
@@ -43,15 +52,19 @@ if [[ ! -e "$config_source" ]]; then
     config_source="$install/empty-config.json"
     printf '{}\n' > "$config_source"
 fi
-jq -e '
+jq -e --argjson playwright "$playwright" '
     if type != "object" then error("opencode.json must contain an object") else . end
     | if (.plugins // []) | type != "array" then error("opencode.json plugins must be an array") else . end
     | .["$schema"] = (.["$schema"] // "https://opencode.ai/config.json")
     | .plugins = (.plugins // [])
-    | if any(.plugins[]?; .package == "./plugins/stanp-opencode") then . else
+    | if any(.plugins[]?; .package == "./plugins/stanp-opencode") then
+        if $playwright then
+            .plugins |= map(if .package == "./plugins/stanp-opencode" then .options = ((.options // {}) + { playwright: true }) else . end)
+        else . end
+      else
         .plugins += [{
             package: "./plugins/stanp-opencode",
-            options: { mode: "gpt", provider: "openai", lunaPriority: true }
+            options: ({ mode: "gpt", provider: "openai", lunaPriority: true } + if $playwright then { playwright: true } else {} end)
         }]
       end
 ' "$config_source" > "$install/opencode.json"

install.sh

@@ -1,6 +1,15 @@
 #!/usr/bin/env bash
 set -euo pipefail
 
+playwright=false
+if (($#)); then
+    if [[ "$#" -ne 1 || "$1" != "--playwright" ]]; then
+        printf 'Usage: %s [--playwright]\n' "$0" >&2
+        exit 2
+    fi
+    playwright=true
+fi
+
 for command in bun curl jq sha256sum tar; do
     command -v "$command" >/dev/null || { printf 'Required command not found: %s\n' "$command" >&2; exit 1; }
 done
@@ -57,15 +66,19 @@ if [[ ! -e "$config_source" ]]; then
     config_source="$install/empty-config.json"
     printf '{}\n' > "$config_source"
 fi
-jq -e '
+jq -e --argjson playwright "$playwright" '
     if type != "object" then error("opencode.json must contain an object") else . end
     | if (.plugins // []) | type != "array" then error("opencode.json plugins must be an array") else . end
     | .["$schema"] = (.["$schema"] // "https://opencode.ai/config.json")
     | .plugins = (.plugins // [])
-    | if any(.plugins[]?; .package == "./plugins/stanp-opencode") then . else
+    | if any(.plugins[]?; .package == "./plugins/stanp-opencode") then
+        if $playwright then
+            .plugins |= map(if .package == "./plugins/stanp-opencode" then .options = ((.options // {}) + { playwright: true }) else . end)
+        else . end
+      else
         .plugins += [{
             package: "./plugins/stanp-opencode",
-            options: { mode: "gpt", provider: "openai", lunaPriority: true }
+            options: ({ mode: "gpt", provider: "openai", lunaPriority: true } + if $playwright then { playwright: true } else {} end)
         }]
       end
 ' "$config_source" > "$install/opencode.json"

src/config.ts

@@ -17,6 +17,7 @@ const assignment = z.strictObject({
 const common = {
     provider: z.string().trim().min(1).default("openai"),
     lunaPriority: z.boolean().default(true),
+    playwright: z.boolean().default(false),
 };
 
 export const settings = z.discriminatedUnion("mode", [

src/index.ts

@@ -114,10 +114,12 @@ export default Plugin.define({
         const assignments = selected.mode === "gpt" ? gptAgents : selected.agents;
         let workers: (typeof workerNames)[number][] = [];
         const active = new Set<string>();
-        const playwrightCommand = await playwrightMcpCommand(context.location.directory);
+        const playwrightCommand = selected.playwright
+            ? await playwrightMcpCommand(context.location.directory)
+            : undefined;
 
         await context.mcp.transform((editor) => {
-            if (!editor.get("playwright"))
+            if (playwrightCommand && !editor.get("playwright"))
                 editor.set("playwright", {
                     type: "local",
                     command: playwrightCommand,
@@ -166,6 +168,9 @@ export default Plugin.define({
                 const assignment = assignments[name];
                 const available = models.get(assignment.model)!;
                 editor.update(name, (agent) => {
+                    const hasVariant = available.model.variants.some(
+                        ({ id }) => id === assignment.effort,
+                    );
                     agent.name = Agent.Name.make(name);
                     agent.mode = "subagent";
                     agent.hidden = false;
@@ -174,10 +179,9 @@ export default Plugin.define({
                         providerID: Provider.ID.make(available.model.providerID),
                         id: Model.ID.make(available.model.id),
                     };
+                    if (hasVariant) agent.model.variant = Model.VariantID.make(assignment.effort);
                     agent.request = {
-                        settings: {
+                        settings: hasVariant ? {} : { reasoningEffort: assignment.effort },
-                            reasoningEffort: assignment.effort,
-                        },
                         headers: {},
                         body: {},
                     };
@@ -191,6 +195,9 @@ export default Plugin.define({
             }
             editor.update("coordinate", (agent) => {
                 const available = models.get(coordinate.model)!;
+                const hasVariant = available.model.variants.some(
+                    ({ id }) => id === coordinate.effort,
+                );
                 agent.name = Agent.Name.make("coordinate");
                 agent.mode = "primary";
                 agent.hidden = false;
@@ -200,8 +207,9 @@ export default Plugin.define({
                     providerID: Provider.ID.make(available.model.providerID),
                     id: Model.ID.make(available.model.id),
                 };
+                if (hasVariant) agent.model.variant = Model.VariantID.make(coordinate.effort);
                 agent.request = {
-                    settings: { reasoningEffort: coordinate.effort },
+                    settings: hasVariant ? {} : { reasoningEffort: coordinate.effort },
                     headers: {},
                     body: {},
                 };

src/tui.tsx

@@ -3,42 +3,44 @@ import { createResource, createSignal, For, Show } from "solid-js";
 import type { QuotaStatus, QuotaWindow } from "./quota.js";
 import { QuotaRpc } from "./quota-rpc.js";
 
+function percent(value: number) {
+    return value < 0.1 && value > 0 ? "<0.1" : value.toFixed(value < 10 ? 1 : 0);
+}
+
 function duration(seconds: number) {
     if (seconds % 86_400 === 0) return `${seconds / 86_400}d`;
     if (seconds % 3_600 === 0) return `${seconds / 3_600}h`;
     return `${Math.round(seconds / 60)}m`;
 }
 
-function percent(value: number) {
-    return value < 0.1 && value > 0 ? "<0.1" : value.toFixed(value < 10 ? 1 : 0);
-}
-
 function resetTiming(resetsAt: number, now: number) {
-    if (!Number.isFinite(resetsAt) || resetsAt <= 0 || !Number.isFinite(now))
+    if (!Number.isFinite(resetsAt) || resetsAt <= 0 || !Number.isFinite(now)) return "unavailable";
-        return "reset unavailable";
     const remaining = resetsAt - now;
-    if (remaining <= 0) return "reset due";
+    if (remaining <= 0) return "due";
-    if (remaining < 60_000) return "resets in <1m";
+    if (remaining < 60_000) return "<1m";
 
     const minutes = Math.floor(remaining / 60_000);
     const days = Math.floor(minutes / 1_440);
     if (days > 0) {
         const hours = Math.floor((minutes % 1_440) / 60);
-        return `resets in ${days}d${hours > 0 ? ` ${hours}h` : ""}`;
+        return `${days}d${hours > 0 ? ` ${hours}h` : ""}`;
     }
     const hours = Math.floor(minutes / 60);
     if (hours > 0) {
         const rest = minutes % 60;
-        return `resets in ${hours}h${rest > 0 ? ` ${rest}m` : ""}`;
+        return `${hours}h${rest > 0 ? ` ${rest}m` : ""}`;
     }
-    return `resets in ${minutes}m`;
+    return `${minutes}m`;
 }
 
-export function formatQuotaWindow(window: QuotaWindow & { remaining: number }, now = Date.now()) {
+export function formatQuotaWindow(
+    window: QuotaWindow & { remaining: number; sessionConsumed: number },
+    now = Date.now(),
+) {
     const label = window.label
         ? `${window.label} ${duration(window.seconds)}`
         : duration(window.seconds);
-    return `${label}: ${percent(window.remaining)}% left · ${resetTiming(window.resetsAt, now)}`;
+    return `${label}: ${percent(window.remaining)}% · ~${percent(window.sessionConsumed)}pp · ${resetTiming(window.resetsAt, now)}`;
 }
 
 export async function loadQuotaStatus(load: () => Promise<QuotaStatus | null>) {

tests/install-local.test.ts

@@ -199,6 +199,18 @@ test("reinstall replaces bundles while preserving existing settings and plugin o
     expect(await readFile(join(setup.installed, "index.js"), "utf8")).toContain('return "second"');
 });
 
+test("enables Playwright only when requested", async () => {
+    const setup = await fixture();
+    const result = await run(
+        ["bash", join(setup.checkout, "install-local.sh"), "--playwright"],
+        setup.cwd,
+        setup.env,
+    );
+    expect(result, result.stderr).toMatchObject({ exitCode: 0 });
+    const installed = JSON.parse(await readFile(join(setup.config, "opencode.json"), "utf8"));
+    expect(installed.plugins.at(-1).options.playwright).toBe(true);
+});
+
 for (const mode of ["dependency", "build", "validation", "config"] as const) {
     test(`${mode} failure leaves the previous installation and configuration untouched`, async () => {
         const invalidConfig = mode === "config" ? { plugins: "not-an-array" } : undefined;

tests/install.test.ts

@@ -44,7 +44,11 @@ async function run(command: string[], cwd: string, env: NodeJS.ProcessEnv) {
     return { exitCode, stdout, stderr };
 }
 
-async function install(mode: "native" | "checksum" | "legacy", configured = false) {
+async function install(
+    mode: "native" | "checksum" | "legacy",
+    configured = false,
+    playwright = false,
+) {
     const directory = await mkdtemp(join(tmpdir(), "stanp-install-"));
     directories.push(directory);
     const config =
@@ -187,7 +191,11 @@ exec sha256sum "$@"
         ),
     ]);
     await Promise.all([chmod(join(bin, "curl"), 0o755), chmod(join(bin, "sha256sum"), 0o755)]);
-    const result = await run(["bash", installer], directory, env);
+    const result = await run(
+        ["bash", installer, ...(playwright ? ["--playwright"] : [])],
+        directory,
+        env,
+    );
     return { directory, config, env, result };
 }
 
@@ -249,6 +257,13 @@ test("preserves an existing stanp-opencode configuration", async () => {
     expect(installed.permissions).toEqual(jsonConfig.permissions);
 }, 60_000);
 
+test("enables Playwright only when requested", async () => {
+    const { config, result } = await install("native", false, true);
+    expect(result, result.stderr).toMatchObject({ exitCode: 0 });
+    const installed = JSON.parse(await readFile(join(config, "opencode.json"), "utf8"));
+    expect(installed.plugins.at(-1).options.playwright).toBe(true);
+}, 60_000);
+
 for (const mode of ["checksum", "legacy"] as const) {
     test(`${mode} failure preserves the installed plugin in a custom path containing spaces`, async () => {
         const { directory, config, env, result } = await install(mode);

tests/plugin.test.ts

@@ -17,7 +17,16 @@ const providerID = Provider.ID.make("test");
 function model(id: string): ModelInfo {
     return Model.Info.default(providerID, Model.ID.make(id));
 }
-const allModels = ["gpt-5.6-luna", "gpt-5.6-sol", "gpt-6-astra"].map(model);
+function reasoningModel(id: string): ModelInfo {
+    return {
+        ...model(id),
+        variants: ["low", "medium", "high"].map((effort) => ({
+            id: Model.VariantID.make(effort),
+            settings: { reasoningEffort: effort },
+        })),
+    };
+}
+const allModels = ["gpt-5.6-luna", "gpt-5.6-sol", "gpt-6-astra"].map(reasoningModel);
 const cleanups: Plugin.Cleanup[] = [];
 afterEach(async () => {
     for (const cleanup of cleanups.splice(0)) await cleanup();
@@ -442,7 +451,13 @@ test("playwright MCP prefers direct MCP, uses npx for Playwright, and preserves
         await writeFile(executable, "#!/usr/bin/env node\n");
         await chmod(executable, 0o755);
 
-        const local = await host(allModels, { provider: "test" }, {}, [], localProject);
+        const local = await host(
+            allModels,
+            { provider: "test", playwright: true },
+            {},
+            [],
+            localProject,
+        );
         expect(local.mcpServers.get("playwright")).toEqual({
             type: "local",
             command: [executable],
@@ -475,7 +490,13 @@ test("playwright MCP prefers direct MCP, uses npx for Playwright, and preserves
         );
         await writeFile(ordinaryExecutable, "#!/usr/bin/env node\n");
         await chmod(ordinaryExecutable, 0o755);
-        const ordinary = await host(allModels, { provider: "test" }, {}, [], ordinaryProject);
+        const ordinary = await host(
+            allModels,
+            { provider: "test", playwright: true },
+            {},
+            [],
+            ordinaryProject,
+        );
         expect(ordinary.mcpServers.get("playwright")?.command).toEqual([
             "npx",
             "-y",
@@ -534,7 +555,7 @@ test("playwright MCP prefers direct MCP, uses npx for Playwright, and preserves
         await chmod(fakePlaywright, 0o755);
         expect((await playwrightMcpCommand(globalProject, ordinaryBin))[0]).toBe("docker");
 
-        const fallback = await host();
+        const fallback = await host(allModels, { provider: "test", playwright: true });
         expect(fallback.mcpServers.get("playwright")).toEqual({
             type: "local",
             command: ["docker", "run", "-i", "--rm", "--init", "mcr.microsoft.com/playwright/mcp"],
@@ -561,6 +582,9 @@ test("playwright MCP prefers direct MCP, uses npx for Playwright, and preserves
             timeout: { startup: 12_345 },
         });
         expect(existing.mcpServers.get("docs")?.codemode).toBe(true);
+
+        const disabled = await host();
+        expect(disabled.mcpServers.has("playwright")).toBe(false);
     } finally {
         if (originalPath === undefined) delete process.env.PATH;
         else process.env.PATH = originalPath;
@@ -594,7 +618,8 @@ test("reasoning settings and Luna priority use native advertised or official sup
         settings: { baseURL: "https://api.openai.com/v1" },
     };
     const ordinary = await host();
-    expect(ordinary.agents.get("fast-lo")?.request.settings).toEqual({ reasoningEffort: "low" });
+    expect(ordinary.agents.get("fast-lo")?.request.settings).toEqual({});
+    expect(String(ordinary.agents.get("fast-lo")?.model?.variant)).toBe("low");
     const official = await host([luna, ...allModels.slice(1)]);
     expect(official.agents.get("fast-hi")?.request.settings).toEqual({
         reasoningEffort: "high",
@@ -617,13 +642,43 @@ test("reasoning settings and Luna priority use native advertised or official sup
     expect(disabled.agents.get("fast-lo")?.request.settings).toEqual({ reasoningEffort: "low" });
 });
 
+test("coordinator defaults expose their effort variant without relying on it for requests", async () => {
+    for (const options of [{ provider: "test" }, { mode: "gpt", provider: "test" }] as const) {
+        const result = await host(allModels, options);
+        expect(result.agents.get("coordinate")?.model).toMatchObject({
+            id: "gpt-6-astra",
+            variant: "medium",
+        });
+        expect(result.agents.get("coordinate")?.request.settings.reasoningEffort).toBeUndefined();
+    }
+
+    const fallback = await host(allModels.slice(0, 2));
+    expect(fallback.agents.get("coordinate")?.model).toMatchObject({
+        id: "gpt-5.6-sol",
+        variant: "medium",
+    });
+    expect(fallback.agents.get("coordinate")?.request.settings.reasoningEffort).toBeUndefined();
+
+    const custom = await host([...allModels, model("local")], {
+        mode: "custom",
+        provider: "test",
+        agents: { ...gptAgents, coordinate: { model: "local", effort: "xhigh" } },
+    });
+    expect(custom.agents.get("coordinate")?.model?.variant).toBeUndefined();
+    expect(custom.agents.get("coordinate")?.request.settings.reasoningEffort).toBe("xhigh");
+});
+
 test("catalog updates activate late models and remove unavailable agents", async () => {
     const result = await host([]);
     await result.refresh(allModels);
     expect(String(result.agents.get("coordinate")?.model?.id)).toBe("gpt-6-astra");
+    expect(String(result.agents.get("coordinate")?.model?.variant)).toBe("medium");
+    expect(result.agents.get("coordinate")?.request.settings.reasoningEffort).toBeUndefined();
     await result.refresh(allModels.slice(0, 2));
     expect(result.agents.has("deep-lo")).toBe(false);
     expect(String(result.agents.get("coordinate")?.model?.id)).toBe("gpt-5.6-sol");
+    expect(String(result.agents.get("coordinate")?.model?.variant)).toBe("medium");
+    expect(result.agents.get("coordinate")?.request.settings.reasoningEffort).toBeUndefined();
     await result.refresh([]);
     expect(result.agents.has("coordinate")).toBe(false);
 });

tests/quota.test.ts

@@ -424,11 +424,12 @@ test("sidebar text places reset timing beside latest remaining quota", () => {
                 label: null,
                 used: 25,
                 remaining: 75,
+                sessionConsumed: 1.5,
                 resetsAt: 100_000 + 2 * 86_400_000 + 5 * 3_600_000,
             },
             100_000,
         ),
-    ).toBe("7d: 75% left · resets in 2d 5h");
+    ).toBe("7d: 75% · ~1.5pp · 2d 5h");
 });
 
 test("sidebar reset timing uses compact future units and handles boundaries", () => {
@@ -441,31 +442,33 @@ test("sidebar reset timing uses compact future units and handles boundaries", ()
                 label: null,
                 used: 25,
                 remaining: 75,
+                sessionConsumed: 1.5,
                 resetsAt: now + resetIn,
             },
             now,
         );
 
-    expect(format(2 * 3_600_000 + 15 * 60_000)).toBe("5h: 75% left · resets in 2h 15m");
+    expect(format(2 * 3_600_000 + 15 * 60_000)).toBe("5h: 75% · ~1.5pp · 2h 15m");
-    expect(format(3_600_000)).toBe("5h: 75% left · resets in 1h");
+    expect(format(3_600_000)).toBe("5h: 75% · ~1.5pp · 1h");
-    expect(format(15 * 60_000)).toBe("5h: 75% left · resets in 15m");
+    expect(format(15 * 60_000)).toBe("5h: 75% · ~1.5pp · 15m");
-    expect(format(60_000)).toBe("5h: 75% left · resets in 1m");
+    expect(format(60_000)).toBe("5h: 75% · ~1.5pp · 1m");
-    expect(format(59_999)).toBe("5h: 75% left · resets in <1m");
+    expect(format(59_999)).toBe("5h: 75% · ~1.5pp · <1m");
 });
 
-test("sidebar reset timing distinguishes stale and unavailable timestamps", () => {
+test("sidebar reset timing preserves labels and distinguishes stale and unavailable timestamps", () => {
     const window = {
         seconds: 18_000,
         id: "codex:primary",
         label: "GPT-5.6 Luna",
         used: 99.95,
         remaining: 0.05,
+        sessionConsumed: 1.5,
     };
     expect(formatQuotaWindow({ ...window, resetsAt: 999 }, 1_000)).toBe(
-        "GPT-5.6 Luna 5h: <0.1% left · reset due",
+        "GPT-5.6 Luna 5h: <0.1% · ~1.5pp · due",
     );
     expect(formatQuotaWindow({ ...window, resetsAt: Number.NaN }, 1_000)).toBe(
-        "GPT-5.6 Luna 5h: <0.1% left · reset unavailable",
+        "GPT-5.6 Luna 5h: <0.1% · ~1.5pp · unavailable",
     );
 });
 

tests/tui.test.ts

@@ -84,7 +84,7 @@ test("built quota slot stays reactive across pending, empty, failure, update, an
 
         pending.resolve(status("plus", 75));
         await setup.waitForFrame((frame) => frame.includes("ChatGPT quota (plus)"));
-        expect(setup.captureCharFrame()).toContain("5h: 75% left · reset due");
+        expect(setup.captureCharFrame()).toContain("5h: 75% · ~1.5pp · due");
 
         responses.push(async () => null);
         notifyUpdated();