@tbui17/omp-neuralwatt-provider 0.1.0 → 1.0.0

package/CHANGELOG.md

@@ -0,0 +1,11 @@
+# 1.0.0 (2026-07-05)
+
+
+### Bug Fixes
+
+* **ci:** add @semantic-release/git to devDependencies ([58c2108](https://github.com/tbui17/omp-neuralwatt-provider/commit/58c210855ee9321fc508c21db8302f47558e5224))
+* **ci:** add semantic-release devDependencies and lockfile ([ccc946e](https://github.com/tbui17/omp-neuralwatt-provider/commit/ccc946ead626131e6d99511d4a39594678c5384a))
+* **ci:** restore publish.yml workflow name with semantic-release ([f1a5224](https://github.com/tbui17/omp-neuralwatt-provider/commit/f1a52245fe989880f2090234c4774b955d5ee788))
+* **ci:** restore setup-bun and use bunx for semantic-release ([11560d0](https://github.com/tbui17/omp-neuralwatt-provider/commit/11560d0cd763cf3030e7dcc393f43915be336421))
+
+# Changelog

package/README.md

@@ -2,46 +2,121 @@
 
 Neuralwatt OpenAI-compatible provider plugin for [oh-my-pi](https://github.com/can1357/oh-my-pi).
 
-Provides GLM, Kimi, and Qwen models hosted on the Neuralwatt API (`https://api.neuralwatt.com/v1`).
+[![](https://img.shields.io/npm/v/@tbui17/omp-neuralwatt-provider.svg)](https://www.npmjs.com/package/@tbui17/omp-neuralwatt-provider)
 
-## Install
+## About Neuralwatt
+
+[Neuralwatt](https://www.neuralwatt.com) is an energy-optimized AI inference cloud. Its API is OpenAI-compatible and accessible at `https://api.neuralwatt.com/v1`, hosting frontier models from GLM (ZhipuAI), Kimi (MoonshotAI), and Qwen.
+
+Neuralwatt's main differentiator is **energy-based pricing**: instead of per-token markups, you pay a flat **$5.00/kWh** for actual GPU compute consumed. One rate applies to every model, so efficient mixture-of-experts models (e.g. Kimi K2.6, Qwen3.5) cost up to 95% less than traditional per-token pricing. Per-token billing is also available if you prefer the familiar model. See the [pricing page](https://portal.neuralwatt.com/pricing) for details.
+
+- **Company / technology:** [neuralwatt.com](https://www.neuralwatt.com) · [How optimization works](https://www.neuralwatt.com/technology)
+- **Cloud portal:** [portal.neuralwatt.com](https://portal.neuralwatt.com) — dashboard, API keys, playground
+- **Quickstart (raw API):** [portal.neuralwatt.com/docs/quickstart](https://portal.neuralwatt.com/docs/quickstart)
+
+## Quickstart
+
+### Prerequisites
+
+- [Bun](https://bun.sh) `>= 1.3.0`
+- [`omp`](https://github.com/can1357/oh-my-pi) installed and on your `PATH`
+- A Neuralwatt account and API key (starts with `sk-`)
+
+### Get an API key
+
+1. Sign up at [portal.neuralwatt.com](https://portal.neuralwatt.com).
+2. Open **Dashboard → API Keys**.
+3. Create a new key and copy it.
+
+Keep your key secret — never commit it to version control.
+
+### Install the provider
 
 ```
 omp install @tbui17/omp-neuralwatt-provider
 ```
 
-## Set your API key
+### Set your API key
 
+The plugin reads the key from the `NEURALWATT_API_KEY` environment variable.
+
+```sh
+# Linux / macOS (current shell)
+export NEURALWATT_API_KEY=sk-...
+
+# Persist for future shells — add the line above to:
+#   ~/.bashrc   (bash)
+#   ~/.zshrc    (zsh)
 ```
-export NEURALWATT_API_KEY=your-key-here
+
+```powershell
+# Windows PowerShell (current session)
+$env:NEURALWATT_API_KEY = "sk-..."
+
+# Persist across sessions
+[Environment]::SetEnvironmentVariable("NEURALWATT_API_KEY", "sk-...", "User")
 ```
 
-The plugin reads the key from the `NEURALWATT_API_KEY` environment variable. Set it before starting `omp`.
+### Smoke test
+
+omp -p --model neuralwatt/glm-5.2 "Say hello."
+```
+
+If `omp` returns a response, the provider is wired up correctly.
 
 ## Models
 
-16 models are registered under the `neuralwatt` provider:
-
-| Model ID | Family | Reasoning | Input |
-|---|---|---|---|
-| `glm-5.2` | GLM | yes | text |
-| `glm-5.2-fast` | GLM | no | text |
-| `glm-5.2-short` | GLM (200K ctx) | yes | text |
-| `glm-5.2-short-fast` | GLM (200K ctx) | no | text |
-| `kimi-k2.6` | Kimi | yes | text + image |
-| `kimi-k2.6-fast` | Kimi | no | text + image |
-| `qwen3.5-397b` | Qwen | yes | text |
-| `qwen3.5-397b-fast` | Qwen | no | text |
-| `qwen3.6-35b` | Qwen | yes | text + image |
-| `kimi-k2.7-code` | Kimi | yes | text + image |
-| `glm-5.2-short-fast-flex` | GLM flex | no | text |
-| `glm-5.2-short-flex` | GLM flex | yes | text |
-| `kimi-k2.6-flex` | Kimi flex | yes | text + image |
-| `kimi-k2.7-code-flex` | Kimi flex | yes | text + image |
-| `glm-5.2-flex` | GLM flex | yes | text |
-| `qwen3.6-35b-fast` | Qwen | no | text + image |
-
-Use with: `omp -p --model neuralwatt/glm-5.2 "your prompt"`
+16 canonical models are registered under the `neuralwatt` provider, spanning three families:
+
+- **GLM-5.2** (ZhipuAI) — long-context reasoning, up to 1M token context window
+- **Kimi K2.6 / K2.7** (MoonshotAI) — reasoning and vision, 256K context
+- **Qwen3.5 / Qwen3.6** (Qwen) — reasoning and vision, up to 256K context
+
+Naming conventions:
+
+| Suffix | Meaning |
+|---|---|
+| `-fast` | Reasoning disabled — lower latency, lower cost |
+| `-short` | 200K context window (vs 1M on full GLM-5.2) |
+| `-flex` | Flex variant — power adjusts to grid demand |
+| (none) | Standard variant, full reasoning |
+
+Vision-capable models (accept `image` input in addition to text): `kimi-k2.6`, `kimi-k2.6-fast`, `kimi-k2.7-code`, `kimi-k2.6-flex`, `kimi-k2.7-code-flex`, `qwen3.6-35b`, `qwen3.6-35b-fast`.
+
+For live per-model pricing, capabilities, and context windows, browse the catalog at [portal.neuralwatt.com/models](https://portal.neuralwatt.com/models). Energy-rate details live at [portal.neuralwatt.com/pricing](https://portal.neuralwatt.com/pricing).
+
+## Usage
+
+All examples use the `omp` CLI. Reference models as `neuralwatt/<model-id>`.
+
+### Default reasoning
+
+```sh
+omp -p --model neuralwatt/glm-5.2 "Explain quicksort."
+```
+
+### Fast (no reasoning, lower latency)
+
+```sh
+omp -p --model neuralwatt/glm-5.2-fast "Summarize this thread."
+```
+
+### Long context (200K window)
+
+```sh
+omp -p --model neuralwatt/glm-5.2-short "Analyze this document."
+```
+
+### Vision-capable
+
+```sh
+# Kimi K2.6 accepts image input
+omp -p --model neuralwatt/kimi-k2.6 "What's in this screenshot?"
+```
+
+### Reasoning effort
+
+Reasoning models accept an effort level via `thinkingLevelMap`. Supported levels vary per model but generally include: `off`, `minimal`, `low`, `medium`, `high`, `xhigh`. Lower effort trades reasoning depth for speed and lower energy use. Refer to the model detail page on [portal.neuralwatt.com/models](https://portal.neuralwatt.com/models) for the levels each model honors.
 
 ## License
 

package/extension.ts

@@ -449,36 +449,6 @@ const NEURALWATT_MODELS: ProviderModelConfig[] = [
   },
 ];
 
-// Legacy model ID aliases that map to canonical public models.
-const LEGACY_MODEL_ALIAS_MAP: Record<string, string> = {
-  "glm-5.1": "glm-5.2",
-  "glm-5.1-fast": "glm-5.2-fast",
-  "zai-org/GLM-5.1-FP8": "glm-5.2",
-  "moonshotai/Kimi-K2.5": "kimi-k2.6",
-  "kimi-k2.5-fast": "kimi-k2.6-fast",
-  "moonshotai/Kimi-K2.6": "kimi-k2.6",
-  "Qwen/Qwen3.5-397B-A17B-FP8": "qwen3.5-397b",
-  "Qwen/Qwen3.6-35B-A3B": "qwen3.6-35b",
-};
-
-function resolveLegacyModels(): ProviderModelConfig[] {
-  return Object.entries(LEGACY_MODEL_ALIAS_MAP).map(
-    ([legacyId, canonicalId]) => {
-      const canonical = NEURALWATT_MODELS.find(
-        (model) => model.id === canonicalId,
-      );
-      if (!canonical) {
-        throw new Error(`Missing canonical model for legacy alias ${legacyId}`);
-      }
-      return {
-        ...canonical,
-        id: legacyId,
-        name: `${canonical.name} (legacy ID)`,
-      };
-    },
-  );
-}
-
 // ============================================================================
 // Extension entry point
 // ============================================================================

package/package.json

@@ -1,20 +1,41 @@
 {
   "name": "@tbui17/omp-neuralwatt-provider",
-  "version": "0.1.0",
+  "version": "1.0.0",
   "description": "Neuralwatt OpenAI-compatible provider for oh-my-pi",
   "type": "module",
-  "main": "./extension.ts",
-  "files": ["extension.ts", "README.md", "LICENSE", "test"],
+  "files": [
+    "extension.ts",
+    "README.md",
+    "LICENSE",
+    "CHANGELOG.md"
+  ],
   "scripts": {
     "test": "bun test"
   },
-  "keywords": ["omp", "omp-extension", "omp-plugin", "neuralwatt", "ai", "llm"],
+  "devDependencies": {
+    "semantic-release": "^25.0.5",
+    "@semantic-release/changelog": "^6.0.3",
+    "@semantic-release/git": "^10.0.1"
+  },
+  "keywords": [
+    "omp",
+    "omp-extension",
+    "omp-plugin",
+    "neuralwatt",
+    "ai",
+    "llm"
+  ],
   "omp": {
-    "extensions": ["./extension.ts"]
+    "extensions": [
+      "./extension.ts"
+    ]
   },
   "engines": {
     "bun": ">=1.3.0"
   },
+  "publishConfig": {
+    "access": "public"
+  },
   "license": "MIT",
   "repository": {
     "type": "git",

package/test/integration.test.ts

@@ -1,38 +0,0 @@
-import { describe, test, expect } from "bun:test";
-
-const NEURALWATT_API_KEY = process.env.NEURALWATT_API_KEY;
-const BASE_URL = "https://api.neuralwatt.com/v1";
-
-describe.skipIf(!NEURALWATT_API_KEY)("neuralwatt integration", () => {
-	test("qwen3.6-35b-fast responds to a chat completion request", async () => {
-		const res = await fetch(`${BASE_URL}/chat/completions`, {
-			method: "POST",
-			headers: {
-				Authorization: `Bearer ${NEURALWATT_API_KEY}`,
-				"Content-Type": "application/json",
-			},
-			body: JSON.stringify({
-				model: "qwen3.6-35b-fast",
-				messages: [{ role: "user", content: "reply with exactly: pong" }],
-				max_tokens: 50,
-			}),
-		});
-
-		expect(res.ok).toBe(true);
-		expect(res.status).toBe(200);
-
-		const data = await res.json();
-		expect(data.object).toBe("chat.completion");
-		expect(data.model).toBe("qwen3.6-35b-fast");
-		expect(data.choices).toBeInstanceOf(Array);
-		expect(data.choices.length).toBeGreaterThan(0);
-
-		const message = data.choices[0]?.message;
-		expect(message?.role).toBe("assistant");
-		expect(message?.content).toBe("pong");
-
-		expect(data.usage).toBeDefined();
-		expect(typeof data.usage.prompt_tokens).toBe("number");
-		expect(typeof data.usage.completion_tokens).toBe("number");
-	});
-});