nvicode 0.1.5 → 0.1.6

package/README.md

@@ -1,6 +1,7 @@
-# nvicode
+# Navicode - one click Nvidia NIM to Claude code connection
 
-Run Claude Code through NVIDIA-hosted models using a local Anthropic-compatible gateway.
+Run Claude Code through NVIDIA-hosted models or OpenRouter using a simple CLI wrapper.
+Use top open-source model APIs on NVIDIA Build for free, with `nvicode` paced to `40 RPM` by default.
 
 Supported environments:
 - macOS
@@ -16,21 +17,16 @@ Install the published package:
 npm install -g nvicode
 ```
 
-Save your NVIDIA API key:
+Set up provider, key, and model:
 
-Get a free key from [NVIDIA Build API Keys](https://build.nvidia.com/settings/api-keys).
-
-```sh
-nvicode auth
-```
-
-Choose a model:
+- NVIDIA: get a free key from [NVIDIA Build API Keys](https://build.nvidia.com/settings/api-keys)
+- OpenRouter: use your OpenRouter API key
 
 ```sh
 nvicode select model
 ```
 
-Launch Claude Code through NVIDIA:
+Launch Claude Code through your selected provider:
 
 ```sh
 nvicode launch claude
@@ -46,7 +42,7 @@ nvicode launch claude
 
 ![nvicode select model](https://raw.githubusercontent.com/dineshpotla/nvicode/main/assets/screenshots/select-model.png)
 
-### Launch Claude Code through NVIDIA
+### Launch Claude Code through your selected provider
 
 ![nvicode launch claude](https://raw.githubusercontent.com/dineshpotla/nvicode/main/assets/screenshots/launch.png)
 
@@ -64,12 +60,18 @@ nvicode auth
 nvicode launch claude -p "Reply with exactly OK"
 ```
 
-The launcher starts a local proxy on `127.0.0.1:8788`, points Claude Code at it with `ANTHROPIC_BASE_URL`, and forwards requests to NVIDIA `chat/completions`.
+Provider behavior:
+- NVIDIA: starts a local proxy on `127.0.0.1:8788`, points Claude Code at it with `ANTHROPIC_BASE_URL`, and forwards requests to NVIDIA `chat/completions`.
+- OpenRouter: points Claude Code directly at `https://openrouter.ai/api` using OpenRouter credentials and Anthropic-compatible model ids.
+
+In an interactive terminal, `nvicode usage` refreshes live every 2 seconds. When piped or redirected, it prints a single snapshot.
 
-If no NVIDIA API key is saved yet, `nvicode` prompts for one on first use.
+`nvicode select model` now asks for provider, optional API key update, and model choice in one guided flow.
+If no API key is saved for the active provider yet, `nvicode` prompts for one on first use.
 By default, the proxy paces upstream NVIDIA requests at `40 RPM`. Override that with `NVICODE_MAX_RPM` if your account has a different limit.
 The usage dashboard compares your local NVIDIA run cost against Claude Opus 4.6 at `$5 / MTok input` and `$25 / MTok output`, based on Anthropic pricing as of `2026-03-30`.
 If your NVIDIA endpoint is not free, override local cost estimates with `NVICODE_INPUT_USD_PER_MTOK` and `NVICODE_OUTPUT_USD_PER_MTOK`.
+Local `usage`, `activity`, and `dashboard` commands are available for NVIDIA proxy sessions. OpenRouter sessions use OpenRouter's direct connection path instead.
 
 ## Requirements
 
@@ -92,4 +94,4 @@ npm link
 - `thinking` is disabled by default because some NVIDIA reasoning models can consume the entire output budget and return no visible answer to Claude Code.
 - The proxy supports basic text, tool calls, tool results, and token count estimation.
 - The proxy includes upstream request pacing and retries on NVIDIA `429` responses.
-- Claude Code remains the frontend; the selected NVIDIA model becomes the backend.
+- Claude Code remains the frontend; the selected provider/model becomes the backend.

package/dist/cli.js

@@ -7,18 +7,18 @@ import path from "node:path";
 import process from "node:process";
 import { spawn } from "node:child_process";
 import { fileURLToPath } from "node:url";
-import { getNvicodePaths, loadConfig, saveConfig, } from "./config.js";
+import { getActiveApiKey, getActiveModel, getNvicodePaths, loadConfig, saveConfig, } from "./config.js";
 import { createProxyServer } from "./proxy.js";
-import { CURATED_MODELS, getRecommendedModels } from "./models.js";
+import { getRecommendedModels } from "./models.js";
 import { filterRecordsSince, formatDuration, formatInteger, formatTimestamp, formatUsd, readUsageRecords, summarizeUsage, } from "./usage.js";
 const __filename = fileURLToPath(import.meta.url);
 const usage = () => {
     console.log(`nvicode
 
 Commands:
-  nvicode select model        Select and save a NVIDIA model
-  nvicode models              Show recommended coding models
-  nvicode auth                Save or update NVIDIA API key
+  nvicode select model        Guided provider, key, and model selection
+  nvicode models              Show recommended models for the active provider
+  nvicode auth                Save or update the API key for the active provider
   nvicode config              Show current nvicode config
   nvicode usage               Show token usage and cost comparison
   nvicode activity            Show recent request activity
@@ -40,6 +40,7 @@ const getPathExts = () => {
         .map((ext) => ext.toLowerCase());
 };
 const unique = (values) => [...new Set(values)];
+const getProviderLabel = (provider) => provider === "openrouter" ? "OpenRouter" : "NVIDIA";
 const question = async (prompt) => {
     const rl = createInterface({
         input: process.stdin,
@@ -52,28 +53,85 @@ const question = async (prompt) => {
         rl.close();
     }
 };
+const promptProviderSelection = async (initialProvider) => {
+    console.log("Choose a provider:");
+    console.log("1. NVIDIA");
+    console.log("   Uses the local nvicode proxy and usage dashboard.");
+    console.log("2. OpenRouter");
+    console.log("   Uses Claude Code direct Anthropic-compatible connection.");
+    const defaultChoice = initialProvider === "openrouter" ? "2" : "1";
+    const answer = (await question(`Provider selection [${defaultChoice}]: `)).toLowerCase();
+    const normalized = answer || defaultChoice;
+    if (normalized === "1" || normalized === "nvidia") {
+        return "nvidia";
+    }
+    if (normalized === "2" ||
+        normalized === "openrouter" ||
+        normalized === "open-router") {
+        return "openrouter";
+    }
+    throw new Error("Provider selection is required.");
+};
+const promptApiKeyUpdate = async (config, provider) => {
+    const providerLabel = getProviderLabel(provider);
+    const currentApiKey = provider === "openrouter" ? config.openrouterApiKey : config.nvidiaApiKey;
+    if (currentApiKey) {
+        const answer = (await question(`${providerLabel} API key already saved. Update it? [y/N]: `)).toLowerCase();
+        if (answer !== "y" && answer !== "yes") {
+            return provider === "openrouter"
+                ? { openrouterApiKey: currentApiKey, nvidiaApiKey: config.nvidiaApiKey }
+                : { nvidiaApiKey: currentApiKey, openrouterApiKey: config.openrouterApiKey };
+        }
+        const nextKey = await question(`${providerLabel} API key (press Enter or type "skip" to keep current): `);
+        if (!nextKey || nextKey.toLowerCase() === "skip") {
+            return provider === "openrouter"
+                ? { openrouterApiKey: currentApiKey, nvidiaApiKey: config.nvidiaApiKey }
+                : { nvidiaApiKey: currentApiKey, openrouterApiKey: config.openrouterApiKey };
+        }
+        return provider === "openrouter"
+            ? { openrouterApiKey: nextKey, nvidiaApiKey: config.nvidiaApiKey }
+            : { nvidiaApiKey: nextKey, openrouterApiKey: config.openrouterApiKey };
+    }
+    const nextKey = await question(`${providerLabel} API key (press Enter or type "skip" to skip): `);
+    if (!nextKey || nextKey.toLowerCase() === "skip") {
+        return {
+            nvidiaApiKey: config.nvidiaApiKey,
+            openrouterApiKey: config.openrouterApiKey,
+        };
+    }
+    return provider === "openrouter"
+        ? { openrouterApiKey: nextKey, nvidiaApiKey: config.nvidiaApiKey }
+        : { nvidiaApiKey: nextKey, openrouterApiKey: config.openrouterApiKey };
+};
 const ensureConfigured = async () => {
     let config = await loadConfig();
     let changed = false;
-    if (!config.apiKey) {
+    const providerLabel = getProviderLabel(config.provider);
+    const activeApiKey = getActiveApiKey(config);
+    const activeModel = getActiveModel(config);
+    if (!activeApiKey) {
         if (!process.stdin.isTTY) {
-            throw new Error("Missing NVIDIA API key. Run `nvicode auth` first.");
+            throw new Error(`Missing ${providerLabel} API key. Run \`nvicode auth\` first.`);
         }
-        const apiKey = await question("NVIDIA API key: ");
+        const apiKey = await question(`${providerLabel} API key: `);
         if (!apiKey) {
-            throw new Error("NVIDIA API key is required.");
+            throw new Error(`${providerLabel} API key is required.`);
         }
         config = {
             ...config,
-            apiKey,
+            ...(config.provider === "openrouter"
+                ? { openrouterApiKey: apiKey }
+                : { nvidiaApiKey: apiKey }),
         };
         changed = true;
     }
-    if (!config.model) {
-        const [first] = await getRecommendedModels(config.apiKey);
+    if (!activeModel) {
+        const [first] = await getRecommendedModels(config.provider, getActiveApiKey(config));
         config = {
             ...config,
-            model: first?.id || CURATED_MODELS[0].id,
+            ...(config.provider === "openrouter"
+                ? { openrouterModel: first?.id || "anthropic/claude-sonnet-4.6" }
+                : { nvidiaModel: first?.id || "moonshotai/kimi-k2.5" }),
         };
         changed = true;
     }
@@ -84,22 +142,28 @@ const ensureConfigured = async () => {
 };
 const runAuth = async () => {
     const config = await loadConfig();
-    const apiKey = await question(config.apiKey ? "NVIDIA API key (leave blank to keep current): " : "NVIDIA API key: ");
-    if (!apiKey && config.apiKey) {
-        console.log("Kept existing NVIDIA API key.");
+    const providerLabel = getProviderLabel(config.provider);
+    const currentApiKey = getActiveApiKey(config);
+    const apiKey = await question(currentApiKey
+        ? `${providerLabel} API key (leave blank to keep current): `
+        : `${providerLabel} API key: `);
+    if (!apiKey && currentApiKey) {
+        console.log(`Kept existing ${providerLabel} API key.`);
         return;
     }
     if (!apiKey) {
-        throw new Error("NVIDIA API key is required.");
+        throw new Error(`${providerLabel} API key is required.`);
     }
     await saveConfig({
         ...config,
-        apiKey,
+        ...(config.provider === "openrouter"
+            ? { openrouterApiKey: apiKey }
+            : { nvidiaApiKey: apiKey }),
     });
-    console.log("Saved NVIDIA API key.");
+    console.log(`Saved ${providerLabel} API key.`);
 };
-const printModels = async (apiKey) => {
-    const models = apiKey ? await getRecommendedModels(apiKey) : CURATED_MODELS;
+const printModels = async (provider, apiKey) => {
+    const models = await getRecommendedModels(provider, apiKey || "");
     models.forEach((model, index) => {
         console.log(`${index + 1}. ${model.label}`);
         console.log(`   ${model.id}`);
@@ -107,11 +171,20 @@ const printModels = async (apiKey) => {
     });
 };
 const runSelectModel = async () => {
-    const config = await ensureConfigured();
-    const models = await getRecommendedModels(config.apiKey);
-    console.log("Recommended NVIDIA coding models:");
-    await printModels(config.apiKey);
-    console.log("Type a number from the list or enter a custom model id.");
+    const config = await loadConfig();
+    const provider = await promptProviderSelection(config.provider);
+    const providerLabel = getProviderLabel(provider);
+    const keyPatch = await promptApiKeyUpdate(config, provider);
+    const nextConfig = await saveConfig({
+        ...config,
+        ...keyPatch,
+        provider,
+    });
+    const models = await getRecommendedModels(provider, getActiveApiKey(nextConfig));
+    console.log(`Top popular ${providerLabel} models:`);
+    await printModels(provider, getActiveApiKey(nextConfig));
+    console.log("Or paste a full model id.");
+    console.log("Example: qwen/qwen3.6-plus-preview:free");
     const answer = await question("Model selection: ");
     const index = Number(answer);
     const chosenModel = Number.isInteger(index) && index >= 1 && index <= models.length
@@ -121,8 +194,10 @@ const runSelectModel = async () => {
         throw new Error("Model selection is required.");
     }
     await saveConfig({
-        ...config,
-        model: chosenModel,
+        ...nextConfig,
+        ...(provider === "openrouter"
+            ? { openrouterModel: chosenModel }
+            : { nvidiaModel: chosenModel }),
     });
     console.log(`Saved model: ${chosenModel}`);
 };
@@ -132,36 +207,46 @@ const runConfig = async () => {
     console.log(`Config file: ${paths.configFile}`);
     console.log(`State dir:   ${paths.stateDir}`);
     console.log(`Usage log:   ${paths.usageLogFile}`);
-    console.log(`Model:       ${config.model}`);
+    console.log(`Provider:    ${getProviderLabel(config.provider)}`);
+    console.log(`Model:       ${getActiveModel(config)}`);
     console.log(`Proxy port:  ${config.proxyPort}`);
     console.log(`Max RPM:     ${config.maxRequestsPerMinute}`);
     console.log(`Thinking:    ${config.thinking ? "on" : "off"}`);
-    console.log(`API key:     ${config.apiKey ? "saved" : "missing"}`);
+    console.log(`NVIDIA key:  ${config.nvidiaApiKey ? "saved" : "missing"}`);
+    console.log(`OpenRouter key: ${config.openrouterApiKey ? "saved" : "missing"}`);
 };
 const printUsageBlock = (label, records) => {
     const summary = summarizeUsage(records);
     console.log(label);
     console.log(`Requests: ${formatInteger(summary.requests)} (${formatInteger(summary.successes)} ok, ${formatInteger(summary.errors)} error)`);
-    console.log(`Input tokens: ${formatInteger(summary.inputTokens)}`);
-    console.log(`Output tokens: ${formatInteger(summary.outputTokens)}`);
+    console.log(`Turn input tokens: ${formatInteger(summary.turnInputTokens)}`);
+    console.log(`Billed input tokens: ${formatInteger(summary.inputTokens)}`);
+    console.log(`Turn output tokens: ${formatInteger(summary.turnOutputTokens)}`);
+    console.log(`Billed output tokens: ${formatInteger(summary.outputTokens)}`);
     console.log(`NVIDIA cost: ${formatUsd(summary.providerCostUsd)}`);
-    console.log(`Opus 4.6 equivalent: ${formatUsd(summary.compareCostUsd)}`);
     console.log(`Estimated savings: ${formatUsd(summary.savingsUsd)}`);
 };
-const runUsage = async () => {
+const getUsageView = async () => {
     const records = await readUsageRecords();
     if (records.length === 0) {
-        console.log("No usage recorded yet.");
-        return;
+        return [
+            "nvicode usage",
+            "",
+            "No usage recorded yet.",
+            "Keep this open and new activity will appear automatically.",
+        ].join("\n");
     }
     const now = Date.now();
     const latestPricing = records[0]?.pricing;
+    const lines = ["nvicode usage", ""];
     if (latestPricing) {
-        console.log("Pricing basis:");
-        console.log(`- NVIDIA configured cost: ${formatUsd(latestPricing.providerInputUsdPerMTok)} / MTok input, ${formatUsd(latestPricing.providerOutputUsdPerMTok)} / MTok output`);
-        console.log(`- ${latestPricing.compareModel}: ${formatUsd(latestPricing.compareInputUsdPerMTok)} / MTok input, ${formatUsd(latestPricing.compareOutputUsdPerMTok)} / MTok output`);
-        console.log(`- Comparison source: ${latestPricing.comparePricingSource} (${latestPricing.comparePricingUpdatedAt})`);
-        console.log("");
+        lines.push("Pricing basis:");
+        lines.push(`- NVIDIA configured cost: ${formatUsd(latestPricing.providerInputUsdPerMTok)} / MTok input, ${formatUsd(latestPricing.providerOutputUsdPerMTok)} / MTok output`);
+        lines.push(`- ${latestPricing.compareModel}: ${formatUsd(latestPricing.compareInputUsdPerMTok)} / MTok input, ${formatUsd(latestPricing.compareOutputUsdPerMTok)} / MTok output`);
+        lines.push(`- Comparison source: ${latestPricing.comparePricingSource} (${latestPricing.comparePricingUpdatedAt})`);
+        lines.push("- In/Out columns show current-turn tokens.");
+        lines.push("- Billed In/Billed Out include the full Claude Code request context.");
+        lines.push("");
     }
     const windows = [
         { label: "Last 1 hour", durationMs: 1 * 60 * 60 * 1000 },
@@ -176,34 +261,87 @@ const runUsage = async () => {
         return {
             window: window.label,
             requests: `${formatInteger(summary.requests)} (${formatInteger(summary.successes)} ok/${formatInteger(summary.errors)} err)`,
-            inputTokens: formatInteger(summary.inputTokens),
-            outputTokens: formatInteger(summary.outputTokens),
+            inputTokens: formatInteger(summary.turnInputTokens),
+            billedInputTokens: formatInteger(summary.inputTokens),
+            outputTokens: formatInteger(summary.turnOutputTokens),
+            billedOutputTokens: formatInteger(summary.outputTokens),
             nvidiaCost: formatUsd(summary.providerCostUsd),
             savings: formatUsd(summary.savingsUsd),
         };
     });
-    console.log("Window        Requests         Input Tok  Output Tok  NVIDIA      Saved");
+    lines.push(`Snapshot: ${formatTimestamp(new Date(now).toISOString())}`);
+    lines.push("");
+    lines.push("Window        Requests         In Tok   Billed In  Out Tok  Billed Out  NVIDIA      Saved");
     rows.forEach((row) => {
-        console.log(`${row.window.padEnd(13)} ${row.requests.padEnd(16)} ${row.inputTokens.padStart(10)} ${row.outputTokens.padStart(11)} ${row.nvidiaCost.padStart(10)} ${row.savings.padStart(10)}`);
+        lines.push(`${row.window.padEnd(13)} ${row.requests.padEnd(16)} ${row.inputTokens.padStart(8)} ${row.billedInputTokens.padStart(11)} ${row.outputTokens.padStart(8)} ${row.billedOutputTokens.padStart(11)} ${row.nvidiaCost.padStart(10)} ${row.savings.padStart(10)}`);
     });
+    return lines.join("\n");
+};
+const sleep = async (ms) => new Promise((resolve) => setTimeout(resolve, ms));
+const clearTerminal = () => {
+    process.stdout.write("\x1b[2J\x1b[H");
+};
+const runUsage = async () => {
+    const config = await loadConfig();
+    if (config.provider === "openrouter") {
+        console.log("OpenRouter uses a direct Claude Code connection.");
+        console.log("Local nvicode usage stats are only available for NVIDIA proxy sessions.");
+        console.log("Use the OpenRouter activity dashboard for OpenRouter usage.");
+        return;
+    }
+    const interactive = process.stdout.isTTY && process.stdin.isTTY;
+    if (!interactive) {
+        console.log(await getUsageView());
+        return;
+    }
+    let stopped = false;
+    const stop = () => {
+        stopped = true;
+    };
+    process.on("SIGINT", stop);
+    process.on("SIGTERM", stop);
+    try {
+        while (!stopped) {
+            clearTerminal();
+            process.stdout.write(await getUsageView());
+            process.stdout.write("\n\nRefreshing every 2s. Press Ctrl+C to exit.\n");
+            await sleep(2_000);
+        }
+    }
+    finally {
+        process.off("SIGINT", stop);
+        process.off("SIGTERM", stop);
+    }
 };
 const runActivity = async () => {
+    const config = await loadConfig();
+    if (config.provider === "openrouter") {
+        console.log("OpenRouter uses a direct Claude Code connection.");
+        console.log("Local nvicode activity logs are only available for NVIDIA proxy sessions.");
+        return;
+    }
     const records = await readUsageRecords();
     if (records.length === 0) {
         console.log("No activity recorded yet.");
         return;
     }
-    console.log("Timestamp             Status  Model                           In Tok  Out Tok  Latency  NVIDIA     Saved");
+    console.log("Timestamp             Status  Model                         In Tok  Bill In  Out Tok Bill Out  Latency  NVIDIA     Saved");
     for (const record of records.slice(0, 15)) {
-        const model = record.model.length > 30 ? `${record.model.slice(0, 27)}...` : record.model;
+        const model = record.model.length > 28 ? `${record.model.slice(0, 25)}...` : record.model;
         const status = record.status === "success" ? "ok" : "error";
-        console.log(`${formatTimestamp(record.timestamp).padEnd(21)} ${status.padEnd(6)} ${model.padEnd(31)} ${formatInteger(record.inputTokens).padStart(7)} ${formatInteger(record.outputTokens).padStart(8)} ${formatDuration(record.latencyMs).padStart(8)} ${formatUsd(record.providerCostUsd).padStart(10)} ${formatUsd(record.savingsUsd).padStart(10)}`);
+        console.log(`${formatTimestamp(record.timestamp).padEnd(21)} ${status.padEnd(6)} ${model.padEnd(29)} ${formatInteger(record.turnInputTokens ?? record.visibleInputTokens ?? record.inputTokens).padStart(7)} ${formatInteger(record.inputTokens).padStart(8)} ${formatInteger(record.turnOutputTokens ?? record.visibleOutputTokens ?? record.outputTokens).padStart(8)} ${formatInteger(record.outputTokens).padStart(8)} ${formatDuration(record.latencyMs).padStart(8)} ${formatUsd(record.providerCostUsd).padStart(10)} ${formatUsd(record.savingsUsd).padStart(10)}`);
         if (record.error) {
             console.log(`  error: ${record.error}`);
         }
     }
 };
 const runDashboard = async () => {
+    const config = await loadConfig();
+    if (config.provider === "openrouter") {
+        console.log("OpenRouter uses a direct Claude Code connection.");
+        console.log("Local nvicode dashboards are only available for NVIDIA proxy sessions.");
+        return;
+    }
     const records = await readUsageRecords();
     if (records.length === 0) {
         console.log("No usage recorded yet.");
@@ -371,19 +509,39 @@ const spawnClaudeProcess = (claudeBinary, args, env) => {
 };
 const runLaunchClaude = async (args) => {
     const config = await ensureConfigured();
-    await ensureProxyRunning(config);
     const claudeBinary = await resolveClaudeBinary();
-    const child = spawnClaudeProcess(claudeBinary, args, {
-        ...process.env,
-        ANTHROPIC_BASE_URL: `http://127.0.0.1:${config.proxyPort}`,
-        ANTHROPIC_AUTH_TOKEN: config.proxyToken,
-        ANTHROPIC_API_KEY: "",
-        ANTHROPIC_MODEL: config.model,
-        CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS: "1",
-        ANTHROPIC_CUSTOM_MODEL_OPTION: config.model,
-        ANTHROPIC_CUSTOM_MODEL_OPTION_NAME: "nvicode custom model",
-        ANTHROPIC_CUSTOM_MODEL_OPTION_DESCRIPTION: "Claude Code via local NVIDIA gateway",
-    });
+    const activeModel = getActiveModel(config);
+    const activeApiKey = getActiveApiKey(config);
+    const env = config.provider === "openrouter"
+        ? {
+            ...process.env,
+            ANTHROPIC_BASE_URL: "https://openrouter.ai/api",
+            ANTHROPIC_AUTH_TOKEN: activeApiKey,
+            ANTHROPIC_API_KEY: "",
+            ANTHROPIC_MODEL: activeModel,
+            ANTHROPIC_DEFAULT_SONNET_MODEL: activeModel,
+            ANTHROPIC_DEFAULT_OPUS_MODEL: activeModel,
+            ANTHROPIC_DEFAULT_HAIKU_MODEL: activeModel,
+            CLAUDE_CODE_SUBAGENT_MODEL: activeModel,
+            CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS: "1",
+        }
+        : (() => {
+            return {
+                ...process.env,
+                ANTHROPIC_BASE_URL: `http://127.0.0.1:${config.proxyPort}`,
+                ANTHROPIC_AUTH_TOKEN: config.proxyToken,
+                ANTHROPIC_API_KEY: "",
+                ANTHROPIC_MODEL: activeModel,
+                CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS: "1",
+                ANTHROPIC_CUSTOM_MODEL_OPTION: activeModel,
+                ANTHROPIC_CUSTOM_MODEL_OPTION_NAME: "nvicode custom model",
+                ANTHROPIC_CUSTOM_MODEL_OPTION_DESCRIPTION: "Claude Code via local NVIDIA gateway",
+            };
+        })();
+    if (config.provider === "nvidia") {
+        await ensureProxyRunning(config);
+    }
+    const child = spawnClaudeProcess(claudeBinary, args, env);
     await new Promise((resolve, reject) => {
         child.on("exit", (code, signal) => {
             if (signal) {
@@ -398,12 +556,15 @@ const runLaunchClaude = async (args) => {
 };
 const runServe = async () => {
     const config = await ensureConfigured();
+    if (config.provider !== "nvidia") {
+        throw new Error("`nvicode serve` is only available for the NVIDIA provider.");
+    }
     const server = createProxyServer(config);
     await new Promise((resolve, reject) => {
         server.once("error", reject);
         server.listen(config.proxyPort, "127.0.0.1", () => resolve());
     });
-    console.error(`nvicode proxy listening on http://127.0.0.1:${config.proxyPort} using ${config.model}`);
+    console.error(`nvicode proxy listening on http://127.0.0.1:${config.proxyPort} using ${config.nvidiaModel}`);
     const shutdown = () => {
         server.close(() => process.exit(0));
     };
@@ -423,7 +584,7 @@ const main = async () => {
     }
     if (command === "models") {
         const config = await loadConfig();
-        await printModels(config.apiKey || undefined);
+        await printModels(config.provider, getActiveApiKey(config) || undefined);
         return;
     }
     if (command === "auth") {

package/dist/config.js

@@ -3,7 +3,9 @@ import { promises as fs } from "node:fs";
 import os from "node:os";
 import path from "node:path";
 const DEFAULT_PROXY_PORT = 8788;
-const DEFAULT_MODEL = "moonshotai/kimi-k2.5";
+const DEFAULT_PROVIDER = "nvidia";
+const DEFAULT_NVIDIA_MODEL = "moonshotai/kimi-k2.5";
+const DEFAULT_OPENROUTER_MODEL = "anthropic/claude-sonnet-4.6";
 const DEFAULT_MAX_REQUESTS_PER_MINUTE = 40;
 const getEnvNumber = (name) => {
     const raw = process.env[name];
@@ -54,9 +56,14 @@ export const getNvicodePaths = () => {
 };
 const withDefaults = (config) => {
     const envMaxRequestsPerMinute = getEnvNumber("NVICODE_MAX_RPM");
+    const legacyApiKey = config.apiKey?.trim() || "";
+    const legacyModel = config.model?.trim() || DEFAULT_NVIDIA_MODEL;
     return {
-        apiKey: config.apiKey?.trim() || "",
-        model: config.model?.trim() || DEFAULT_MODEL,
+        provider: config.provider === "openrouter" ? "openrouter" : DEFAULT_PROVIDER,
+        nvidiaApiKey: config.nvidiaApiKey?.trim() || legacyApiKey,
+        nvidiaModel: config.nvidiaModel?.trim() || legacyModel,
+        openrouterApiKey: config.openrouterApiKey?.trim() || "",
+        openrouterModel: config.openrouterModel?.trim() || DEFAULT_OPENROUTER_MODEL,
         proxyPort: Number.isInteger(config.proxyPort) && config.proxyPort > 0
             ? config.proxyPort
             : DEFAULT_PROXY_PORT,
@@ -97,3 +104,5 @@ export const updateConfig = async (patch) => {
         ...patch,
     });
 };
+export const getActiveApiKey = (config) => config.provider === "openrouter" ? config.openrouterApiKey : config.nvidiaApiKey;
+export const getActiveModel = (config) => config.provider === "openrouter" ? config.openrouterModel : config.nvidiaModel;

package/dist/models.js

@@ -1,4 +1,4 @@
-export const CURATED_MODELS = [
+export const NVIDIA_CURATED_MODELS = [
     {
         id: "moonshotai/kimi-k2.5",
         label: "Kimi K2.5",
@@ -30,6 +30,28 @@ export const CURATED_MODELS = [
         description: "Smaller coding-focused Qwen model.",
     },
 ];
+export const OPENROUTER_CURATED_MODELS = [
+    {
+        id: "qwen/qwen3.6-plus-preview:free",
+        label: "Qwen 3.6 Plus Preview (Free)",
+        description: "Free OpenRouter Qwen preview model.",
+    },
+    {
+        id: "anthropic/claude-sonnet-4.6",
+        label: "Claude Sonnet 4.6",
+        description: "Recommended OpenRouter model for Claude Code compatibility.",
+    },
+    {
+        id: "anthropic/claude-opus-4.6",
+        label: "Claude Opus 4.6",
+        description: "Higher-end Anthropic model through OpenRouter.",
+    },
+    {
+        id: "anthropic/claude-haiku-4.5",
+        label: "Claude Haiku 4.5",
+        description: "Faster lower-cost Anthropic model through OpenRouter.",
+    },
+];
 const MODELS_URL = "https://integrate.api.nvidia.com/v1/models";
 export const fetchAvailableModelIds = async (apiKey) => {
     const response = await fetch(MODELS_URL, {
@@ -49,13 +71,16 @@ export const fetchAvailableModelIds = async (apiKey) => {
     }
     return ids;
 };
-export const getRecommendedModels = async (apiKey) => {
+export const getRecommendedModels = async (provider, apiKey) => {
+    if (provider === "openrouter") {
+        return OPENROUTER_CURATED_MODELS;
+    }
     try {
         const available = await fetchAvailableModelIds(apiKey);
-        const curated = CURATED_MODELS.filter((model) => available.has(model.id));
-        return curated.length > 0 ? curated : CURATED_MODELS;
+        const curated = NVIDIA_CURATED_MODELS.filter((model) => available.has(model.id));
+        return curated.length > 0 ? curated : NVIDIA_CURATED_MODELS;
     }
     catch {
-        return CURATED_MODELS;
+        return NVIDIA_CURATED_MODELS;
     }
 };

package/dist/proxy.js

@@ -336,9 +336,58 @@ const estimateTokens = (payload) => {
     const raw = JSON.stringify(payload);
     return Math.max(1, Math.ceil(raw.length / 4));
 };
+const getCurrentTurnMessages = (messages) => {
+    const entries = messages ?? [];
+    for (let index = entries.length - 1; index >= 0; index -= 1) {
+        if (entries[index]?.role === "assistant") {
+            return entries.slice(index + 1);
+        }
+    }
+    return entries;
+};
+const extractPromptInput = (messages) => {
+    const parts = [];
+    for (const message of messages) {
+        if (message.role !== "user") {
+            continue;
+        }
+        if (typeof message.content === "string") {
+            if (message.content.trim().length > 0) {
+                parts.push(message.content);
+            }
+            continue;
+        }
+        for (const block of message.content) {
+            if (block.type === "text" && block.text.trim().length > 0) {
+                parts.push(block.text);
+                continue;
+            }
+            if (block.type === "image" && block.source?.data) {
+                parts.push({
+                    type: "image_url",
+                    image_url: {
+                        url: `data:${block.source.media_type || "application/octet-stream"};base64,${block.source.data}`,
+                    },
+                });
+            }
+        }
+    }
+    return parts;
+};
+const estimateTurnInputTokens = (payload) => {
+    const currentTurnMessages = getCurrentTurnMessages(payload.messages);
+    const promptInput = extractPromptInput(currentTurnMessages);
+    if (promptInput.length === 0) {
+        return 0;
+    }
+    return estimateTokens({
+        prompt: promptInput,
+    });
+};
+const estimateTurnOutputTokens = (content) => estimateTokens(content);
 const resolveTargetModel = (config, payload) => payload.model && payload.model.includes("/") && !payload.model.startsWith("claude-")
     ? payload.model
-    : config.model;
+    : config.nvidiaModel;
 const callNvidia = async (config, scheduleRequest, payload) => {
     const targetModel = resolveTargetModel(config, payload);
     const requestBody = {
@@ -374,7 +423,7 @@ const callNvidia = async (config, scheduleRequest, payload) => {
             const response = await fetch(NVIDIA_URL, {
                 method: "POST",
                 headers: {
-                    Authorization: `Bearer ${config.apiKey}`,
+                    Authorization: `Bearer ${config.nvidiaApiKey}`,
                     Accept: "application/json",
                     "Content-Type": "application/json",
                 },
@@ -412,7 +461,7 @@ export const createProxyServer = (config) => {
             if (url.pathname === "/health") {
                 sendJson(response, 200, {
                     ok: true,
-                    model: config.model,
+                    model: config.nvidiaModel,
                     port: config.proxyPort,
                     thinking: config.thinking,
                     maxRequestsPerMinute: config.maxRequestsPerMinute,
@@ -445,12 +494,14 @@ export const createProxyServer = (config) => {
                     messages: payload.messages ?? [],
                     tools: payload.tools ?? [],
                 });
+                const estimatedTurnInputTokens = estimateTurnInputTokens(payload);
                 const startedAt = Date.now();
                 const pricing = getPricingSnapshot();
                 try {
                     const { upstream } = await callNvidia(config, scheduleNvidiaRequest, payload);
                     const choice = upstream.choices?.[0];
                     const mappedContent = mapResponseContent(choice);
+                    const estimatedTurnOutputTokens = estimateTurnOutputTokens(mappedContent);
                     const anthropicResponse = {
                         id: upstream.id || `msg_${randomUUID()}`,
                         type: "message",
@@ -470,6 +521,8 @@ export const createProxyServer = (config) => {
                         model: targetModel,
                         inputTokens: anthropicResponse.usage.input_tokens,
                         outputTokens: anthropicResponse.usage.output_tokens,
+                        turnInputTokens: estimatedTurnInputTokens,
+                        turnOutputTokens: estimatedTurnOutputTokens,
                         latencyMs: Date.now() - startedAt,
                         stopReason: anthropicResponse.stop_reason,
                         pricing,
@@ -570,6 +623,8 @@ export const createProxyServer = (config) => {
                         model: targetModel,
                         inputTokens: estimatedInputTokens,
                         outputTokens: 0,
+                        turnInputTokens: estimatedTurnInputTokens,
+                        turnOutputTokens: 0,
                         latencyMs: Date.now() - startedAt,
                         error: message,
                         pricing,

package/dist/usage.js

@@ -26,7 +26,7 @@ export const getPricingSnapshot = () => ({
 });
 export const estimateCostUsd = (inputTokens, outputTokens, inputUsdPerMTok, outputUsdPerMTok) => (inputTokens / 1_000_000) * inputUsdPerMTok +
     (outputTokens / 1_000_000) * outputUsdPerMTok;
-export const buildUsageRecord = ({ id, timestamp = new Date().toISOString(), status, model, inputTokens, outputTokens, latencyMs, stopReason, error, pricing = getPricingSnapshot(), }) => {
+export const buildUsageRecord = ({ id, timestamp = new Date().toISOString(), status, model, inputTokens, outputTokens, turnInputTokens, turnOutputTokens, visibleInputTokens, visibleOutputTokens, latencyMs, stopReason, error, pricing = getPricingSnapshot(), }) => {
     const providerCostUsd = estimateCostUsd(inputTokens, outputTokens, pricing.providerInputUsdPerMTok, pricing.providerOutputUsdPerMTok);
     const compareCostUsd = estimateCostUsd(inputTokens, outputTokens, pricing.compareInputUsdPerMTok, pricing.compareOutputUsdPerMTok);
     return {
@@ -36,6 +36,18 @@ export const buildUsageRecord = ({ id, timestamp = new Date().toISOString(), sta
         model,
         inputTokens,
         outputTokens,
+        ...(turnInputTokens !== undefined
+            ? { turnInputTokens }
+            : {}),
+        ...(turnOutputTokens !== undefined
+            ? { turnOutputTokens }
+            : {}),
+        ...(visibleInputTokens !== undefined
+            ? { visibleInputTokens }
+            : {}),
+        ...(visibleOutputTokens !== undefined
+            ? { visibleOutputTokens }
+            : {}),
         latencyMs,
         providerCostUsd,
         compareCostUsd,
@@ -75,6 +87,16 @@ export const summarizeUsage = (records) => records.reduce((summary, record) => {
     summary.errors += record.status === "error" ? 1 : 0;
     summary.inputTokens += record.inputTokens;
     summary.outputTokens += record.outputTokens;
+    summary.turnInputTokens +=
+        record.turnInputTokens ??
+            record.visibleInputTokens ??
+            record.inputTokens;
+    summary.turnOutputTokens +=
+        record.turnOutputTokens ??
+            record.visibleOutputTokens ??
+            record.outputTokens;
+    summary.visibleInputTokens += record.visibleInputTokens ?? record.inputTokens;
+    summary.visibleOutputTokens += record.visibleOutputTokens ?? record.outputTokens;
     summary.providerCostUsd += record.providerCostUsd;
     summary.compareCostUsd += record.compareCostUsd;
     summary.savingsUsd += record.savingsUsd;
@@ -85,6 +107,10 @@ export const summarizeUsage = (records) => records.reduce((summary, record) => {
     errors: 0,
     inputTokens: 0,
     outputTokens: 0,
+    turnInputTokens: 0,
+    turnOutputTokens: 0,
+    visibleInputTokens: 0,
+    visibleOutputTokens: 0,
     providerCostUsd: 0,
     compareCostUsd: 0,
     savingsUsd: 0,

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "nvicode",
-  "version": "0.1.5",
-  "description": "Run Claude Code through NVIDIA-hosted models using a local Anthropic-compatible gateway.",
+  "version": "0.1.6",
+  "description": "Run Claude Code through NVIDIA-hosted models or OpenRouter using a simple CLI wrapper.",
   "author": "Dinesh Potla",
   "keywords": [
     "claude-code",