codeep 2.16.0 → 2.18.0

package/dist/utils/tokenTracker.js

@@ -2,13 +2,16 @@
  * Token and cost tracking for API usage
  */
 import { AsyncLocalStorage } from 'node:async_hooks';
-// Context window sizes per model (in tokens).
-// Keep this table in lockstep with `providers.ts` — entries for models that
-// aren't in the provider catalogue only show up if a user types an id by hand
-// and produce phantom estimates against the wrong context size.
+import { isFlatFeeProvider } from '../config/providers.js';
+import { formatResourceImpactReport } from './resourceImpact.js';
+// Context window sizes per model (in tokens). Primarily mirrors providers.ts;
+// retired aliases remain only where restored historical sessions still need a
+// meaningful context/cost display.
 const MODEL_CONTEXT_WINDOWS = {
     // Z.AI / ZhipuAI
-    'glm-5.2': 200_000,
+    'glm-5.3': 1_000_000,
+    'glm-5.2': 1_000_000,
+    'glm-5.1': 200_000,
     'glm-5-turbo': 202_752,
     // OpenAI
     'gpt-5.6-sol': 1_050_000,
@@ -28,31 +31,36 @@ const MODEL_CONTEXT_WINDOWS = {
     'deepseek-v4-flash': 1_000_000,
     // Google
     'gemini-3.1-pro-preview': 1_048_576,
-    'gemini-3.5-flash': 1_000_000,
-    'gemini-3.1-flash-lite': 1_048_576,
+    'gemini-3.7-flash': 1_048_576,
+    'gemini-3.6-flash': 1_048_576,
+    'gemini-3.5-flash': 1_048_576,
+    'gemini-3.5-flash-lite': 1_048_576,
     'gemini-3-flash-preview': 1_000_000,
     // MiniMax
-    'MiniMax-M3': 524_288,
-    // Kimi (Moonshot) — 1M on the K3 line, 256K across K2.x
-    'kimi-k3-code': 1_000_000,
-    'kimi-k3-code-highspeed': 1_000_000,
-    'kimi-k3-thinking': 1_000_000,
+    'MiniMax-M3': 1_000_000,
+    // Kimi (Moonshot) — 1M on K3, 256K across K2.x
+    'kimi-k3': 1_000_000,
     'kimi-k2.7-code': 262_144,
     'kimi-k2.7-code-highspeed': 262_144,
     'kimi-k2.6': 262_144,
-    'kimi-k2.5': 262_144,
     'kimi-for-coding': 262_144,
+    'kimi-for-coding-highspeed': 262_144,
+    'k3': 1_000_000,
+    'k3-256k': 262_144,
     // Grok (xAI)
+    'grok-4.6': 500_000,
     'grok-4.5': 500_000,
     'grok-build-0.1': 256_000,
     'grok-4.3': 1_000_000,
     'grok-code-fast-1': 256_000,
     'grok-4-fast-reasoning': 2_000_000,
-    // Qwen (Alibaba) — 256K native (1M with extrapolation)
-    'qwen3-coder-plus': 262_144,
-    'qwen3-coder-next': 262_144,
-    'qwen3-coder-flash': 262_144,
+    // Qwen (Alibaba) — current hosted generation
     'qwen3.7-max': 1_000_000,
+    'qwen3.8-max-preview': 1_000_000,
+    'qwen3.7-plus': 1_000_000,
+    'qwen3.6-plus': 1_000_000,
+    'qwen3.5-plus': 1_000_000,
+    'qwen3.6-flash': 1_000_000,
     'Qwen/Qwen3-Coder-480B-A35B-Instruct': 262_144,
 };
 const DEFAULT_CONTEXT_WINDOW = 128_000;
@@ -63,20 +71,21 @@ export function getModelContextWindow(model) {
     return MODEL_CONTEXT_WINDOWS[model] ?? DEFAULT_CONTEXT_WINDOW;
 }
 // Pricing table — USD per 1M tokens. Same rule as MODEL_CONTEXT_WINDOWS:
-// only list model ids that exist in `providers.ts`, otherwise typing an id
-// by hand can produce phantom cost estimates against stale rates.
+// Primarily mirrors `providers.ts`. A few retired aliases remain so restored
+// historical sessions still show the rate that applied when they were created.
 const MODEL_PRICING = {
     // Z.AI / ZhipuAI
-    // GLM-5.2 per-token pricing isn't published yet — mirror GLM-5.1 (same tier,
-    // its successor) provisionally so /cost stays sane; update when z.ai posts it.
-    // Note: on the GLM Coding Plan (the default `z.ai` provider) billing is a flat
-    // subscription, so this only affects the pay-per-use estimate.
-    'glm-5.2': { inputPer1M: 1.00, outputPer1M: 3.20 },
+    // Coding Plan is flat-fee; these official rates apply to pay-per-use.
+    // `glm-5.3` is GLM Coding Plan only — Z.AI publishes no per-token rate for it
+    // (the standalone model API is still "coming soon"), so it stays unpriced
+    // rather than borrowing GLM-5.2's.
+    'glm-5.2': { inputPer1M: 1.40, outputPer1M: 4.40 },
+    'glm-5.1': { inputPer1M: 1.40, outputPer1M: 4.40 },
     'glm-5-turbo': { inputPer1M: 1.20, outputPer1M: 4.00 },
     // OpenAI
     'gpt-5.6-sol': { inputPer1M: 5.00, outputPer1M: 30.00 },
-    'gpt-5.6-terra': { inputPer1M: 2.50, outputPer1M: 15.00 },
-    'gpt-5.6-luna': { inputPer1M: 1.00, outputPer1M: 6.00 },
+    'gpt-5.6-terra': { inputPer1M: 2.00, outputPer1M: 12.00 },
+    'gpt-5.6-luna': { inputPer1M: 0.20, outputPer1M: 1.20 },
     'gpt-5.5': { inputPer1M: 5.00, outputPer1M: 30.00 },
     'gpt-5.4': { inputPer1M: 2.50, outputPer1M: 15.00 },
     'gpt-5.4-mini': { inputPer1M: 0.75, outputPer1M: 4.50 },
@@ -87,37 +96,48 @@ const MODEL_PRICING = {
     'claude-sonnet-5': { inputPer1M: 3.00, outputPer1M: 15.00 },
     'claude-haiku-4-5-20251001': { inputPer1M: 1.00, outputPer1M: 5.00 },
     // DeepSeek (cache-miss input pricing)
-    'deepseek-v4-pro': { inputPer1M: 1.74, outputPer1M: 3.48 },
+    'deepseek-v4-pro': { inputPer1M: 0.435, outputPer1M: 0.87 },
     'deepseek-v4-flash': { inputPer1M: 0.14, outputPer1M: 0.28 },
     // Google
+    // Gemini 3.6/3.7 Flash carry PROMOTIONAL rates that run through 2026-12-31 and
+    // step up to 1.50/7.50 on 2027-01-01 — revisit both rows on that date.
     'gemini-3.1-pro-preview': { inputPer1M: 2.00, outputPer1M: 12.00 },
+    'gemini-3.7-flash': { inputPer1M: 0.75, outputPer1M: 3.75 },
+    'gemini-3.6-flash': { inputPer1M: 0.75, outputPer1M: 3.75 },
     'gemini-3.5-flash': { inputPer1M: 1.50, outputPer1M: 9.00 },
-    'gemini-3.1-flash-lite': { inputPer1M: 0.25, outputPer1M: 1.50 },
+    'gemini-3.5-flash-lite': { inputPer1M: 0.30, outputPer1M: 2.50 },
     'gemini-3-flash-preview': { inputPer1M: 0.50, outputPer1M: 3.00 },
     // MiniMax
     'MiniMax-M3': { inputPer1M: 0.60, outputPer1M: 2.40 },
     // Kimi (Moonshot) — pay-per-use cache-miss rates; `kimi-for-coding` is the
     // subscription alias (flat-fee in reality, priced notionally like K2.7 Code).
-    'kimi-k3-code': { inputPer1M: 0.60, outputPer1M: 2.50 },
-    'kimi-k3-code-highspeed': { inputPer1M: 0.60, outputPer1M: 2.50 },
-    'kimi-k3-thinking': { inputPer1M: 0.60, outputPer1M: 2.50 },
-    'kimi-k2.7-code': { inputPer1M: 0.60, outputPer1M: 2.50 },
-    'kimi-k2.7-code-highspeed': { inputPer1M: 0.60, outputPer1M: 2.50 },
-    'kimi-k2.6': { inputPer1M: 0.55, outputPer1M: 2.20 },
-    'kimi-k2.5': { inputPer1M: 0.40, outputPer1M: 1.90 },
-    'kimi-for-coding': { inputPer1M: 0.60, outputPer1M: 2.50 },
-    // Grok (xAI)
+    'kimi-k3': { inputPer1M: 3.00, outputPer1M: 15.00 },
+    'kimi-k2.7-code': { inputPer1M: 0.95, outputPer1M: 4.00 },
+    // Kimi doesn't publish a distinct high-speed price in its main table.
+    // Leave that variant unpriced rather than presenting an invented estimate.
+    'kimi-k2.6': { inputPer1M: 0.95, outputPer1M: 4.00 },
+    'kimi-for-coding': { inputPer1M: 0.95, outputPer1M: 4.00 },
+    'kimi-for-coding-highspeed': { inputPer1M: 0.95, outputPer1M: 4.00 },
+    'k3': { inputPer1M: 3.00, outputPer1M: 15.00 },
+    'k3-256k': { inputPer1M: 3.00, outputPer1M: 15.00 },
+    // Grok (xAI) — base-tier rates. xAI doubles Grok 4.5/4.6 at prompts ≥200K;
+    // this table stores one flat rate per model, so the base tier is what we show.
+    'grok-4.6': { inputPer1M: 2.00, outputPer1M: 6.00 },
     'grok-4.5': { inputPer1M: 2.00, outputPer1M: 6.00 },
     'grok-build-0.1': { inputPer1M: 1.00, outputPer1M: 2.00 },
     'grok-4.3': { inputPer1M: 1.25, outputPer1M: 2.50 },
     'grok-code-fast-1': { inputPer1M: 0.20, outputPer1M: 1.50 },
     'grok-4-fast-reasoning': { inputPer1M: 0.20, outputPer1M: 0.50 },
-    // Qwen (Alibaba) — qwen3-coder-* 0–256K tier; the Coding Plan is flat-fee so
-    // this only affects the pay-per-use estimate.
-    'qwen3-coder-plus': { inputPer1M: 0.28, outputPer1M: 1.65 },
-    'qwen3-coder-next': { inputPer1M: 0.28, outputPer1M: 1.65 },
-    'qwen3-coder-flash': { inputPer1M: 0.10, outputPer1M: 0.50 },
+    // Qwen (Alibaba) — list prices for the first context tier. Coding Plan
+    // variants are flat-fee; these rates describe pay-per-use calls.
+    // `qwen3.8-max-preview` is Token-Plan-only (credit-metered, promotional
+    // preview rate); Alibaba publishes no pay-per-use per-token price for it.
+    // Leave it unpriced rather than borrowing the GA qwen3.8-max rate.
     'qwen3.7-max': { inputPer1M: 2.50, outputPer1M: 7.50 },
+    'qwen3.7-plus': { inputPer1M: 0.40, outputPer1M: 1.60 },
+    'qwen3.6-plus': { inputPer1M: 0.40, outputPer1M: 2.40 },
+    'qwen3.5-plus': { inputPer1M: 0.40, outputPer1M: 2.40 },
+    'qwen3.6-flash': { inputPer1M: 0.25, outputPer1M: 1.50 },
     // ModelScope free tier — no per-token charge.
     'Qwen/Qwen3-Coder-480B-A35B-Instruct': { inputPer1M: 0, outputPer1M: 0 },
 };
@@ -285,13 +305,20 @@ export function getSessionStats() {
         totalCacheCreationTokens += record.cacheCreationTokens ?? 0;
         totalCacheReadTokens += record.cacheReadTokens ?? 0;
     }
-    const estimatedCost = getCostBreakdown().reduce((s, b) => s + b.estimatedCost, 0);
+    const breakdown = getCostBreakdown();
+    const estimatedCost = breakdown.reduce((s, b) => s + b.estimatedCost, 0);
+    const billableCost = breakdown
+        .filter(b => !isFlatFeeProvider(b.provider))
+        .reduce((s, b) => s + b.estimatedCost, 0);
+    const hasFlatFeeUsage = breakdown.some(b => isFlatFeeProvider(b.provider));
     return {
         totalPromptTokens,
         totalCompletionTokens,
         totalTokens,
         requestCount: currentRecords().length,
         estimatedCost,
+        billableCost,
+        hasFlatFeeUsage,
         totalCacheCreationTokens,
         totalCacheReadTokens,
     };
@@ -342,18 +369,26 @@ export function formatCostReport() {
         return '_No API requests in this session yet._';
     }
     const breakdown = getCostBreakdown();
+    // Flat-fee providers (subscriptions / free tiers) bill nothing per token, so
+    // their notional cost is never shown and never folded into the total. A
+    // session can mix them with pay-per-use models, so decide per entry.
+    const planEntries = breakdown.filter(b => isFlatFeeProvider(b.provider));
+    const costLine = planEntries.length === breakdown.length
+        ? '**Estimated cost:** included in plan'
+        : `**Estimated cost:** $${stats.billableCost.toFixed(4)}${planEntries.length > 0 ? ' + usage included in plan' : ''}`;
     const lines = [
         '## Session Cost',
         '',
         `**Requests:** ${stats.requestCount}  ·  **Input:** ${formatTokenCount(stats.totalPromptTokens)}  ·  **Output:** ${formatTokenCount(stats.totalCompletionTokens)}  ·  **Total:** ${formatTokenCount(stats.totalTokens)}`,
-        `**Estimated cost:** $${stats.estimatedCost.toFixed(4)}`,
+        costLine,
         '',
     ];
-    if (breakdown.length > 1 || (breakdown.length === 1 && breakdown[0].estimatedCost > 0)) {
+    if (breakdown.length > 1 || (breakdown.length === 1 && (breakdown[0].estimatedCost > 0 || planEntries.length > 0))) {
         lines.push('| Provider / Model | Input | Output | Cost |');
         lines.push('|---|---:|---:|---:|');
         for (const b of breakdown) {
-            lines.push(`| \`${b.provider}\` / \`${b.model}\` | ${formatTokenCount(b.promptTokens)} | ${formatTokenCount(b.completionTokens)} | $${b.estimatedCost.toFixed(4)} |`);
+            const cost = isFlatFeeProvider(b.provider) ? 'included in plan' : `$${b.estimatedCost.toFixed(4)}`;
+            lines.push(`| \`${b.provider}\` / \`${b.model}\` | ${formatTokenCount(b.promptTokens)} | ${formatTokenCount(b.completionTokens)} | ${cost} |`);
         }
     }
     // Prompt caching summary — only shown if at least one cached call landed.
@@ -368,9 +403,10 @@ export function formatCostReport() {
             lines.push(`**Estimated savings vs no caching:** $${cache.estimatedSavingsUsd.toFixed(4)}`);
         }
     }
+    lines.push('', ...formatResourceImpactReport(stats.totalTokens));
     // Models with no pricing entry don't contribute to cost — flag so users
     // aren't surprised the total looks low.
-    const untracked = breakdown.filter(b => b.estimatedCost === 0 && (b.promptTokens + b.completionTokens) > 0);
+    const untracked = breakdown.filter(b => b.estimatedCost === 0 && (b.promptTokens + b.completionTokens) > 0 && !isFlatFeeProvider(b.provider));
     if (untracked.length > 0) {
         lines.push('', `_Note: ${untracked.length} model${untracked.length === 1 ? '' : 's'} (${untracked.map(u => `\`${u.model}\``).join(', ')}) have no pricing entry — token counts are tracked but not priced._`);
     }

package/dist/version.d.ts

@@ -1 +1 @@
-export declare const VERSION = "2.16.0";
+export declare const VERSION = "2.18.0";

package/dist/version.js

@@ -1,4 +1,4 @@
 // AUTO-GENERATED by scripts/gen-version.js — do not edit by hand.
 // Baked from package.json at build time so the bun-compiled binary reports
 // the right version (it has no package.json on disk to read at runtime).
-export const VERSION = '2.16.0';
+export const VERSION = '2.18.0';

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "codeep",
-  "version": "2.16.0",
+  "version": "2.18.0",
   "description": "AI-powered coding assistant built for the terminal. Multiple LLM providers, project-aware context, and a seamless development workflow.",
   "type": "module",
   "main": "dist/index.js",
@@ -15,6 +15,7 @@
     "test": "vitest run",
     "test:watch": "vitest",
     "test:coverage": "vitest run --coverage",
+    "version": "node scripts/gen-version.js && git add src/version.ts",
     "release": "node scripts/release.js"
   },
   "repository": {