pi-cost 0.1.0

package/lib/pi-pricing.js

@@ -0,0 +1,197 @@
+'use strict';
+
+const LITELLM_PRICING_URL =
+  'https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json';
+
+const TIERED_THRESHOLD = 200_000;
+const PRICING_REFRESH_MS = 6 * 60 * 60 * 1000;
+
+const PI_ROUTING_PROVIDERS = new Set(['dial', 'exa', 'github-copilot', 'github_copilot', 'auto']);
+
+// Strip pi-side decoration so we can match LiteLLM canonical keys.
+// Examples: claude-sonnet-4.6 -> claude-sonnet-4-5  (pi uses dotted minor)
+function piIdNormalizations(modelId) {
+  if (!modelId) return [];
+  const lower = String(modelId).toLowerCase().trim();
+  const out = new Set();
+  out.add(lower);
+  const dotToDash = lower.replace(/(\d)\.(\d)/g, '$1-$2');
+  out.add(dotToDash);
+  // claude-sonnet-4.6 -> claude-sonnet-4-5 (pi numbering, LiteLLM uses prior step)
+  const stripMinor = lower.replace(/(claude-(?:sonnet|haiku|opus)-\d+)\.\d+/g, '$1');
+  out.add(stripMinor);
+  return [...out];
+}
+
+const OFFLINE_PRICING = {
+  'anthropic/claude-sonnet-4-5': {
+    input_cost_per_token: 3e-6, output_cost_per_token: 1.5e-5,
+    cache_creation_input_token_cost: 3.75e-6, cache_read_input_token_cost: 3e-7,
+    input_cost_per_token_above_200k_tokens: 6e-6, output_cost_per_token_above_200k_tokens: 2.25e-5,
+    cache_creation_input_token_cost_above_200k_tokens: 7.5e-6, cache_read_input_token_cost_above_200k_tokens: 6e-7,
+  },
+  'anthropic/claude-haiku-4-5': {
+    input_cost_per_token: 8e-7, output_cost_per_token: 4e-6,
+    cache_creation_input_token_cost: 1e-6, cache_read_input_token_cost: 8e-8,
+  },
+  'anthropic/claude-opus-4-1': {
+    input_cost_per_token: 1.5e-5, output_cost_per_token: 7.5e-5,
+    cache_creation_input_token_cost: 1.875e-5, cache_read_input_token_cost: 1.5e-6,
+  },
+  'openai/gpt-5': { input_cost_per_token: 2.5e-6, output_cost_per_token: 1e-5, cache_read_input_token_cost: 2.5e-7 },
+  'openai/gpt-5-mini': { input_cost_per_token: 2.5e-7, output_cost_per_token: 2e-6, cache_read_input_token_cost: 2.5e-8 },
+  'openai/gpt-4': { input_cost_per_token: 3e-5, output_cost_per_token: 6e-5 },
+  'openai/gpt-4o': { input_cost_per_token: 2.5e-6, output_cost_per_token: 1e-5 },
+  'openai/gpt-4o-mini': { input_cost_per_token: 1.5e-7, output_cost_per_token: 6e-7 },
+  'gemini/gemini-2.5-pro': { input_cost_per_token: 1.25e-6, output_cost_per_token: 1e-5 },
+};
+
+let cachedPricing = null;
+let pricingFetchedAt = 0;
+let pricingSource = 'none';
+
+async function fetchPricing() {
+  const now = Date.now();
+  if (cachedPricing && (now - pricingFetchedAt) < PRICING_REFRESH_MS) return cachedPricing;
+  try {
+    const resp = await fetch(LITELLM_PRICING_URL);
+    if (!resp.ok) throw new Error(`HTTP ${resp.status}`);
+    const data = await resp.json();
+    const pricing = new Map();
+    for (const [name, info] of Object.entries(data)) {
+      if (typeof info !== 'object' || info == null) continue;
+      if (info.input_cost_per_token != null || info.output_cost_per_token != null) {
+        pricing.set(name, info);
+      }
+    }
+    cachedPricing = pricing;
+    pricingFetchedAt = now;
+    pricingSource = 'litellm';
+    console.log(`[pi-cost pricing] loaded ${pricing.size} models from LiteLLM`);
+    return pricing;
+  } catch (err) {
+    console.warn('[pi-cost pricing] fetch failed, using offline fallback:', err.message);
+    if (!cachedPricing) {
+      cachedPricing = new Map(Object.entries(OFFLINE_PRICING));
+      pricingFetchedAt = now;
+      pricingSource = 'offline';
+    }
+    return cachedPricing;
+  }
+}
+
+function refresh() {
+  cachedPricing = null;
+  pricingFetchedAt = 0;
+  pricingSource = 'none';
+}
+
+function getMeta() {
+  return {
+    fetchedAt: pricingFetchedAt ? new Date(pricingFetchedAt).toISOString() : null,
+    source: pricingSource,
+    modelCount: cachedPricing ? cachedPricing.size : 0,
+  };
+}
+
+function lookupExact(pricing, key) {
+  if (!key) return null;
+  const direct = pricing.get(key);
+  if (direct) return { key, info: direct };
+  return null;
+}
+
+function resolveKey(pricing, provider, modelId) {
+  if (!pricing || !modelId) return null;
+  const candidates = piIdNormalizations(modelId);
+  const providerLower = provider ? String(provider).toLowerCase() : null;
+  const isRouting = providerLower ? PI_ROUTING_PROVIDERS.has(providerLower) : false;
+
+  // 1. Exact match on every normalized id.
+  for (const c of candidates) {
+    const hit = lookupExact(pricing, c);
+    if (hit) return hit;
+  }
+
+  // 2. provider/model when the provider is a real upstream (not a pi billing route).
+  if (providerLower && !isRouting) {
+    for (const c of candidates) {
+      const hit = lookupExact(pricing, `${providerLower}/${c}`);
+      if (hit) return hit;
+    }
+  }
+
+  // 3. Anthropic / openai / gemini prefixes for unrouted pi model IDs.
+  const knownPrefixes = ['anthropic/', 'openai/', 'gemini/', 'mistral/', 'groq/'];
+  for (const c of candidates) {
+    for (const p of knownPrefixes) {
+      const hit = lookupExact(pricing, `${p}${c}`);
+      if (hit) return hit;
+    }
+  }
+
+  // 4. Fuzzy: longest LiteLLM key whose tail-segment matches any candidate.
+  let best = null;
+  let bestScore = 0;
+  for (const c of candidates) {
+    for (const [key, info] of pricing) {
+      const kLower = key.toLowerCase();
+      const tail = kLower.split('/').pop();
+      // exact tail match wins
+      if (tail === c && c.length > bestScore) {
+        best = { key, info }; bestScore = c.length + 10;
+        continue;
+      }
+      if (tail && (tail.includes(c) || c.includes(tail))) {
+        const score = Math.min(tail.length, c.length);
+        if (score > bestScore) { best = { key, info }; bestScore = score; }
+      }
+    }
+  }
+  return best;
+}
+
+function calculateTieredCost(tokens, basePrice, tieredPrice, threshold = TIERED_THRESHOLD) {
+  if (!tokens || tokens <= 0) return 0;
+  if (tokens > threshold && tieredPrice != null) {
+    const below = Math.min(tokens, threshold);
+    const above = Math.max(0, tokens - threshold);
+    let cost = above * tieredPrice;
+    if (basePrice != null) cost += below * basePrice;
+    return cost;
+  }
+  if (basePrice != null) return tokens * basePrice;
+  return 0;
+}
+
+function computeCost(info, usage) {
+  const input = calculateTieredCost(
+    usage.input || 0, info.input_cost_per_token,
+    info.input_cost_per_token_above_200k_tokens);
+  const output = calculateTieredCost(
+    usage.output || 0, info.output_cost_per_token,
+    info.output_cost_per_token_above_200k_tokens);
+  const cacheWrite = calculateTieredCost(
+    usage.cacheWrite || 0, info.cache_creation_input_token_cost,
+    info.cache_creation_input_token_cost_above_200k_tokens);
+  const cacheRead = calculateTieredCost(
+    usage.cacheRead || 0, info.cache_read_input_token_cost,
+    info.cache_read_input_token_cost_above_200k_tokens);
+  return input + output + cacheWrite + cacheRead;
+}
+
+// Pi-shaped usage: { input, output, cacheRead, cacheWrite, cost: { total } }
+// Returns { cost, source, modelKey }
+function estimate(pricing, provider, modelId, usage) {
+  if (!usage) return { cost: 0, source: 'none', modelKey: null };
+  const jsonlTotal = usage.cost && typeof usage.cost.total === 'number' ? usage.cost.total : 0;
+  if (jsonlTotal > 0) {
+    return { cost: jsonlTotal, source: 'jsonl', modelKey: null };
+  }
+  const match = resolveKey(pricing, provider, modelId);
+  if (!match) return { cost: 0, source: 'none', modelKey: null };
+  const cost = computeCost(match.info, usage);
+  return { cost, source: 'pricing', modelKey: match.key };
+}
+
+module.exports = { fetchPricing, refresh, getMeta, estimate, resolveKey };

package/package.json

@@ -0,0 +1,61 @@
+{
+  "name": "pi-cost",
+  "version": "0.1.0",
+  "description": "Cost dashboard for the pi coding agent",
+  "keywords": [
+    "pi-package",
+    "pi-extension",
+    "cost",
+    "dashboard",
+    "litellm",
+    "observability"
+  ],
+  "license": "MIT",
+  "author": "nikiforovall",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/NikiforovAll/pi-cost.git"
+  },
+  "bugs": {
+    "url": "https://github.com/NikiforovAll/pi-cost/issues"
+  },
+  "homepage": "https://nikiforovall.github.io/pi-cost/",
+  "bin": { "pi-cost": "./server.js" },
+  "main": "server.js",
+  "files": [
+    "server.js",
+    "lib/",
+    "public/",
+    "extensions/",
+    "themes/",
+    "README.md"
+  ],
+  "scripts": {
+    "start": "node server.js",
+    "dev": "node server.js --open",
+    "docs:dev": "vitepress dev docs",
+    "docs:build": "vitepress build docs",
+    "docs:preview": "vitepress preview docs"
+  },
+  "dependencies": {
+    "express": "^4.19.2",
+    "open": "^10.1.0"
+  },
+  "devDependencies": {
+    "vitepress": "^1.6.4"
+  },
+  "peerDependencies": {
+    "@earendil-works/pi-coding-agent": "*"
+  },
+  "peerDependenciesMeta": {
+    "@earendil-works/pi-coding-agent": { "optional": true }
+  },
+  "pi": {
+    "extensions": ["./extensions"],
+    "image": "https://raw.githubusercontent.com/NikiforovAll/pi-cost/main/assets/overview.png",
+    "video": "https://raw.githubusercontent.com/NikiforovAll/pi-cost/main/assets/demo.mp4"
+  },
+  "engines": {
+    "node": ">=20.0.0"
+  }
+}