fraim-hub 2.0.204

package/dist/src/local-mcp-server/agent-token-prices.js

@@ -0,0 +1,167 @@
+"use strict";
+/**
+ * Static price table for converting per-agent token counters to USD.
+ *
+ * Issue #330 / R1.3 / C4 (ISO 27001 A.5.31).
+ *
+ * Each entry MUST carry `source` (a citation pointing back to the vendor's
+ * published pricing page) and `verifiedOn` (an ISO date string). The
+ * `agent-token-prices.test.ts` lint fails CI if any entry exceeds 180 days
+ * since `verifiedOn` so this table is reviewed at least quarterly.
+ *
+ * Prices are USD per million tokens.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.AGENT_TOKEN_PRICES = void 0;
+exports.lookupPrice = lookupPrice;
+exports.computeCostUsd = computeCostUsd;
+exports.AGENT_TOKEN_PRICES = [
+    // Anthropic — Claude (used by Claude Code).
+    // Note: when Claude Code emits cost via OTLP we read it directly; this
+    // table backstops cost computation for legacy/no-cost-metric snapshots
+    // and lets per-agent comparison panels report a model-attributed cost
+    // even when a snapshot lacks a costUsd field.
+    {
+        agent: 'claude-code',
+        model: 'claude-opus-4-7',
+        inputPerMTok: 15.00,
+        outputPerMTok: 75.00,
+        cacheReadPerMTok: 1.50,
+        cacheCreationPerMTok: 18.75,
+        source: 'https://www.anthropic.com/pricing',
+        verifiedOn: '2026-04-29',
+    },
+    {
+        agent: 'claude-code',
+        model: 'claude-sonnet-4-6',
+        inputPerMTok: 3.00,
+        outputPerMTok: 15.00,
+        cacheReadPerMTok: 0.30,
+        cacheCreationPerMTok: 3.75,
+        source: 'https://www.anthropic.com/pricing',
+        verifiedOn: '2026-04-29',
+    },
+    {
+        agent: 'claude-code',
+        model: 'claude-haiku-4-5',
+        inputPerMTok: 1.00,
+        outputPerMTok: 5.00,
+        cacheReadPerMTok: 0.10,
+        cacheCreationPerMTok: 1.25,
+        source: 'https://www.anthropic.com/pricing',
+        verifiedOn: '2026-04-29',
+    },
+    // OpenAI — GPT-5 Codex (used by Codex CLI).
+    {
+        agent: 'codex',
+        model: 'gpt-5-codex',
+        inputPerMTok: 5.00,
+        outputPerMTok: 15.00,
+        cacheReadPerMTok: 0.50,
+        cacheCreationPerMTok: 0,
+        source: 'https://openai.com/api/pricing/',
+        verifiedOn: '2026-04-29',
+    },
+    {
+        agent: 'codex',
+        model: 'gpt-5',
+        inputPerMTok: 5.00,
+        outputPerMTok: 15.00,
+        cacheReadPerMTok: 0.50,
+        cacheCreationPerMTok: 0,
+        source: 'https://openai.com/api/pricing/',
+        verifiedOn: '2026-04-29',
+    },
+    // Real Codex CLI 0.125.0 emits `gpt-5.5` on turn_context.payload.model.
+    // Treated as the gpt-5 family for pricing until OpenAI publishes a
+    // distinct rate card. Conservative estimate; refresh on next quarterly
+    // verification pass.
+    {
+        agent: 'codex',
+        model: 'gpt-5.5',
+        inputPerMTok: 5.00,
+        outputPerMTok: 15.00,
+        cacheReadPerMTok: 0.50,
+        cacheCreationPerMTok: 0,
+        source: 'https://openai.com/api/pricing/ (gpt-5 row, applied to gpt-5.5 as nearest published rate)',
+        verifiedOn: '2026-04-30',
+    },
+    // Google — Gemini CLI. Standard paid-tier text/image/video token rates.
+    // Context caching maps to cache reads; the published storage-hour price
+    // cannot be derived from the CLI token stats and is intentionally omitted.
+    {
+        agent: 'gemini',
+        model: 'gemini-3.1-flash-lite',
+        inputPerMTok: 0.25,
+        outputPerMTok: 1.50,
+        cacheReadPerMTok: 0.025,
+        cacheCreationPerMTok: 0,
+        source: 'https://ai.google.dev/gemini-api/docs/pricing',
+        verifiedOn: '2026-06-02',
+    },
+    {
+        agent: 'gemini',
+        model: 'gemini-3-flash-preview',
+        inputPerMTok: 0.50,
+        outputPerMTok: 3.00,
+        cacheReadPerMTok: 0.05,
+        cacheCreationPerMTok: 0,
+        source: 'https://ai.google.dev/gemini-api/docs/pricing',
+        verifiedOn: '2026-06-02',
+    },
+    // GitHub Copilot CLI. The CLI's agentic mode uses GPT-4.1 (also branded
+    // "gpt-4.1" in the Copilot API stream). GitHub charges for Copilot via
+    // premium-request quota rather than per-token billing on most plans; the
+    // per-token rates below apply to the Copilot API / enterprise API billing
+    // path (pay-per-token, not included-requests) as published at
+    // https://docs.github.com/en/copilot/about-github-copilot/plans-for-github-copilot
+    // and the underlying OpenAI model pricing for gpt-4.1 at openai.com/api/pricing/.
+    // Coverage is 'partial' for Copilot runs where the CLI does not emit a cost
+    // field — the table backstops cost computation when a token count is present
+    // but no direct costUsd is in the stream.
+    {
+        agent: 'copilot',
+        model: 'gpt-4.1',
+        inputPerMTok: 2.00,
+        outputPerMTok: 8.00,
+        cacheReadPerMTok: 0.50,
+        cacheCreationPerMTok: 0,
+        source: 'https://openai.com/api/pricing/ (gpt-4.1 row; GitHub Copilot CLI agentic mode)',
+        verifiedOn: '2026-06-09',
+    },
+    {
+        agent: 'copilot',
+        model: 'gpt-4o',
+        inputPerMTok: 2.50,
+        outputPerMTok: 10.00,
+        cacheReadPerMTok: 1.25,
+        cacheCreationPerMTok: 0,
+        source: 'https://openai.com/api/pricing/ (gpt-4o row; GitHub Copilot CLI fallback model)',
+        verifiedOn: '2026-06-09',
+    },
+];
+/**
+ * Look up the price entry for an agent + model. Agent is matched
+ * case-insensitively. Returns null when no matching entry exists — the
+ * adapter is then expected to skip cost computation rather than guess.
+ */
+function lookupPrice(agent, model) {
+    if (!agent || !model)
+        return null;
+    const a = agent.toLowerCase();
+    return exports.AGENT_TOKEN_PRICES.find(e => e.agent === a && e.model === model) || null;
+}
+/**
+ * Compute USD cost from cumulative token counts. Returns 0 when no price
+ * entry exists for the agent/model — the caller decides whether to drop
+ * the snapshot entirely or surface a partial-coverage warning.
+ */
+function computeCostUsd(agent, model, counts) {
+    const price = lookupPrice(agent, model);
+    if (!price)
+        return 0;
+    return ((counts.inputTokens / 1_000_000) * price.inputPerMTok +
+        (counts.outputTokens / 1_000_000) * price.outputPerMTok +
+        (counts.cacheReadTokens / 1_000_000) * price.cacheReadPerMTok +
+        (counts.cacheCreationTokens / 1_000_000) * price.cacheCreationPerMTok);
+}