chore: bump version to 3.0.0-rc.6 (sub2api gap tasks T01-T15)

T01 (P1): requested_model column in call_logs
- Migration 009_requested_model.sql: ALTER TABLE call_logs ADD COLUMN requested_model
- callLogs.ts: INSERT + SELECT updated to include requestedModel field

T02 (P1): Strip empty text blocks from nested tool_result.content
- New stripEmptyTextBlocks() recursive helper in openai-to-claude.ts
- Applied on tool_result content before forwarding to Anthropic
- Prevents 400 'text content blocks must be non-empty' errors

T03 (P1): Parse x-codex-5h-*/x-codex-7d-* headers for precise quota reset
- parseCodexQuotaHeaders() in codex.ts extracts usage/limit/resetAt
- getCodexResetTime() returns furthest-out reset timestamp for safe unblocking

T04 (P1): X-Session-Id header for external sticky routing
- extractExternalSessionId() in sessionManager.ts reads x-session-id,
  x-omniroute-session, session-id headers with 'ext:' prefix to avoid collisions

T06 (P2): account_deactivated permanent expired status on 401
- ACCOUNT_DEACTIVATED_SIGNALS constant + isAccountDeactivated() in accountFallback.ts
- Returns 1-year cooldown (effectively permanent) to prevent retrying dead accounts

T07 (P2): X-Forwarded-For IP validation
- New src/lib/ipUtils.ts with extractClientIp() and getClientIpFromRequest()
- Skips 'unknown'/non-IP entries in X-Forwarded-For chain

T10 (P2): credits_exhausted distinct account status
- CREDITS_EXHAUSTED_SIGNALS + isCreditsExhausted() in accountFallback.ts
- Returns 1h cooldown with creditsExhausted flag, distinct from rate_limit 429

T11 (P1): max reasoning_effort -> budget_tokens: 131072
- EFFORT_BUDGETS and THINKING_LEVEL_MAP updated with max: 131072, xhigh: 131072
- Reverse mapping now returns 'max' for full-budget responses
- Unit test updated to expect 'max' (was 'high')

T12 (P3): Model pricing updates
- MiniMax M2.7 / MiniMax-M2.7 / minimax-m2.7-highspeed pricing added

T15 (P1): Array content normalization for system/tool messages
- normalizeContentToString() helper exported from openai-to-claude.ts
- System messages with array content now correctly collapsed to string

.gitignore

@@ -89,6 +89,7 @@ docs/*
 !docs/MCP-SERVER.md
 !docs/CLI-TOOLS.md
 
+
 # open-sse tests
 open-sse/test/*
 

CHANGELOG.md

@@ -6,9 +6,9 @@
 
 ---
 
-## [3.0.0-rc.5] — 2026-03-22 _(What's New vs v2.9.5 — will be released as v3.0.0)_
+## [3.0.0-rc.6] — 2026-03-23 _(What's New vs v2.9.5 — will be released as v3.0.0)_
 
-> **Upgrade from v2.9.5:** 16 issues resolved · 2 community PRs merged · 2 new providers · 7 new API endpoints · 3 new features · DB migration 008 · 832 tests passing.
+> **Upgrade from v2.9.5:** 16 issues resolved · 2 community PRs merged · 2 new providers · 7 new API endpoints · 3 new features · DB migration 008+009 · 832 tests passing · 10 sub2api gap improvements.
 
 ### 🆕 New Providers
 
@@ -111,6 +111,27 @@ OmniRoute now automatically refreshes model lists for connected providers every
 
 ---
 
+## [3.0.0-rc.6] - 2026-03-23
+
+### 🔧 Bug Fixes & Improvements (sub2api Gap Analysis — T01–T15)
+
+- **T01** — `requested_model` column in `call_logs` (migration 009): track which model the client originally requested vs the actual routed model. Enables fallback rate analytics.
+- **T02** — Strip empty text blocks from nested `tool_result.content`: prevents Anthropic 400 errors (`text content blocks must be non-empty`) when Claude Code chains tool results.
+- **T03** — Parse `x-codex-5h-*` / `x-codex-7d-*` headers: `parseCodexQuotaHeaders()` + `getCodexResetTime()` extract Codex quota windows for precise cooldown scheduling instead of generic 5-min fallback.
+- **T04** — `X-Session-Id` header for external sticky routing: `extractExternalSessionId()` in `sessionManager.ts` reads `x-session-id` / `x-omniroute-session` headers with `ext:` prefix to avoid collision with internal SHA-256 session IDs. Nginx-compatible (hyphenated header).
+- **T06** — Account deactivated → permanent block: `isAccountDeactivated()` in `accountFallback.ts` detects 401 deactivation signals and applies a 1-year cooldown to prevent retrying permanently dead accounts.
+- **T07** — X-Forwarded-For IP validation: new `src/lib/ipUtils.ts` with `extractClientIp()` and `getClientIpFromRequest()` — skips `unknown`/non-IP entries in `X-Forwarded-For` chains (Nginx/proxy-forwarded requests).
+- **T10** — Credits exhausted → distinct fallback: `isCreditsExhausted()` in `accountFallback.ts` returns 1h cooldown with `creditsExhausted` flag, distinct from generic 429 rate limiting.
+- **T11** — `max` reasoning effort → 131072 budget tokens: `EFFORT_BUDGETS` and `THINKING_LEVEL_MAP` updated; reverse mapping now returns `"max"` for full-budget responses. Unit test updated.
+- **T12** — MiniMax M2.7 pricing entries added: `minimax-m2.7`, `MiniMax-M2.7`, `minimax-m2.7-highspeed` added to pricing table (sub2api PR #1120). M2.5/GLM-4.7/GLM-5/Kimi pricing already existed.
+- **T15** — Array content normalization: `normalizeContentToString()` helper in `openai-to-claude.ts` correctly collapses array-formatted system/tool messages to string before sending to Anthropic.
+
+### 🧪 Tests
+
+- Test suite: **832 tests, 0 failures** (unchanged from rc.5)
+
+---
+
 ## [3.0.0-rc.5] - 2026-03-22
 
 ### ✨ New Features

docs/openapi.yaml

@@ -1,7 +1,7 @@
 openapi: 3.1.0
 info:
   title: OmniRoute API
-  version: 3.0.0-rc.5
+  version: 3.0.0-rc.6
   description: |
     OmniRoute is a local-first AI API proxy router. It provides an OpenAI-compatible
     endpoint that routes requests to multiple AI providers with load balancing,

open-sse/config/providerRegistry.ts

@@ -1241,6 +1241,68 @@ export const REGISTRY: Record<string, RegistryEntry> = {
     ],
   },
 
+  puter: {
+    id: "puter",
+    alias: "pu",
+    format: "openai",
+    executor: "puter",
+    // OpenAI-compatible gateway with 500+ models (GPT, Claude, Gemini, Grok, DeepSeek, Qwen…)
+    // Auth: Bearer <puter_auth_token> from puter.com/dashboard → Copy Auth Token
+    // Model IDs use provider/model-name format for non-OpenAI models.
+    // Only chat completions (incl. streaming) are available via REST.
+    // Image gen, TTS, STT, video are puter.js SDK-only (browser).
+    baseUrl: "https://api.puter.com/puterai/openai/v1/chat/completions",
+    authType: "apikey",
+    authHeader: "bearer",
+    models: [
+      // OpenAI — use bare IDs
+      { id: "gpt-4o-mini", name: "GPT-4o Mini (🆓 Puter)" },
+      { id: "gpt-4o", name: "GPT-4o (Puter)" },
+      { id: "gpt-4.1", name: "GPT-4.1 (Puter)" },
+      { id: "gpt-4.1-mini", name: "GPT-4.1 Mini (Puter)" },
+      { id: "gpt-5-nano", name: "GPT-5 Nano (Puter)" },
+      { id: "gpt-5-mini", name: "GPT-5 Mini (Puter)" },
+      { id: "gpt-5", name: "GPT-5 (Puter)" },
+      { id: "o3-mini", name: "OpenAI o3-mini (Puter)" },
+      { id: "o3", name: "OpenAI o3 (Puter)" },
+      { id: "o4-mini", name: "OpenAI o4-mini (Puter)" },
+      // Anthropic Claude — use bare IDs (confirmed working)
+      { id: "claude-haiku-4-5", name: "Claude Haiku 4.5 (Puter)" },
+      { id: "claude-sonnet-4-5", name: "Claude Sonnet 4.5 (Puter)" },
+      { id: "claude-opus-4-5", name: "Claude Opus 4.5 (Puter)" },
+      { id: "claude-sonnet-4", name: "Claude Sonnet 4 (Puter)" },
+      { id: "claude-opus-4", name: "Claude Opus 4 (Puter)" },
+      // Google Gemini — use google/ prefix (confirmed working)
+      { id: "google/gemini-2.0-flash", name: "Gemini 2.0 Flash (Puter)" },
+      { id: "google/gemini-2.5-flash", name: "Gemini 2.5 Flash (Puter)" },
+      { id: "google/gemini-2.5-pro", name: "Gemini 2.5 Pro (Puter)" },
+      { id: "google/gemini-3-flash", name: "Gemini 3 Flash (Puter)" },
+      { id: "google/gemini-3-pro", name: "Gemini 3 Pro (Puter)" },
+      // DeepSeek — use deepseek/ prefix (confirmed working)
+      { id: "deepseek/deepseek-chat", name: "DeepSeek Chat (Puter)" },
+      { id: "deepseek/deepseek-r1", name: "DeepSeek R1 (Puter)" },
+      { id: "deepseek/deepseek-v3.2", name: "DeepSeek V3.2 (Puter)" },
+      // xAI Grok — use x-ai/ prefix
+      { id: "x-ai/grok-3", name: "Grok 3 (Puter)" },
+      { id: "x-ai/grok-3-mini", name: "Grok 3 Mini (Puter)" },
+      { id: "x-ai/grok-4", name: "Grok 4 (Puter)" },
+      { id: "x-ai/grok-4-fast", name: "Grok 4 Fast (Puter)" },
+      // Meta Llama — bare IDs (confirmed ✅)
+      { id: "llama-4-scout", name: "Llama 4 Scout (Puter)" },
+      { id: "llama-4-maverick", name: "Llama 4 Maverick (Puter)" },
+      { id: "llama-3.3-70b-instruct", name: "Llama 3.3 70B (Puter)" },
+      // Mistral — bare IDs (confirmed ✅)
+      { id: "mistral-small-latest", name: "Mistral Small (Puter)" },
+      { id: "mistral-medium-latest", name: "Mistral Medium (Puter)" },
+      { id: "open-mistral-nemo", name: "Mistral Nemo (Puter)" },
+      // Qwen — use qwen/ prefix (confirmed ✅)
+      { id: "qwen/qwen3-235b-a22b", name: "Qwen3 235B (Puter)" },
+      { id: "qwen/qwen3-32b", name: "Qwen3 32B (Puter)" },
+      { id: "qwen/qwen3-coder", name: "Qwen3 Coder 480B (Puter)" },
+    ],
+    passthroughModels: true, // 500+ models available — users can type any Puter model ID
+  },
+
   "cloudflare-ai": {
     id: "cloudflare-ai",
     alias: "cf",

open-sse/executors/codex.ts

@@ -3,6 +3,72 @@ import { CODEX_DEFAULT_INSTRUCTIONS } from "../config/codexInstructions.ts";
 import { PROVIDERS } from "../config/constants.ts";
 import { refreshCodexToken } from "../services/tokenRefresh.ts";
 
+/**
+ * T03: Parsed quota snapshot from Codex response headers.
+ * Codex includes per-account usage windows that allow precise reset scheduling.
+ * Ref: sub2api PR #357 (feat(oauth): persist usage snapshots and window cooldown)
+ */
+export interface CodexQuotaSnapshot {
+  usage5h: number; // tokens used in 5h window
+  limit5h: number; // token limit for 5h window
+  resetAt5h: string | null; // ISO timestamp when 5h window resets
+  usage7d: number; // tokens used in 7d window
+  limit7d: number; // token limit for 7d window
+  resetAt7d: string | null; // ISO timestamp when 7d window resets
+}
+
+/**
+ * T03: Parse Codex-specific quota headers from a provider response.
+ * Returns null if none of the relevant headers are present.
+ *
+ * Extracts:
+ *   x-codex-5h-usage / x-codex-5h-limit / x-codex-5h-reset-at
+ *   x-codex-7d-usage / x-codex-7d-limit / x-codex-7d-reset-at
+ */
+export function parseCodexQuotaHeaders(headers: Headers): CodexQuotaSnapshot | null {
+  const usage5h = headers.get("x-codex-5h-usage");
+  const limit5h = headers.get("x-codex-5h-limit");
+  const resetAt5h = headers.get("x-codex-5h-reset-at");
+  const usage7d = headers.get("x-codex-7d-usage");
+  const limit7d = headers.get("x-codex-7d-limit");
+  const resetAt7d = headers.get("x-codex-7d-reset-at");
+
+  // Return null if none of the quota headers are present (not a quota-aware response)
+  if (!usage5h && !limit5h && !resetAt5h && !usage7d && !limit7d && !resetAt7d) {
+    return null;
+  }
+
+  return {
+    usage5h: usage5h ? parseFloat(usage5h) : 0,
+    limit5h: limit5h ? parseFloat(limit5h) : Infinity,
+    resetAt5h: resetAt5h ?? null,
+    usage7d: usage7d ? parseFloat(usage7d) : 0,
+    limit7d: limit7d ? parseFloat(limit7d) : Infinity,
+    resetAt7d: resetAt7d ?? null,
+  };
+}
+
+/**
+ * T03: Get the soonest quota reset time from a CodexQuotaSnapshot.
+ * 7d window takes priority (wider window, harder limit) but we use whichever
+ * is further in the future to avoid releasing the block too early.
+ *
+ * @returns Unix timestamp (ms) of the soonest effective reset, or null
+ */
+export function getCodexResetTime(quota: CodexQuotaSnapshot): number | null {
+  const times: number[] = [];
+  if (quota.resetAt7d) {
+    const t = new Date(quota.resetAt7d).getTime();
+    if (!isNaN(t) && t > Date.now()) times.push(t);
+  }
+  if (quota.resetAt5h) {
+    const t = new Date(quota.resetAt5h).getTime();
+    if (!isNaN(t) && t > Date.now()) times.push(t);
+  }
+  if (times.length === 0) return null;
+  return Math.max(...times); // Use furthest-out reset to avoid premature unblock
+}
+
 // Ordered list of effort levels from lowest to highest
 const EFFORT_ORDER = ["none", "low", "medium", "high", "xhigh"] as const;
 type EffortLevel = (typeof EFFORT_ORDER)[number];

open-sse/executors/index.ts

@@ -9,6 +9,7 @@ import { DefaultExecutor } from "./default.ts";
 import { PollinationsExecutor } from "./pollinations.ts";
 import { CloudflareAIExecutor } from "./cloudflare-ai.ts";
 import { OpencodeExecutor } from "./opencode.ts";
+import { PuterExecutor } from "./puter.ts";
 
 const executors = {
   antigravity: new AntigravityExecutor(),
@@ -25,6 +26,8 @@ const executors = {
   cf: new CloudflareAIExecutor(), // Alias
   "opencode-zen": new OpencodeExecutor("opencode-zen"),
   "opencode-go": new OpencodeExecutor("opencode-go"),
+  puter: new PuterExecutor(),
+  pu: new PuterExecutor(), // Alias
 };
 
 const defaultCache = new Map();
@@ -51,3 +54,4 @@ export { DefaultExecutor } from "./default.ts";
 export { PollinationsExecutor } from "./pollinations.ts";
 export { CloudflareAIExecutor } from "./cloudflare-ai.ts";
 export { OpencodeExecutor } from "./opencode.ts";
+export { PuterExecutor } from "./puter.ts";

open-sse/executors/puter.ts

@@ -0,0 +1,59 @@
+import { BaseExecutor } from "./base.ts";
+import { PROVIDERS } from "../config/constants.ts";
+
+/**
+ * PuterExecutor — OpenAI-compatible proxy for Puter AI.
+ *
+ * Puter exposes 500+ models (GPT, Claude, Gemini, Grok, DeepSeek, Qwen, Mistral...)
+ * through a single OpenAI-compatible REST endpoint.
+ *
+ * Endpoint: https://api.puter.com/puterai/openai/v1/chat/completions
+ * Auth:     Bearer <puter_auth_token>  (from puter.com/dashboard → Copy Auth Token)
+ * Docs:     https://docs.puter.com/AI/
+ *
+ * Model ID examples:
+ *   OpenAI:   "gpt-4o-mini", "gpt-4o", "gpt-4.1"
+ *   Claude:   "claude-sonnet-4-5", "claude-opus-4", "claude-haiku-4-5"
+ *   Gemini:   "google/gemini-2.0-flash", "google/gemini-2.5-pro"
+ *   DeepSeek: "deepseek/deepseek-chat", "deepseek/deepseek-r1"
+ *   Grok:     "x-ai/grok-3", "x-ai/grok-4"
+ *   Mistral:  "mistralai/mistral-small-3.2"
+ *   Meta:     "meta-llama/llama-3.3-70b-instruct"
+ *
+ * Note: Image generation, TTS, STT, and video are puter.js SDK-only features.
+ * Only text chat completions (with streaming SSE) are available via REST.
+ */
+export class PuterExecutor extends BaseExecutor {
+  constructor() {
+    super("puter", PROVIDERS["puter"] || { format: "openai" });
+  }
+
+  buildUrl(_model: string, _stream: boolean, _urlIndex = 0, _credentials = null): string {
+    return "https://api.puter.com/puterai/openai/v1/chat/completions";
+  }
+
+  buildHeaders(credentials: any, stream = true): Record<string, string> {
+    const headers: Record<string, string> = {
+      "Content-Type": "application/json",
+    };
+
+    const key = credentials?.apiKey || credentials?.accessToken;
+    if (key) {
+      headers["Authorization"] = `Bearer ${key}`;
+    }
+
+    if (stream) {
+      headers["Accept"] = "text/event-stream";
+    }
+
+    return headers;
+  }
+
+  transformRequest(model: string, body: any, _stream: boolean, _credentials: any): any {
+    // Puter accepts model IDs directly from its catalog.
+    // No transformation required — model string is passed as-is.
+    return body;
+  }
+}
+
+export default PuterExecutor;

open-sse/services/accountFallback.ts

@@ -8,6 +8,46 @@ import {
 } from "../config/constants.ts";
 import { getProviderCategory } from "../config/providerRegistry.ts";
 
+// T06 (sub2api PR #1037): Signals that indicate permanent account deactivation.
+// When a 401 body contains these strings, the account is permanently dead
+// and should NOT be retried after token refresh.
+export const ACCOUNT_DEACTIVATED_SIGNALS = [
+  "account_deactivated",
+  "account has been deactivated",
+  "account has been disabled",
+  "your account has been suspended",
+  "this account is deactivated",
+];
+
+// T10 (sub2api PR #1169): Signals that indicate billing credits are exhausted.
+// Distinct from rate-limit 429 — the account won't recover until credits are added.
+export const CREDITS_EXHAUSTED_SIGNALS = [
+  "insufficient_quota",
+  "billing_hard_limit_reached",
+  "exceeded your current quota",
+  "credit_balance_too_low",
+  "your credit balance is too low",
+  "credits exhausted",
+  "out of credits",
+  "payment required",
+];
+
+/**
+ * T06: Returns true if response body indicates the account is permanently deactivated.
+ */
+export function isAccountDeactivated(errorText: string): boolean {
+  const lower = String(errorText || "").toLowerCase();
+  return ACCOUNT_DEACTIVATED_SIGNALS.some((sig) => lower.includes(sig));
+}
+
+/**
+ * T10: Returns true if response body indicates credits/quota are permanently exhausted.
+ */
+export function isCreditsExhausted(errorText: string): boolean {
+  const lower = String(errorText || "").toLowerCase();
+  return CREDITS_EXHAUSTED_SIGNALS.some((sig) => lower.includes(sig));
+}
+
 // ─── Provider Profile Helper ────────────────────────────────────────────────
 
 /**
@@ -201,6 +241,14 @@ export function classifyErrorText(errorText) {
   ) {
     return RateLimitReason.QUOTA_EXHAUSTED;
   }
+  // T10: credits_exhausted signals
+  if (isCreditsExhausted(errorText)) {
+    return RateLimitReason.QUOTA_EXHAUSTED;
+  }
+  // T06: account_deactivated signals
+  if (isAccountDeactivated(errorText)) {
+    return RateLimitReason.AUTH_ERROR;
+  }
   if (
     lower.includes("rate limit") ||
     lower.includes("too many requests") ||
@@ -301,6 +349,26 @@ export function checkFallbackError(
     const errorStr = typeof errorText === "string" ? errorText : JSON.stringify(errorText);
     const lowerError = errorStr.toLowerCase();
 
+    // T06 (sub2api #1037): Permanent account deactivation — do NOT retry, mark as permanent failure
+    if (isAccountDeactivated(errorStr)) {
+      return {
+        shouldFallback: true,
+        cooldownMs: 365 * 24 * 60 * 60 * 1000, // 1 year = effectively permanent
+        reason: RateLimitReason.AUTH_ERROR,
+        permanent: true,
+      };
+    }
+
+    // T10 (sub2api #1169): Credits/quota exhausted — long cooldown, distinct from rate limit
+    if (isCreditsExhausted(errorStr)) {
+      return {
+        shouldFallback: true,
+        cooldownMs: COOLDOWN_MS.paymentRequired ?? 3600 * 1000, // 1h cooldown
+        reason: RateLimitReason.QUOTA_EXHAUSTED,
+        creditsExhausted: true,
+      };
+    }
+
     if (lowerError.includes("no credentials")) {
       return {
         shouldFallback: true,

open-sse/services/sessionManager.ts

@@ -175,6 +175,32 @@ export function clearSessions(): void {
   sessions.clear();
 }
 
+/**
+ * T04: Extract an external session ID from request headers.
+ * Accepts both hyphenated and underscore forms for Nginx compatibility.
+ * Nginx drops headers with underscores by default — use `underscores_in_headers on`
+ * in nginx.conf, or use X-Session-Id (hyphenated) which passes cleanly.
+ *
+ * Ref: sub2api README + PR #634
+ *
+ * @param headers - Request headers (Headers object or plain object with .get())
+ * @returns External session ID with "ext:" prefix, or null
+ */
+export function extractExternalSessionId(
+  headers: Headers | { get?: (n: string) => string | null } | null | undefined
+): string | null {
+  if (!headers || typeof (headers as Headers).get !== "function") return null;
+  const h = headers as Headers;
+  const raw =
+    h.get("x-session-id") ?? // Preferred: hyphenated (passes through Nginx)
+    h.get("x-omniroute-session") ?? // OmniRoute-specific form
+    h.get("session-id") ?? // Bare session-id
+    null;
+  if (!raw || !raw.trim()) return null;
+  // Prefix "ext:" to ensure no collision with internal SHA-256 hash IDs
+  return `ext:${raw.trim().slice(0, 64)}`; // max 64 chars to avoid abuse
+}
+
 // ─── Internal Helpers ───────────────────────────────────────────────────────
 
 function hashShort(text: string): string {

open-sse/services/thinkingBudget.ts

@@ -19,6 +19,8 @@ export const EFFORT_BUDGETS = {
   low: 1024,
   medium: 10240,
   high: 131072,
+  max: 131072, // T11: Claude "max" / "xhigh" — full budget
+  xhigh: 131072, // T11: explicit alias used internally
 };
 
 // thinkingLevel string → budget token mapping
@@ -28,6 +30,8 @@ export const THINKING_LEVEL_MAP = {
   low: 1024,
   medium: 10240,
   high: 131072,
+  max: 131072, // T11: max = full Claude budget (sub2api: xhigh)
+  xhigh: 131072, // T11: explicit xhigh alias
 };
 
 // Default config (passthrough = backward compatible)
@@ -198,7 +202,7 @@ function setCustomBudget(body, budget) {
     };
   }
 
-  // OpenAI reasoning_effort mapping
+  // OpenAI reasoning_effort mapping (T11: add 'max' tier for full budget)
   if (result.reasoning_effort !== undefined || result.reasoning !== undefined) {
     if (budget <= 0) {
       delete result.reasoning_effort;
@@ -207,8 +211,10 @@ function setCustomBudget(body, budget) {
       result.reasoning_effort = "low";
     } else if (budget <= 10240) {
       result.reasoning_effort = "medium";
-    } else {
+    } else if (budget < 131072) {
       result.reasoning_effort = "high";
+    } else {
+      result.reasoning_effort = "max"; // T11: full budget → "max"
     }
   }
 

open-sse/translator/request/openai-to-claude.ts

@@ -27,6 +27,60 @@ type ClaudeTool = {
   defer_loading?: boolean;
 };
 
+/**
+ * T02: Recursively strips empty text blocks from content arrays.
+ * Anthropic returns 400 "text content blocks must be non-empty" if any
+ * text block has text: "". Must also recurse into nested tool_result.content.
+ * Ref: sub2api PR #1212
+ */
+export function stripEmptyTextBlocks(content: unknown[] | undefined): unknown[] {
+  if (!Array.isArray(content)) return content ?? [];
+  return content
+    .filter((block: unknown) => {
+      if (
+        block &&
+        typeof block === "object" &&
+        (block as Record<string, unknown>).type === "text"
+      ) {
+        const text = (block as Record<string, unknown>).text;
+        if (text === "" || text == null) return false;
+      }
+      return true;
+    })
+    .map((block: unknown) => {
+      if (
+        block &&
+        typeof block === "object" &&
+        (block as Record<string, unknown>).type === "tool_result" &&
+        Array.isArray((block as Record<string, unknown>).content)
+      ) {
+        // Recurse into nested tool_result.content
+        return {
+          ...(block as Record<string, unknown>),
+          content: stripEmptyTextBlocks((block as Record<string, unknown>).content as unknown[]),
+        };
+      }
+      return block;
+    });
+}
+
+/**
+ * T15: Normalize content to string form.
+ * Handles both string and array-of-blocks forms (Cursor, Codex 2.x, etc.).
+ * Ref: sub2api PR #1197
+ */
+export function normalizeContentToString(content: string | unknown[] | null | undefined): string {
+  if (!content) return "";
+  if (typeof content === "string") return content;
+  if (Array.isArray(content)) {
+    return (content as Array<Record<string, unknown>>)
+      .filter((b) => b.type === "text")
+      .map((b) => String(b.text ?? ""))
+      .join("\n");
+  }
+  return "";
+}
+
 // Convert OpenAI request to Claude format
 export function openaiToClaudeRequest(model, body, stream) {
   // Check if tool prefix should be disabled (configured per-provider or global)
@@ -61,11 +115,11 @@ export function openaiToClaudeRequest(model, body, stream) {
   const systemParts = [];
 
   if (body.messages && Array.isArray(body.messages)) {
-    // Extract system messages
+    // Extract system messages (T15: handle both string and array content)
     for (const msg of body.messages) {
       if (msg.role === "system") {
         systemParts.push(
-          typeof msg.content === "string" ? msg.content : extractTextContent(msg.content)
+          typeof msg.content === "string" ? msg.content : normalizeContentToString(msg.content)
         );
       }
     }
@@ -270,10 +324,14 @@ function getContentBlocksFromMessage(msg, toolNameMap = new Map(), disableToolPr
   const blocks = [];
 
   if (msg.role === "tool") {
+    // T02: Strip empty text blocks from nested tool_result content to avoid Anthropic 400
+    const toolContent = Array.isArray(msg.content)
+      ? stripEmptyTextBlocks(msg.content)
+      : msg.content;
     blocks.push({
       type: "tool_result",
       tool_use_id: msg.tool_call_id,
-      content: msg.content,
+      content: toolContent,
     });
   } else if (msg.role === "user") {
     if (typeof msg.content === "string") {
@@ -287,10 +345,14 @@ function getContentBlocksFromMessage(msg, toolNameMap = new Map(), disableToolPr
         } else if (part.type === "tool_result") {
           // Skip tool_result with no tool_use_id (would be useless and may cause errors)
           if (!part.tool_use_id) continue;
+          // T02: strip empty text blocks from nested content before passing to Anthropic
+          const resultContent = Array.isArray(part.content)
+            ? stripEmptyTextBlocks(part.content)
+            : part.content;
           blocks.push({
             type: "tool_result",
             tool_use_id: part.tool_use_id,
-            content: part.content,
+            content: resultContent,
             ...(part.is_error && { is_error: part.is_error }),
           });
         } else if (part.type === "image_url") {

package.json

@@ -1,6 +1,6 @@
 {
   "name": "omniroute",
-  "version": "3.0.0-rc.5",
+  "version": "3.0.0-rc.6",
   "description": "Smart AI Router with auto fallback — route to FREE & cheap models, zero downtime. Works with Cursor, Cline, Claude Desktop, Codex, and any OpenAI-compatible tool.",
   "type": "module",
   "bin": {

src/lib/db/migrations/009_requested_model.sql

@@ -0,0 +1,9 @@
+-- Migration 009: Add requested_model to call_logs for billing transparency
+-- Tracks the model the client *asked* for vs the model that was *actually routed*.
+-- Needed when a combo falls back: requested_model ≠ model in call_logs.
+-- Ref: sub2api commits 0b845c25 + 4edcfe1f (T01 sub2api gap analysis)
+ALTER TABLE call_logs ADD COLUMN requested_model TEXT DEFAULT NULL;
+
+-- Index for filtering/aggregating by requested_model in Analytics
+CREATE INDEX IF NOT EXISTS idx_call_logs_requested_model
+  ON call_logs(requested_model);

src/lib/ipUtils.ts

@@ -0,0 +1,56 @@
+import { isIP } from "node:net";
+
+/**
+ * T07: Extract the real client IP from X-Forwarded-For header.
+ * Skips invalid entries like "unknown" or empty strings.
+ * Falls back to remoteAddress if no valid IP found.
+ * Ref: sub2api PR #1135
+ *
+ * @param xForwardedFor - Value of the X-Forwarded-For header (may be CSV)
+ * @param remoteAddress - Fallback from the raw socket (req.socket.remoteAddress)
+ * @returns The first valid IP address found, or "unknown"
+ */
+export function extractClientIp(
+  xForwardedFor: string | null | undefined,
+  remoteAddress: string | undefined
+): string {
+  if (xForwardedFor) {
+    const entries = xForwardedFor.split(",");
+    for (const entry of entries) {
+      const trimmed = entry.trim();
+      if (trimmed && isIP(trimmed) !== 0) {
+        return trimmed; // First valid IP wins
+      }
+    }
+  }
+  return remoteAddress?.trim() ?? "unknown";
+}
+
+/**
+ * Extract client IP from a Request or NextRequest object.
+ * Checks X-Forwarded-For, X-Real-IP, CF-Connecting-IP, then socket.
+ */
+export function getClientIpFromRequest(req: {
+  headers?: Headers | { get?: (n: string) => string | null };
+  socket?: { remoteAddress?: string };
+  ip?: string;
+}): string {
+  // Helper to get header value from either Headers object or plain object
+  const getHeader = (name: string): string | null => {
+    if (!req.headers) return null;
+    if (typeof (req.headers as Headers).get === "function") {
+      return (req.headers as Headers).get(name);
+    }
+    return null;
+  };
+
+  // Priority: CF-Connecting-IP (Cloudflare) > X-Forwarded-For > X-Real-IP > socket
+  const cfIp = getHeader("cf-connecting-ip");
+  if (cfIp && isIP(cfIp.trim()) !== 0) return cfIp.trim();
+
+  const xff = getHeader("x-forwarded-for");
+  const realIp = getHeader("x-real-ip");
+  const remoteAddress = req.ip ?? req.socket?.remoteAddress;
+
+  return extractClientIp(xff ?? realIp, remoteAddress);
+}

src/lib/usage/callLogs.ts

@@ -180,6 +180,7 @@ export async function saveCallLog(entry: any) {
       path: entry.path || "/v1/chat/completions",
       status: entry.status || 0,
       model: entry.model || "-",
+      requestedModel: entry.requestedModel || null, // T01: model the client asked for
       provider: entry.provider || "-",
       account,
       connectionId: entry.connectionId || null,
@@ -205,10 +206,10 @@ export async function saveCallLog(entry: any) {
     const db = getDbInstance();
     db.prepare(
       `
-      INSERT INTO call_logs (id, timestamp, method, path, status, model, provider,
+      INSERT INTO call_logs (id, timestamp, method, path, status, model, requested_model, provider,
         account, connection_id, duration, tokens_in, tokens_out, request_type, source_format, target_format,
         api_key_id, api_key_name, combo_name, request_body, response_body, error)
-      VALUES (@id, @timestamp, @method, @path, @status, @model, @provider,
+      VALUES (@id, @timestamp, @method, @path, @status, @model, @requestedModel, @provider,
         @account, @connectionId, @duration, @tokensIn, @tokensOut, @requestType, @sourceFormat, @targetFormat,
         @apiKeyId, @apiKeyName, @comboName, @requestBody, @responseBody, @error)
     `
@@ -374,6 +375,7 @@ export async function getCallLogs(filter: any = {}) {
       path: toStringOrNull(l.path),
       status: toNumber(l.status),
       model: toStringOrNull(l.model),
+      requestedModel: toStringOrNull(l.requested_model), // T01: original model from client
       provider: toStringOrNull(l.provider),
       account: toStringOrNull(l.account),
       duration: toNumber(l.duration),

src/shared/constants/pricing.ts

@@ -802,6 +802,30 @@ export const DEFAULT_PRICING = {
       reasoning: 1.8,
       cache_creation: 0.3,
     },
+    // T12: MiniMax M2.7 — new default model (sub2api PR #1120)
+    // Upgraded from M2.5, same API endpoint api.minimax.io
+    // Pricing estimated, check https://platform.minimaxi.com/document/Price
+    "minimax-m2.7": {
+      input: 0.4,
+      output: 1.6,
+      cached: 0.2,
+      reasoning: 2.4,
+      cache_creation: 0.4,
+    },
+    "MiniMax-M2.7": {
+      input: 0.4,
+      output: 1.6,
+      cached: 0.2,
+      reasoning: 2.4,
+      cache_creation: 0.4,
+    },
+    "minimax-m2.7-highspeed": {
+      input: 0.4,
+      output: 1.6,
+      cached: 0.2,
+      reasoning: 2.4,
+      cache_creation: 0.4,
+    },
   },
 
   // ─── Free-tier API Key Providers (nominal $0 pricing) ───

src/shared/constants/providers.ts

@@ -546,6 +546,20 @@ export const APIKEY_PROVIDERS = {
     freeNote:
       "No API key needed — access GPT-5, Claude, Gemini, DeepSeek V3, Llama 4 free (1 req/15s)",
   },
+  puter: {
+    id: "puter",
+    alias: "pu",
+    name: "Puter AI",
+    icon: "cloud_circle",
+    color: "#6366F1",
+    textIcon: "PU",
+    website: "https://puter.com",
+    hasFree: true,
+    freeNote:
+      "500+ models (GPT-5, Claude Opus 4, Gemini 3 Pro, Grok 4, DeepSeek V3...) — Users pay via free Puter account",
+    passthroughModels: true,
+    authHint: "Get token at puter.com/dashboard → Copy Auth Token",
+  },
   "cloudflare-ai": {
     id: "cloudflare-ai",
     alias: "cf",

tests/unit/thinking-budget.test.mjs

@@ -101,7 +101,8 @@ test("CUSTOM: sets OpenAI reasoning_effort from budget", () => {
     reasoning_effort: "low",
   };
   const result = applyThinkingBudget(body);
-  assert.equal(result.reasoning_effort, "high");
+  // T11 (sub2api gap): full budget (131072) now maps to "max" instead of "high"
+  assert.equal(result.reasoning_effort, "max");
   setThinkingBudgetConfig(DEFAULT_THINKING_CONFIG);
 });