Never stop coding. Free MIT AI gateway: one endpoint, 375 providers (150+ free), 1,400+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 900+ contributors
TypeScript a2aai-agentsai-gatewayanthropicclaudeclaude-codeclinecodexcopilotcursordeepseekfree-aigeminikimillm-gatewaymcpopenaiopenai-proxyqwentoken-saver
fix(tooling): strip empty arrays from WebSearch and fix Codex prompt body propagation (#1852, #1853)
Antigravity Assistantc766aa540369
diff --git a/README.md b/README.mdindex 74e2f571bc..efb8578db3 100644--- a/README.md+++ b/README.md@@ -8,6 +8,16 @@ _The most complete open-source AI proxy — **one endpoint**, **160+ providers****Chat Completions • Responses API • Embeddings • Image Generation • Video • Music • Audio Speech/Transcription • Reranking • Moderations • Web Search • MCP Server • A2A Protocol • 4,600+ Tests • 100% TypeScript**+<br/>++<a href="https://agentrouter.org/register?aff=70LM">+ <img src="https://img.shields.io/badge/🎁_GET_$100_FREE_AI_CREDITS-AgentRouter_(no_card_needed)-ff6600?style=for-the-badge&labelColor=1a1a2e&logo=data:image/svg+xml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNCAyNCIgZmlsbD0id2hpdGUiPjxwYXRoIGQ9Ik0xMiAyTDIgNy41bDEwIDUuNSAxMC01LjVMMTIgMnpNMiAxNy41bDEwIDUuNSAxMC01LjVMMTIgMTIgMiAxNy41eiIvPjwvc3ZnPg==" alt="Get $100 Free AI Credits" height="40"/>+</a>++<sub>🔥 <b>Limited offer:</b> Sign up at <a href="https://agentrouter.org/register?aff=70LM"><b>AgentRouter</b></a> and get <b>$100 in free AI credits</b></br> Access GPT-5, Claude, Gemini, DeepSeek & 100+ models. No credit card required. <a href="https://agentrouter.org/register?aff=70LM"><b>Claim your credits →</b></a></sub>++<br/>+<a href="https://trendshift.io/repositories/23589" target="_blank"><img src="https://trendshift.io/api/badge/repositories/23589" alt="diegosouzapw%2FOmniRoute | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>[🚀 Quick Start](#-quick-start) • [💡 Features](#-key-features) • [🗜️ Compression](#%EF%B8%8F-prompt-compression--save-15-75-tokens-automatically) • [💰 Pricing](#-pricing-at-a-glance) • [🎯 Use Cases](#-use-cases--ready-made-combo-playbooks) • [🌍 Proxy](#-bypass-geographic-blocks--use-ai-from-any-country) • [❓ FAQ](#-frequently-asked-questions) • [📖 Docs](#-documentation) • [💬 WhatsApp](https://chat.whatsapp.com/JI7cDQ1GyaiDHhVBpLxf8b?mode=gi_t)@@ -852,19 +862,19 @@ No proxy? Use the built-in **1proxy** integration for **hundreds of free, valida> Combine all free providers into one unbreakable combo — OmniRoute auto-routes between them when quota runs out.-| Provider | Prefix | Free Models | Quota |-| ----------------- | ----------- | ------------------------------------------------------------- | ----------------- |-| **Kiro** | `kr/` | Claude Sonnet 4.5, Haiku 4.5, Opus 4.6 | ♾️ Unlimited |-| **Qoder** | `if/` | kimi-k2-thinking, qwen3-coder-plus, deepseek-r1, minimax-m2.1 | ♾️ Unlimited |-| **Qwen** | `qw/` | qwen3-coder-plus, qwen3-coder-flash, qwen3-coder-next | ♾️ Unlimited |-| **Pollinations** | `pol/` | GPT-5, Claude, Gemini, DeepSeek, Llama 4, Mistral | No key needed |-| **LongCat** | `lc/` | LongCat-Flash-Lite | 50M tokens/day 🔥 |-| **Gemini CLI** | `gc/` | gemini-3-flash, gemini-2.5-pro | 180K tok/mo |-| **Cloudflare AI** | `cf/` | 50+ models (Llama, Gemma, Mistral, Whisper) | 10K Neurons/day |-| **Groq** | `groq/` | Llama 3.3 70B, Qwen3 32B, Kimi K2 | 14.4K RPD |-| **NVIDIA NIM** | `nvidia/` | 129 models (DeepSeek, Llama, GLM, Kimi) | ~40 RPM |-| **Cerebras** | `cerebras/` | Qwen3 235B, GPT-OSS 120B, Llama 3.1 | 1M tok/day |-| **Scaleway** | `scw/` | Qwen3 235B, Llama 70B, DeepSeek V3 | 1M tokens (EU) |+| Provider | Prefix | Free Models | Quota |+| ----------------- | ----------- | ------------------------------------------------------------- | -------------------- |+| **Kiro** | `kr/` | Claude Sonnet 4.5, Haiku 4.5, Opus 4.6 | 50 CREDITS per month |+| **Qoder** | `if/` | kimi-k2-thinking, qwen3-coder-plus, deepseek-r1, minimax-m2.1 | ♾️ Unlimited |+| **Qwen** | `qw/` | qwen3-coder-plus, qwen3-coder-flash, qwen3-coder-next | ♾️ Unlimited |+| **Pollinations** | `pol/` | GPT-5, Claude, Gemini, DeepSeek, Llama 4, Mistral | No key needed |+| **LongCat** | `lc/` | LongCat-Flash-Lite | 50M tokens/day 🔥 |+| **Gemini CLI** | `gc/` | gemini-3-flash, gemini-2.5-pro | 180K tok/mo |+| **Cloudflare AI** | `cf/` | 50+ models (Llama, Gemma, Mistral, Whisper) | 10K Neurons/day |+| **Groq** | `groq/` | Llama 3.3 70B, Qwen3 32B, Kimi K2 | 14.4K RPD |+| **NVIDIA NIM** | `nvidia/` | 129 models (DeepSeek, Llama, GLM, Kimi) | ~40 RPM |+| **Cerebras** | `cerebras/` | Qwen3 235B, GPT-OSS 120B, Llama 3.1 | 1M tok/day |+| **Scaleway** | `scw/` | Qwen3 235B, Llama 70B, DeepSeek V3 | 1M tokens (EU) |<details><summary><b>📖 25+ more free providers — Groq, Cerebras, Mistral, GitHub Models, OpenRouter, and more</b></summary>diff --git a/open-sse/executors/codex.ts b/open-sse/executors/codex.tsindex 5bc413ba96..e84be095b6 100644--- a/open-sse/executors/codex.ts+++ b/open-sse/executors/codex.ts@@ -1182,6 +1182,24 @@ export class CodexExecutor extends BaseExecutor {body.service_tier = requestDefaults.serviceTier;}+ // Issue #1832 & #1853: Map messages to input for clients like Cursor 5.5 that use responses/compact but send messages instead of input.+ // This MUST run before convertSystemToDeveloperRole and stripStoredItemReferences.+ if (!body.input && Array.isArray(body.messages)) {+ body.input = body.messages.map((msg: any) => ({+ type: "message",+ role: typeof msg.role === "string" ? msg.role : "user",+ content:+ typeof msg.content === "string"+ ? [{ type: "input_text", text: msg.content }]+ : Array.isArray(msg.content)+ ? msg.content.map((c: any) => {+ if (c && c.type === "text") return { type: "input_text", text: c.text };+ return c;+ })+ : [],+ }));+ }+// ── Cache-aware system prompt handling (both paths) ──//// Convert system → developer role IN-PLACE so system prompts remain in the@@ -1251,23 +1269,6 @@ export class CodexExecutor extends BaseExecutor {// so any references to previous response items would cause 404 errors.stripStoredItemReferences(body);- // Issue #1832: Map messages to input for clients like Cursor 5.5 that use responses/compact but send messages instead of input- if (!body.input && Array.isArray(body.messages)) {- body.input = body.messages.map((msg: any) => ({- type: "message",- role: typeof msg.role === "string" ? msg.role : "user",- content:- typeof msg.content === "string"- ? [{ type: "input_text", text: msg.content }]- : Array.isArray(msg.content)- ? msg.content.map((c: any) => {- if (c && c.type === "text") return { type: "input_text", text: c.text };- return c;- })- : [],- }));- }-// Issue #806: Even for native passthrough, some clients (purist completions) might indiscriminately inject// a `messages` or `prompt` array which the strict Codex Responses schema rejects.delete body.messages;diff --git a/open-sse/handlers/responseTranslator.ts b/open-sse/handlers/responseTranslator.tsindex 603e7c64da..c85bf34ea2 100644--- a/open-sse/handlers/responseTranslator.ts+++ b/open-sse/handlers/responseTranslator.ts@@ -123,10 +123,17 @@ export function translateNonStreamingResponse(toString(itemObj.call_id) ||toString(itemObj.id) ||`call_${Date.now()}_${toolCalls.length}`;+ let argsToEmit = itemObj.arguments;+ if (argsToEmit != null && typeof argsToEmit === "object" && !Array.isArray(argsToEmit)) {+ const cleaned = { ...argsToEmit };+ for (const [k, v] of Object.entries(cleaned)) {+ if (v === "" || (Array.isArray(v) && v.length === 0)) delete cleaned[k];+ }+ argsToEmit = cleaned;+ }+const fnArgs =- typeof itemObj.arguments === "string"- ? itemObj.arguments- : JSON.stringify(itemObj.arguments || {});+ typeof argsToEmit === "string" ? argsToEmit : JSON.stringify(argsToEmit || {});const rawName = toString(itemObj.name);// Strip Claude OAuth proxy_ prefix using toolNameMapconst resolvedName = toolNameMap?.get(rawName) ?? rawName;diff --git a/open-sse/transformer/responsesTransformer.ts b/open-sse/transformer/responsesTransformer.tsindex 3006a69322..93e913b46f 100644--- a/open-sse/transformer/responsesTransformer.ts+++ b/open-sse/transformer/responsesTransformer.ts@@ -223,13 +223,13 @@ export function createResponsesApiTransformStream(logger = null) {if (callId && !state.funcItemDone[idx]) {let args = state.funcArgsBuf[idx] || "{}";- // Fix #1674: Final cleanup of empty string placeholders that might have been split across delta chunks+ // Fix #1674 & #1852: Final cleanup of empty string and empty array placeholderstry {const parsed = JSON.parse(args);if (parsed && typeof parsed === "object" && !Array.isArray(parsed)) {let modified = false;for (const [k, v] of Object.entries(parsed)) {- if (v === "") {+ if (v === "" || (Array.isArray(v) && v.length === 0)) {delete parsed[k];modified = true;}@@ -506,12 +506,13 @@ export function createResponsesApiTransformStream(logger = null) {const refCallId = state.funcCallIds[tcIdx] || newCallId;let deltaStr = tc.function.arguments;- // Fix #1674: cx/gpt-5.5 injects empty strings for optional parameters.- // We strip these directly from the streaming deltas to avoid breaking strict clients like Claude Code.- if (deltaStr.includes('""')) {+ // Fix #1674 & #1852: Strip empty strings and empty arrays from streaming deltas+ if (deltaStr.includes('""') || deltaStr.includes("[]") || deltaStr.includes("[ ]")) {deltaStr = deltaStr.replace(/,"[a-zA-Z0-9_]+":""/g, "")- .replace(/"[a-zA-Z0-9_]+":"",/g, "");+ .replace(/"[a-zA-Z0-9_]+":"",/g, "")+ .replace(/,"[a-zA-Z0-9_]+":\s*\[\s*\]/g, "")+ .replace(/"[a-zA-Z0-9_]+":\s*\[\s*\],?/g, "");}if (refCallId) {diff --git a/open-sse/translator/response/openai-responses.ts b/open-sse/translator/response/openai-responses.tsindex 1ab95d06e4..77289dfb71 100644--- a/open-sse/translator/response/openai-responses.ts+++ b/open-sse/translator/response/openai-responses.ts@@ -633,10 +633,10 @@ export function openaiResponsesToOpenAIResponse(chunk, state) {let argsToEmit = item.arguments;if (argsToEmit != null && typeof argsToEmit === "object" && !Array.isArray(argsToEmit)) {- // Fix #1674: Strip empty string placeholders emitted by GPT-5.5 for optional fields+ // Fix #1674 & #1852: Strip empty string and array placeholders emitted by GPT-5.5 for optional fieldsconst cleaned = { ...argsToEmit };for (const [k, v] of Object.entries(cleaned)) {- if (v === "") delete cleaned[k];+ if (v === "" || (Array.isArray(v) && v.length === 0)) delete cleaned[k];}argsToEmit = cleaned;}@@ -685,7 +685,7 @@ export function openaiResponsesToOpenAIResponse(chunk, state) {if (argsToEmit != null && typeof argsToEmit === "object" && !Array.isArray(argsToEmit)) {const cleaned = { ...argsToEmit };for (const [k, v] of Object.entries(cleaned)) {- if (v === "") delete cleaned[k];+ if (v === "" || (Array.isArray(v) && v.length === 0)) delete cleaned[k];}argsToEmit = cleaned;}diff --git a/open-sse/translator/response/openai-to-claude.ts b/open-sse/translator/response/openai-to-claude.tsindex f3e8ed81ec..801594f01e 100644--- a/open-sse/translator/response/openai-to-claude.ts+++ b/open-sse/translator/response/openai-to-claude.ts@@ -184,10 +184,21 @@ export function openaiToClaudeResponse(chunk, state) {if (tc.function?.arguments) {const toolInfo = state.toolCalls.get(idx);if (toolInfo) {+ let deltaStr = tc.function.arguments;++ // Fix #1852: Strip empty string and array placeholders from streaming tool arguments+ if (deltaStr.includes('""') || deltaStr.includes("[]") || deltaStr.includes("[ ]")) {+ deltaStr = deltaStr+ .replace(/,"[a-zA-Z0-9_]+":""/g, "")+ .replace(/"[a-zA-Z0-9_]+":"",/g, "")+ .replace(/,"[a-zA-Z0-9_]+":\s*\[\s*\]/g, "")+ .replace(/"[a-zA-Z0-9_]+":\s*\[\s*\],?/g, "");+ }+results.push({type: "content_block_delta",index: toolInfo.blockIndex,- delta: { type: "input_json_delta", partial_json: tc.function.arguments },+ delta: { type: "input_json_delta", partial_json: deltaStr },});}}