Changelog

Status: Public documentation and API-surface change log.

Use this page to understand what changed in NexoRouter documentation, setup guidance, API behavior expectations, and support workflows. Model catalog and pricing pages remain the source of truth for live model availability and prices.

What appears here

What does not appear here

• Every model catalog adjustment.
• Every upstream provider issue.
• Operational notes not relevant to public API use.
• Work planning details.
• Undocumented product experiments.

For current operational status, use Status and incidents. For current model IDs and prices, use Models and Pricing.

2026-06-29

P1/P2 coverage pass

• Added model support FAQ pages for model selection, availability, versions/aliases, and request cost.
• Added high-intent support entries for OpenAI SDK migration, max_tokens, content safety, password reset, and balance-alert boundaries.
• Added translation tools and community ecosystem integration entries as candidate or not-verified patterns instead of claiming specific client support.
• Added a candidate integration checklist as the shared verification template for tool pages.
• Added explicit candidate/unsupported status labels to llms.txt tool entries for safer AI-agent ingestion.
• Expanded OpenAI-compatible SDK examples to include Python, Node.js, Go, .NET, and the general custom-base-URL rule for other languages.
• Linked the new model and ecosystem entries from docs home and scenario index.

P0 docs depth pass

• Expanded Quickstart with a setup path, tool field mapping, environment variable guidance, first-failure triage, and next-step links.
• Expanded OpenAI-compatible Chat Completions with SDK setup, request body boundaries, wrong-vs-right provider examples, and structured failure recovery.
• Rebuilt the invalid_api_key FAQ as a full support workflow for base URL/key mismatch, headers, env vars, third-party tools, and support handoff.
• Expanded OpenCode from a short candidate setup into an end-to-end candidate guide with credential flow, config example, first test, second test, Usage Logs verification, common issues, and known limits.
• Expanded Status and incidents with a decision table, known-safe test request, support handoff template, and what not to infer.

Boundary cleanup

• Kept Codex CLI, Claude Code, Gemini CLI, Responses API, embeddings, streaming, and tool calling as unsupported or candidate/boundary topics unless explicitly documented.
• Kept llms.txt aligned with the human docs so AI agents do not infer unsupported protocols.

2026-06-28

Documentation 2.0 structure

• Added API capability pages for OpenAI-compatible chat, Models API, API keys, Usage Logs, rate limits, error handling, streaming status, embeddings status, Responses API boundary, tool calling, reasoning models, and model lifecycle.
• Split combined tool pages into per-tool guides for Cline, Roo Code, Aider, Continue, Open WebUI, Cherry Studio, Chatbox, Dify, LangChain, Vercel AI SDK, and LiteLLM.
• Added long-tail support pages for Cursor, Gemini CLI boundary, ChatGPT Next Web, LlamaIndex, status/incidents, API concurrency, billing concepts, network/proxy, server region, key usage investigation, billing review, and registration email guidance.
• Expanded FAQ coverage for model scope, key budgets, browser/CORS, streaming, tool calling, payments, data security, and account access.
• Added real public-site screenshots for onboarding, model selection, pricing, and docs navigation.

API boundary wording

• Clarified that only endpoints documented in API Reference are stable public support commitments.
• Clarified that /v1/responses, embeddings, streaming, and tool calling require explicit documentation or per-tool verification before production reliance.

How to read this changelog

Documentation changes do not automatically mean new API capability support. Always check the relevant capability page and API Reference before changing production code.