VicBilibily/GCMP

What is GCMP?

GCMP (short for Global Chinese Model Provider) is a VS Code extension that lets you pick and use a large variety of Chinese‑origin large‑language‑model (LLM) services – and also OpenAI‑compatible ones – directly from the editor’s Copilot Chat UI. It registers each provider as a selectable model, handles API‑key entry, shows usage/balance in the status bar, and lets you configure which model should be used for different assistant tasks (e.g., commit‑message generation, inline chat, code‑exploration agents, vision‑based analysis, etc.).

---

Who might find it useful?

• Developers in China (or anyone needing a model that runs on Chinese cloud platforms) who want to avoid the cost limits of GitHub Copilot’s built‑in model.
• Teams that need multiple providers – you can switch between ZhipuAI, MiniMax, DeepSeek, Alibaba Bailian, Tencent Hunyuan, Baidu Qianfan, and many others without leaving VS Code.
• People who want fine‑grained control over which model does cheap utility work (title generation, git‑commit messages) versus heavy‑weight coding assistance.
• Anyone who wants to keep an eye on usage – the extension shows remaining quota or balance for each provider right in the status bar.

---

Core features (as described in the README)

Feature	What it does
Provider integration	Pre‑bundled support for 15+ Chinese LLM providers (ZhipuAI, MiniMax, Moonshot, DeepSeek, Alibaba Bailian, Volcano Ark, Tencent Hunyuan, Xiaomi MiMo, Baidu Qianfan, StepFun, Ant Ling, XunFei Astron, LongCat) plus OpenAI/Anthropic‑compatible APIs.
Model selection UI	In the Copilot Chat panel you can open Manage Models → pick a provider, enter its API key, and enable the model for the current session.
Task‑specific model defaults	Settings let you assign different models for:

• chat.utilityModel – general utilities (title, summary, code‑explain, search)
• chat.utilitySmallModel – lightweight tasks (commit messages, branch names)
• inlineChat.defaultModel – inline editor chat
• Various Agent models for explore/plan/ask/implement workflows
• gcmp.commit.model – dedicated commit‑message generator
• gcmp.vision.model – image‑input capable vision model | | Automatic fallback handling | If you haven’t configured a utility model, VS Code will fall back to the built‑in Copilot model, which may consume your monthly free quota. GCMP encourages you to point those tasks to a cheaper local provider. | | Usage / balance display | The status‑bar token‑consumption icon shows remaining quota or account balance for each provider, with colour warnings when the balance gets low. | | Search integration | Several providers expose a #providerWebSearch command that triggers their own web‑search APIs (MCP or standard per‑call billing). | | Retry & rate‑limit controls | Global and per‑provider settings let you enable automatic retries (up to a configurable number of attempts) and define RPM/RPS/TPM/parallel limits to stay within provider quotas. | | Proxy & TLS support | Global proxy (gcmp.proxy) and per‑machine overrides let you route traffic through corporate proxies, with optional system‑certificate loading for self‑signed roots. | | Debug / HAR capture | Optional HAR logging of HTTP traffic (with sensitive data redacted) for troubleshooting. | | CLI‑based OAuth providers | Supports third‑party CLI tools (OpenAI Codex, xAI Grok Build) that use OAuth; the extension can call those CLIs to obtain tokens, though it warns about possible terms‑of‑service risks. |

---

How to get started (quick‑start steps)

1. Install the extension from the VS Code Marketplace (vicanent.gcmp).
2. Open the GitHub Copilot Chat panel, click Manage Models, and select a provider from the list.
3. On first use, you’ll be prompted for the provider’s API key – paste it and save.
4. Choose the newly added model in the model selector and start chatting.
5. (Optional) Use the status‑bar token icon → Set utility models to configure the utilityModel and utilitySmallModel settings in one visual panel.

---

When might you not need GCMP?

• If you only ever use OpenAI’s models and are happy with the default Copilot quota, the extra configuration isn’t necessary.
• Projects that require heavy GPU‑based fine‑tuning or training are outside the scope of this extension – GCMP is purely a runtime inference integration.

---

Bottom line

GCMP is a fairly comprehensive bridge between VS Code’s Copilot Chat experience and the rapidly growing ecosystem of Chinese LLM services. By letting you pick the cheapest or most capable model for each coding‑assistant task and surfacing usage data, it helps developers stay within budget while still getting AI‑driven code help.