feat(models): add OpenCode Go provider

[DaweiTian]

Summary

Add a new OpenCode Go model provider for Dify.

OpenCode Go recently introduced stricter client requirements for its gateway. This plugin implements them natively so Dify apps can call Go models without being treated as a generic/abusive client.

OpenCode Go API client requirements → how this plugin addresses them

OpenCode Go requirement	Implementation
Identify the client with a dedicated User-Agent (not a generic SDK/HTTP library name)	Sends User-Agent: dify-opencode-go-plugin/0.1.0 on every request
Send a stable session id per conversation in x-opencode-session for routing & prompt-cache affinity	Sends x-opencode-session: dify-opencode-go/<client-id>/<dify-user> — namespaced, collision-resistant, stable per Dify user (avoids bare ids like est-user / short numeric ids that can collide across workspaces)
Typical coding-agent traffic against the Go OpenAI-compatible endpoint	Uses POST {base}/chat/completions with default https://opencode.ai/zen/go/v1

Reference: OpenCode Go — Where can I use it?

What else is included

• Predefined models from the Go catalog: GLM-5.x, Kimi K2/K3, DeepSeek V4, MiMo-V2.5, MiniMax M3/M2.x, Qwen3.x, LongCat-2.0, Hy3/Hy4, and more
• Custom model configuration for newly added model IDs
• Streaming and non-streaming completions
• Tool / function calling on supported models
• Official OpenCode logo assets (light/dark)

Release Notes

• Add OpenCode Go as a new LLM model provider
• Comply with OpenCode Go client requirements: dedicated User-Agent and namespaced x-opencode-session header for routing and prompt caching
• Bundle predefined models from the OpenCode Go catalog
• Support custom model IDs for newly added gateway models

Change Type

• Documentation / non-plugin change
• Non-LLM plugin (tools, extensions, datasource, etc.)
• LLM plugin

Screenshots / Videos

Before	After
OpenCode Go models unavailable in Dify	OpenCode Go provider with curated coding models

LLM Plugin Checklist

Areas affected by this change (check all that apply)

• Message flow (system messages, user ↔ assistant turn-taking)
• Tool interaction flow (multi-round usage, Agent App and Agent Node)
• Multimodal input (images, PDFs, audio, video, etc.)
• Multimodal output (images, audio, video, etc.)
• Structured output (JSON, XML, etc.)
• Token consumption metrics
• Other LLM functionality (reasoning, grounding, prompt caching, etc.)
• New models / model parameter fixes

Version

• Bumped top-level version in manifest.yaml (not the one under meta)
• dify_plugin declared in pyproject.toml

Testing

• Local deployment — Dify version: self-hosted
• SaaS (cloud.dify.ai)

Verified against a live OpenCode Go key:

• GET /zen/go/v1/models returns the catalog
• Non-stream and stream chat/completions succeed for GLM / Kimi / DeepSeek / MiMo / Qwen / Hy / MiniMax M3
• Request headers include the dedicated User-Agent and namespaced x-opencode-session
• Models that only speak the OpenAI Responses API (e.g. Grok 4.6, GPT 5.6 Luna) are intentionally excluded from predefined models

Docs: https://opencode.ai/docs/go/