OpenCode Go CLIProxyAPI Plugin

A native dynamic Go plugin for CLIProxyAPI that exposes OpenCode Go as a single provider (opencode-go).

The plugin unifies model discovery, protocol translation, and execution across OpenCode Go's upstream endpoints while leveraging CLIProxyAPI's built-in authentication, scheduling, keys rotation, and cooldown management.

The Problem

OpenCode Go exposes models across multiple API protocols (OpenAI Chat Completions /v1/chat/completions, Anthropic Messages /v1/messages, and OpenAI Responses /v1/responses).

Without this plugin, using OpenCode Go in CLIProxyAPI requires configuring separate provider blocks for each protocol family. This leads to:

• Duplicated configuration & keys: The same API keys must be configured across multiple provider blocks.
• Fragmented scheduling & rotation: Keys rotation, rate limits, and cooldowns cannot be shared across protocols—exhausting quota on one protocol does not coordinate with another.
• Client protocol burden: Clients must know beforehand which upstream protocol and endpoint each model requires.
• Fragmented catalog: Models are split across disjoint provider namespaces instead of a unified model list.

The Solution

This plugin exposes OpenCode Go as a single provider (opencode-go) backed by a shared keys pool:

• Unified auth pool: Configure keys once; CLIProxyAPI schedules, rotates, and cools down keys across all protocols.
• Transparent protocol translation & routing: Clients request models (e.g. opencode-go/glm-5.2, opencode-go/gpt-5.6-luna) without needing to know the upstream protocol format.
• Single model catalog: All models are discovered and published under the opencode-go provider namespace in /v1/models.

Features

• Single Provider Namespace: Exposes models under the opencode-go provider prefix (e.g. opencode-go/glm-5.2, opencode-go/qwen3.7-max, opencode-go/gpt-5.6-luna).
• Multi-Protocol Translation: Translates requests and streaming responses between client formats and upstream endpoints:
  • OpenAI Chat Completions (/v1/chat/completions)
  • Anthropic Messages (/v1/messages)
  • OpenAI Responses (/v1/responses)
• Thinking & Reasoning Support: Maps reasoning effort across supported client and upstream formats.
• Dynamic Catalog Discovery: Fetches remote model catalogs with local fallback and custom route overrides.
• Multi-Key Auth Scheduling: Pools multiple API keys with CLIProxyAPI's native scheduler for rotation, retries, and error cooldowns across all protocols.
• OpenCode Go Quota Page: Management Center includes a separate OpenCode Go Quota page. Page load lists credentials without contacting OpenCode; each card is refreshed manually and independently, and quota values do not affect routing or CPA's native quota page.

Requirements

• CLIProxyAPI: v7.2.138+
• Go Toolchain: Go 1.24+ (with CGO enabled for C-shared build mode)

Build

Build the dynamic shared library for your platform:

Windows (AMD64)

go build -buildmode=c-shared -o plugins/windows/amd64/opencode-go-cliproxyapi.dll .

Linux (AMD64)

go build -buildmode=c-shared -o plugins/linux/amd64/opencode-go-cliproxyapi.so .

macOS (ARM64)

go build -buildmode=c-shared -o plugins/darwin/arm64/opencode-go-cliproxyapi.dylib .

Place the compiled binary into your CLIProxyAPI plugin directory (e.g. <cliproxyapi_root>/plugins/<os>/<arch>/).

Configuration

Configure the plugin in your CLIProxyAPI config.yaml under plugins.configs.opencode-go-cliproxyapi:

plugins:
  configs:
    opencode-go-cliproxyapi:
      # Upstream base URL (default: "https://opencode.ai/zen/go/v1")
      base-url: "https://opencode.ai/zen/go/v1"

      # Optional catalog endpoint override (default: "{base-url}/models")
      # catalog-url: "https://opencode.ai/zen/go/v1/models"

      # Client-facing model ID prefix configuration
      model-prefix:
        enabled: true           # true -> "opencode-go/<model>", false -> bare "<model>" (default: true)
        value: "opencode-go"    # prefix name (default: "opencode-go")

      # OpenCode Go API keys (at least one required). Supports ${ENV_VAR} expansion.
      api-keys:
        - value: "sk-opencode-key-1"
        - value: "sk-opencode-key-2"
        - value: "${OPENCODE_GO_API_KEY}"

      # Catalog discovery settings
      catalog:
        refresh-interval: "15m"          # discovery refresh cadence, min "1m" (default: "15m")
        stale-while-unavailable: true    # retain last good catalog snapshot on refresh failure (default: true)

      # Protocol enable/disable switches (all default to true)
      protocols:
        chat-completions: true   # enables models routed to /v1/chat/completions
        messages: true           # enables models routed to /v1/messages
        responses: true          # enables models routed to /v1/responses

      # Explicit route overrides per model (takes priority over built-in prefix routing)
      route-overrides:
        "custom-model":
          protocol: "messages"           # "chat-completions" | "messages" | "responses"
          endpoint: "/v1/messages"       # must start with /

      # Execution settings
      request-timeout: "5m"              # upstream request timeout (default: "5m")
      max-response-bytes: 67108864       # max non-streaming response body size in bytes (default: 64 MiB)
      allow-http: false                  # allow http:// scheme for local mock/testing (default: false)

Configuration Options

Option	Type	Default	Description
api-keys	[]object	(Required)	List of API keys (- value: "..."). Supports ${ENV_VAR} expansion. Duplicates and empty values are rejected.
base-url	string	https://opencode.ai/zen/go/v1	Upstream base URL. Must be valid HTTPS (or HTTP if allow-http: true) without query parameters, fragments, or userinfo.
catalog-url	string	{base-url}/models	Full URL for catalog discovery. Defaults to {base-url}/models.
model-prefix.enabled	bool	true	When true, client-facing model names use <prefix>/<model>. When false, uses bare model IDs.
model-prefix.value	string	opencode-go	Provider prefix string when prefixing is enabled.
catalog.refresh-interval	duration	15m	Interval between catalog polling refreshes (e.g. 15m, 1h). Minimum is 1m.
catalog.stale-while-unavailable	bool	true	When true, serves the last valid catalog snapshot if an update fails.
protocols.chat-completions	bool	true	Protocol switch for Chat Completions endpoints.
protocols.messages	bool	true	Protocol switch for Messages endpoints.
protocols.responses	bool	true	Protocol switch for Responses endpoints.
route-overrides	map	{}	Map of model ID to { protocol: "...", endpoint: "..." } overriding built-in family routing. Valid protocols: chat-completions, messages, responses.
request-timeout	duration	5m	Upstream HTTP request timeout. Must be positive.
max-response-bytes	int64	67108864 (64 MiB)	Maximum non-streaming response body size in bytes.
allow-http	bool	false	When true, permits http:// scheme in base-url / catalog-url for local testing.

Testing

# Run all tests
go test ./...

# Run tests with coverage
go test ./... -cover

# Run linter / vetting
go vet ./...