alibaba/open-code-review · error

unsupported protocol %q; supported protocols are %q, %q, %q,

Error message

unsupported protocol %q; supported protocols are %q, %q, %q, %q

What it means

ValidateProtocol accepts only the four canonical protocol names built into open-code-review: "anthropic", "openai", "openai-responses", and "anthropic-bedrock". It is the whitelist gate used by config parsing, provider field application, and the LLM client factory. Any other (case-normalized) value fails with this error listing every accepted name.

Source

Thrown at internal/llm/protocol.go:70

	case ProtocolOpenAIChatCompletions:
		return ProtocolOpenAIChatCompletions
	case ProtocolOpenAIResponses:
		return ProtocolOpenAIResponses
	case ProtocolAnthropicBedrock:
		return ProtocolAnthropicBedrock
	default:
		return normalized
	}
}

// ValidateProtocol accepts the four canonical protocol names and rejects
// everything else.
func ValidateProtocol(p string) error {
	switch p {
	case ProtocolAnthropic, ProtocolOpenAIChatCompletions, ProtocolOpenAIResponses, ProtocolAnthropicBedrock:
		return nil
	default:
		return fmt.Errorf("unsupported protocol %q; supported protocols are %q, %q, %q, %q", p, ProtocolAnthropic, ProtocolOpenAIChatCompletions, ProtocolOpenAIResponses, ProtocolAnthropicBedrock)
	}
}

View on GitHub (pinned to 5cf97d0d15)

Solutions

1. Use one of the exact names: "anthropic", "openai", "openai-responses", or "anthropic-bedrock"
2. Call llm.NormalizeProtocol(raw) first, then ValidateProtocol on the result, to see the canonical form in the error
3. If you need Bedrock, use "anthropic-bedrock" with AWS credentials in the environment rather than a custom protocol string

Example fix

// before
protocol: "openai-chat"
// after
protocol: "openai"

p := llm.NormalizeProtocol(raw)
if err := llm.ValidateProtocol(p); err != nil {
    return fmt.Errorf("config protocol %q invalid: %w", raw, err)
}

func isKnownProtocol(p string) bool {
    switch llm.NormalizeProtocol(p) {
    case llm.ProtocolAnthropic, llm.ProtocolOpenAIChatCompletions,
        llm.ProtocolOpenAIResponses, llm.ProtocolAnthropicBedrock:
        return true
    }
    return false
}

When it happens

Trigger: Setting a protocol value in a provider/custom_providers config section, via applyProviderField, or via OCR_LLM_PROTOCOL that is not one of the four canonical names — e.g. "OpenAI" (NormalizeProtocol lowercases, so case is fine, but "openai-chat" or "gpt" are not), a misspelling like "anthropic-bedrock " is trimmed but "bedrock" alone is rejected.

Common situations: Typing a protocol alias by hand in ~/.opencodereview/config.json; copying a vendor doc name like "chat-completions" or "azure-openai"; using an old custom value from before "openai-responses"/"anthropic-bedrock" were introduced.

Understand the failure class

Background: Invalid enum value errors: "Unknown type", "Invalid scope", "must be one of" — when a string is not on the library's allowed list — this error's family across 23 libraries.

Related errors

• invalid max_tokens %q: must be a positive integer
• unknown config key: %s Supported keys: %s Provider fields: a
• invalid URL for %s: %w
• invalid model list for %s: %w
• invalid retry codes for %s: %w

AI-assisted analysis of alibaba/open-code-review@5cf97d0d15 (2026-09-02). Data as JSON: /api/errors/d16d260733f657ef. Report an issue: GitHub.