gcop-rs

English

简体中文

English

简体中文

Appearance

Provider Health Checks (gcop-rs config validate) ​

This page explains what gcop-rs config validate actually checks and how to troubleshoot failures.

Quick Start ​

bash
gcop-rs config validate

The command runs in two steps:

  1. Load and parse configuration
  2. Validate provider connectivity

What Is Validated ​

config validate uses your configured provider chain:

  • Primary provider: [llm].default_provider
  • Optional fallbacks: [llm].fallback_providers

Validation succeeds if at least one provider in that chain validates successfully.

If a configured provider cannot be instantiated (for example, missing/invalid config), it is skipped during chain construction.

Provider-Specific Validation Behavior ​

Claude / OpenAI / Gemini Style Providers ​

For Claude, OpenAI, and Gemini style providers, gcop-rs sends a minimal test request:

  • A small prompt ("test")
  • max_tokens = 1 (minimal token cost)
  • Directly to the configured endpoint

This confirms:

  • Endpoint is reachable
  • API key is accepted
  • Model and request shape are valid enough for a real call

Ollama Provider ​

For Ollama, gcop-rs performs a health-style check on:

  • .../api/tags (derived from your configured Ollama generate endpoint)

Then it verifies the configured model exists in the returned model list.

Common Failure Patterns ​

Endpoint Connection Failure ​

Symptoms:

  • Connection errors
  • Timeout errors

Checks:

  • Verify endpoint URL spelling and scheme (http / https)
  • Confirm network/proxy settings
  • For Ollama, ensure local service is running (ollama serve)

Authentication Failure (401/403) ​

Checks:

  • Confirm API key is valid and not expired
  • Confirm provider key is set in config or environment variable
  • Check provider/account permissions

Rate Limit / Server Errors (429/5xx) ​

Checks:

  • Retry later
  • Test with a backup provider via fallback_providers

Suggested Configuration Pattern ​

toml
[llm]
default_provider = "claude"
fallback_providers = ["openai", "gemini", "ollama"]

[llm.providers.gemini]
api_style = "gemini"
api_key = "AIza..."
model = "gemini-3-flash-preview"

[llm.providers.claude]
api_style = "claude"
api_key = "sk-ant-..."
model = "claude-3-5-haiku-20241022"

[llm.providers.openai]
api_style = "openai"
api_key = "sk-..."
model = "gpt-4o-mini"

[llm.providers.ollama]
api_style = "ollama"
endpoint = "http://localhost:11434/api/generate"
model = "llama3"

See Also ​