everything is working so smooth, Thank u for this and I just request to add pollination.ai provider and models

[gaur-avvv]

I would like to request the integration of Pollinations.ai as a supported model provider in Claudish. Pollinations provides a simple, free-to-use API for a variety of open-source text-to-text models (like Llama 3, Qwen, and Mistral) that would be a valuable addition to the current lineup of providers like OpenRouter and Ollama.Proposed FeatureNew Provider: Add pollinations to the list of available providers.Model Support: Support the text generation endpoint (https://pollinations.ai). Functionality:Allow users to run Claude Code with Pollinations models using a flag like --model pollinations@llama.Include a model discovery/list feature to show current models supported by the Pollinations API.Authentication: Since Pollinations is currently free and doesn't require a standard API key, the implementation should allow for empty or optional credential handling.Why this is usefulPollinations offers a low-friction way for users to test Claude Code with different open-source models without needing complex local setups (like Ollama) or paid API credits (like OpenRouter). It lowers the barrier to entry for new users and provides a reliable fallback for quick testing.

[claudish-bot]

@gaur-avvv Pollinations isn't a built-in provider, but you can wire it up today via customEndpoints in ~/.claudish/config.json since their text API is OpenAI-compatible.

Drop this in your config:

{
  "customEndpoints": {
    "pollinations": {
      "kind": "simple",
      "url": "https://text.pollinations.ai/openai",
      "format": "openai",
      "apiKey": "none"
    }
  }
}

Then run claudish --model pollinations@openai "task" (or whatever model name they expose). The loader is at packages/cli/src/providers/custom-endpoints-loader.ts and the schema is documented in the README under "Custom endpoints (v7.0.0+)".

For a first-class built-in profile (with model discovery via claudish --models), that would mean a new entry in PROVIDER_PROFILES — happy to consider it, but want to gauge usage first. Does the custom endpoint route work for your case?