Pollinations Realm

Pollinations provider realm for Mvge, as a bootstrap-only fallback.

Connects agents to Pollinations' legacy text endpoint — an OpenAI-compatible chat completions API that answers without an API key — so a first-time user can complete a bootstrap turn before setting up a keyed provider.

What it registers

• A realm factory for the pollinations provider.
• Provider configuration with base URL https://text.pollinations.ai/openai.

When installed into ~/.agents/extensions/pollinations-realm, Mvge discovers and activates it automatically.

Hooks

before_provider_request, after_provider_response — the realm shapes outgoing requests (slug stripping, keyless header handling, stream setup) and processes responses.

Anonymous model

Exactly one anonymous-tier model today: openai-fast (GPT-OSS 20B), wire alias openai, shipped default slug pollinations/openai. Published anonymous limit is about 1 request per 15 seconds, with no published token-per-day cap. The legacy lane can close without warning, as OpenCode Zen's anonymous tier did, so this realm is a bootstrap fallback, not a production foundation. See the rune's own README for the full warning and for the engine follow-up (credential-gate exemption and the explicit OpenRouter-key bootstrap prompt).

Dependencies

httpx>=0.27. Python >= 3.13.