name: model-fallback-heuristic title: "Fallback — Model Switching Heuristics" repo: https://github.com/tomkubik/hermes-fallback sha: ce150572ea10de682a2c67902b314590b55683b3 version: "0.2.0" description: >- Multi-provider heuristic model fallback and automatic failover routing across all configured providers (parameter count, context window, release recency, flash tier, tool support, provider preference, and reasoning effort) — works seamlessly with or without an LLM call. Disclosure — session-start and API-error hooks rewrite fallback_providers and plugin settings in ~/.hermes/config.yaml (primary model replacement is opt-in via auto_switch_primary_when_unavailable); queries public model catalog endpoints (models.dev/api.json, openrouter.ai/api/v1/models, and inference-api.nousresearch.com/v1/portal/models) to rank candidate models and check free-tier availability; optional interactive CLI setup saves user-entered provider API keys via hermes_cli.config.save_env_value when explicitly requested. maintainer: tomkubik tier: community category: models requires_hermes: ">=0.19" docs_url: https://github.com/tomkubik/hermes-fallback#readme image: https://raw.githubusercontent.com/tomkubik/hermes-fallback/ce150572ea10de682a2c67902b314590b55683b3/assets/cover.png readme: true platforms: [] capabilities: provides_tools: - fallback_heuristic_resolve - fallback_heuristic_switch - fallback_heuristic_configure - fallback_heuristic_providers provides_hooks: - on_session_start - pre_llm_call - api_request_error provides_middleware: - llm_request - llm_execution requires_env: []