All plugins

pareto

Pareto Inference as a Hermes model provider (`--provider pareto`): GLM 5.3 Flash, DeepSeek V4.1 Flash and MiMo V2.6 Flash over OpenAI-compatible Chat Completions, pay per token, with /reasoning mapped to Pareto's reasoning_effort levels and context windows from models.dev. Disclosure — model provider; registers one provider profile at import and nothing else (no tools, hooks, commands or threads). Hermes sends prompts, tool results and a GET /v1/models catalog request to api.paretoinference.com with PARETO_API_KEY, which Hermes reads; the plugin itself opens no connections, reads or writes no files and adds no telemetry.

pareto-inferenceModelsv1.0.0Updated 2 hours ago

About

Pareto Inference as a Hermes model provider (`--provider pareto`): GLM 5.3 Flash, DeepSeek V4.1 Flash and MiMo V2.6 Flash over OpenAI-compatible Chat Completions, pay per token, with /reasoning mapped to Pareto's reasoning_effort levels and context windows from models.dev. Disclosure — model provider; registers one provider profile at import and nothing else (no tools, hooks, commands or threads). Hermes sends prompts, tool results and a GET /v1/models catalog request to api.paretoinference.com with PARETO_API_KEY, which Hermes reads; the plugin itself opens no connections, reads or writes no files and adds no telemetry.

Pareto Inference for Hermes Agent

Adds Pareto Inference to the Hermes provider list: GLM 5.3 Flash, DeepSeek V4.1 Flash and
MiMo V2.6 Flash, pay per token (pricing).

Install

hermes plugins install pareto
echo 'PARETO_API_KEY=your_key' >> ~/.hermes/.env
hermes model                                        # pick "Pareto Inference"

Or set it directly in ~/.hermes/config.yaml:

model:
  provider: pareto
  default: z-ai/glm-5.3-flash
auxiliary:
  # Pareto serves streamed requests only. Hermes' auxiliary tasks (titles, compression,
  # vision) are non-streamed by default, so list the endpoint here.
  stream_only_base_urls:
    - api.paretoinference.com

What it does

  • Registers one model provider, pareto (aliases pareto-inference, paretoinference), at
    https://api.paretoinference.com/v1 with PARETO_API_KEY.
  • Maps /reasoning levels to Pareto's top-level reasoning_effort (none … max; ultra
    is sent as max). With no level set, Pareto's per-model default applies.
  • Accepts bare model names (glm-5.3-flash) and sends the vendor/model id the API requires.

Disclosure

Hermes sends prompts, tool results and /models requests to api.paretoinference.com with your
PARETO_API_KEY, which Hermes reads. The plugin opens no connections, reads or writes no files,
starts no processes and adds no telemetry.

README from the reviewed commit 647bd6f; it updates when the author re-pins.