Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.rkat.ai/llms.txt

Use this file to discover all available pages before exploring further.

Use this guide when you want Meerkat to talk to models served from your own infrastructure or a local machine.

​
What this guide is for

This page explains the shared self-hosting model in Meerkat:
  • server definitions
  • model aliases
  • OpenAI-compatible transport expectations
  • validation and debugging
It then points to Gemma 4 as the main worked example.

​
The shared Meerkat model

Every self-hosted setup has two parts:
  1. a server definition under self_hosted.servers.<server_id>
  2. one or more model aliases under self_hosted.models.<alias>
The alias is what users type. The upstream server/model identity stays inside the self-hosted config. For current Meerkat usage, prefer:
  • transport = "openai_compatible"
  • api_style = "chat_completions"
for Ollama, LM Studio, and vLLM unless you have explicitly verified a different server-specific workflow.

​
Basic shape

[self_hosted.servers.local]
transport = "openai_compatible"
base_url = "http://127.0.0.1:11434"
api_style = "chat_completions"
bearer_token_env = "LOCAL_LLM_TOKEN"

[self_hosted.models.my-local-model]
server = "local"
remote_model = "provider/model-name"
display_name = "My Local Model"
family = "custom"
context_window = 128000
max_output_tokens = 8192
vision = true
image_tool_results = true
supports_temperature = true
supports_thinking = true
supports_reasoning = true

​
Validation checklist

  • rkat models shows the alias under the self_hosted provider group
  • rkat doctor shows the server as reachable
  • rkat run -m <alias> "say hello" works without --provider
  • the configured alias matches the upstream remote_model

​
Worked example: Gemma 4

For a detailed practical guide using real Gemma 4 aliases and server recipes, see: That page should be treated as the concrete worked example, not the entire generic self-hosting story.

​
See also