Skip to content

Routing Recipes (Ollama + Cloud)

Common workflow:

  • planning: small/local (cheap, fast)
  • coding: stronger cloud model
  • summary: cheap summarizer

Example routing

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
[router.planning]
provider = "ollama"
model = "default"

[router.coding]
provider = "glm_native" # or minimax_native / deepseek / kimi / qwen_native ...
model = "default"

[router.summary]
provider = "ollama"
model = "default"

See how-to/providers.md for full provider examples (GLM/MiniMax native + NVIDIA NIM included).

Tip: validate with small models first

Validate tool-calling + harness flow with Ollama first, then switch routing to bigger models once the loop is stable.