Ollama IDE integration

Wire Ollama into editors via OpenAI-compatible baseURL. Local models degrade faster than frontier models on bloated context — trim attachments per Context hygiene.

Hub: Ollama. API details: OpenAI API.

Zed

Configure Ollama in assistant settings. Route exploration locally; escalate to cloud mid-tier for commits. See Zed local models.

Aider

# .aider.conf.yml
model: ollama/llama3.2

Keep architect mode on cloud if local quality fails on multi-file edits.

Cline / Continue / Open WebUI

Target http://localhost:11434/v1. Narrow @ context — local 7B models fail on whole-repo dumps.

Hybrid routing

Exploration → Ollama (7B–13B)
     ↓ quality / commit gate
Mid-tier cloud → production merge

Log every escalation per Article II: local attempt, failure signal, cloud model, cost delta.

Anti-patterns

Related

• Ollama hub
• Model routing
• Tokenminning in Zed