Live
GPU‑Accelerated Inference on AWS Lambda via Managed InstancesClaude Code’s “Don’t Ask Again” local file leaks credentials to public reposPost‑training RL with PivotRL lifts Nemotron 3 Ultra coding benchmarkIntegrating AI Coding Agents into Development Workflows: Protocols, Practices, and Security ImplicationsAWS Quick’s new Meta ads AI connector: what engineers need to knowAI router data leakage forces engineers to rethink prompt pipelinesRust’s Agent Client Protocol simplifies AI‑agent integration for production workloadsUsing Gaussian Sharpening with Cloudflare Images: Practical Guidance for EngineersGPU‑Accelerated Inference on AWS Lambda via Managed InstancesClaude Code’s “Don’t Ask Again” local file leaks credentials to public reposPost‑training RL with PivotRL lifts Nemotron 3 Ultra coding benchmarkIntegrating AI Coding Agents into Development Workflows: Protocols, Practices, and Security ImplicationsAWS Quick’s new Meta ads AI connector: what engineers need to knowAI router data leakage forces engineers to rethink prompt pipelinesRust’s Agent Client Protocol simplifies AI‑agent integration for production workloadsUsing Gaussian Sharpening with Cloudflare Images: Practical Guidance for Engineers
GitHub

Copilot CLI introduces on‑the‑fly local model discovery for Ollama

AI SummaryPowered by AI

GitHub Copilot CLI 1.0.94-0 adds a /model command that discovers locally hosted Ollama models alongside cloud options without auto‑adding them. This gives engineers a way to test and use local LLMs in‑session while keeping offline mode and telemetry controls explicit.

GitHub Copilot CLI version 1.0.94-0 now includes a local model discovery workflow that lets you query a running Ollama instance for compatible models directly from the CLI. This change removes the need to leave your terminal or pre‑configure models before you can experiment with a locally hosted LLM.

What Changed

The /model command now lists models served by Ollama alongside any cloud‑based models you have configured. Discovery is read‑only – it does not add the model automatically. After selecting a model you can either add it for the current session and start using it immediately, or add it without switching the active model. The CLI picks up the new model without a restart.

Why Practitioners Should Care

AI engineers and platform teams can iterate on locally built or fine‑tuned models without disrupting existing workflows. DevOps and SRE staff gain a faster feedback loop for testing model behavior in CI pipelines or on‑prem environments. Security engineers see that opting for a local model does not implicitly enable offline mode or suppress telemetry, preserving visibility into usage while still allowing isolated execution when desired.

Architectural and Operational Implications

  • Dependency Management: Ollama and the target model must already be installed; the CLI does not perform installation or download actions.
  • Model Capabilities: Only models that support tool calling and streaming appear in the picker, limiting the set to those that can handle Copilot’s interactive features.
  • Provider Errors: Connection failures to a local provider surface in the selection UI with an explanatory message, giving operators immediate diagnostic context.
  • Offline Mode Separation: Enabling COPILOT_OFFLINE=true remains an explicit step. Selecting a local model does not toggle offline mode, and remote providers can still receive prompts over the network even when offline mode is set.
  • Telemetry Continuity: GitHub telemetry remains active unless explicitly disabled, ensuring that usage data continues to flow for monitoring and compliance.

Related CloudNinjas coverage: AI engineering.

What This Means For Practitioners

Adopt the new /model workflow to evaluate locally hosted LLMs without altering your CI/CD pipelines or requiring a CLI restart. Verify that your Ollama deployment satisfies the tool‑calling and streaming requirements before adding it to the picker. Keep offline mode and telemetry settings explicit in your environment configuration to avoid accidental exposure or loss of observability. Finally, monitor the upcoming intelligent routing feature announced by Microsoft for potential automation of model selection once it becomes generally available.

Originally published atGitHub Changelog