The ollama 0.14-rc2 release is available today and it introduces new functionality with ollama run –experimental for in this experimental mode to run an agent loop so that LLMs can use tools like bash and web searching on your system. It’s opt-in for letting ollama/LLMs make use of bash on your local system and there are at least some safeguards in place.
Sure to rub some people the wrong way, the new experimental agent loop with ollama 0.14 allows for LLMs to leverage some tools like Bash. Bash command execution can be driven via LLMs but for making sure the LLMs don’t go too wild, there is an interactive approval user interface. There is also an auto-allowlist for "safe" commands as well as a deny list for blocking po…
The ollama 0.14-rc2 release is available today and it introduces new functionality with ollama run –experimental for in this experimental mode to run an agent loop so that LLMs can use tools like bash and web searching on your system. It’s opt-in for letting ollama/LLMs make use of bash on your local system and there are at least some safeguards in place.
Sure to rub some people the wrong way, the new experimental agent loop with ollama 0.14 allows for LLMs to leverage some tools like Bash. Bash command execution can be driven via LLMs but for making sure the LLMs don’t go too wild, there is an interactive approval user interface. There is also an auto-allowlist for "safe" commands as well as a deny list for blocking potentially dangerous commands like "sudo" and "rm -rf". Though among the safe commands are npm run, pwd, git status, and others. The experimental agent loop will also provide a warning box for commands working on paths outside the project directory.
This merge request that was merged for v0.14-rc2 this week explains more:
"Add –experimental flag to enable an agent loop that allows LLMs to use tools (bash, web_search) with interactive user approval.
Features:
- Built-in tools: bash command execution, web search via Ollama API
- Interactive approval UI with arrow key navigation
- Auto-allowlist for safe commands (pwd, git status, npm run, etc.)
- Denylist for dangerous patterns (rm -rf, sudo, credential access)
- Prefix-based allowlist for approved directories (cat src/ approves cat src/*)
- Warning box for commands targeting paths outside project directory"
Besides this experimental agent loop with Bash support, ollama 0.14-rc2 also brings improved Anthropic API compatibility, more accurate video RAM measurements for integrated AMD graphics, support for highlighting Swift source code, using Zstd compression for ollama Linux install bundles, and experimental support for image generation via MLX.
Per this merge, the image generation support within ollama is still in early form and only merged yesterday:
"This PR adds experimental support for image generation using the Z-Image model, powered by an MLX-based runner. This is not optimized. Both VRAM usage and generation speed will improve dramatically in future iterations. This is a start to iterate on."
Ollama 0.14-rc2 downloads for those interested via GitHub.