[
https://issues.apache.org/jira/browse/CAMEL-24631?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Claus Ibsen reassigned CAMEL-24631:
-----------------------------------
Assignee: Claus Ibsen
> Support Ollama local models in camel doctor and TUI doctor popup
> ----------------------------------------------------------------
>
> Key: CAMEL-24631
> URL: https://issues.apache.org/jira/browse/CAMEL-24631
> Project: Camel
> Issue Type: Improvement
> Components: camel-jbang
> Reporter: Claus Ibsen
> Assignee: Claus Ibsen
> Priority: Minor
>
> Ollama (https://ollama.com) allows running large language models locally on
> developer machines. With modern hardware (Apple Silicon, high-RAM
> workstations), models like qwen2.5:32b or llama3.3:70b run fast enough to be
> useful for interactive development workflows.
> This issue tracks improving the user-facing feedback for Ollama support
> across the Camel CLI and TUI:
> *1. camel doctor — detect local Ollama*
> camel doctor should detect whether Ollama is running at localhost:11434 and
> list pulled models. This helps developers know their AI tooling is ready
> before using camel ask or the F8 TUI panel.
> Example output when running:
> Ollama: Running at localhost:11434 — models: qwen2.5:32b,
> llama3.2:latest
> Example output when not running:
> Ollama: Not detected (optional — start for local AI with F8 in TUI)
> *2. TUI Doctor popup — show Ollama status*
> The doctor popup in the TUI should display the same Ollama detection row,
> including how many models are available and their names. The AI provider row
> should also recognise a running Ollama instance as a valid provider
> (currently it shows a warning if no API key env vars are set, even though
> Ollama needs none).
> *Usage*
> Install Ollama natively (not via camel infra / Docker, which loses Metal GPU
> acceleration on macOS):
> {code}
> brew install ollama
> ollama serve
> ollama pull qwen2.5:32b
> camel tui # press F8, AI auto-detects Ollama
> camel ask "what routes are running?"
> {code}
> *Notes*
> - camel infra run ollama uses Docker/Testcontainers which loses GPU
> acceleration on macOS. For development use, native Ollama is preferred and
> auto-detected at localhost:11434.
> - Models with tool-calling support work best for camel ask and the F8 panel
> (qwen2.5, llama3.1/3.3, deepseek-r1).
> - camel harden already supports Ollama. camel explain and camel ask support
> Ollama via the existing auto-detect.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
