ui/server/src/ollama · Developer Docs

Thin wrapper around Ollama's local HTTP API (default http://localhost:11434, overridable via OLLAMA_HOST for anyone running it on a non-default port/host).

Cockpit server, v0.9. Source: ui/server/src/ollama.mjs.

Functions#

getOllamaStatus()#

getOllamaStatus(): Promise<{ host?: undefined; running: boolean; version?: undefined; } | { host: any; running: boolean; version: any; } | { host: any; running: boolean; version?: undefined; }>

Defined in: ui/server/src/ollama.mjs:21

Is the local Ollama daemon up? Never throws — a non-running daemon (the expected case when Ollama isn't installed, or isn't started) is a normal, reportable state, not an error.

Returns#

Promise<{ host?: undefined; running: boolean; version?: undefined; } | { host: any; running: boolean; version: any; } | { host: any; running: boolean; version?: undefined; }>

listOllamaModels()#

listOllamaModels(): Promise<any>

Defined in: ui/server/src/ollama.mjs:35

Installed models, as Ollama's own /api/tags reports them (name, size, modified date, etc.) — passed through as-is rather than reshaped, so the UI has access to whatever Ollama itself considers a model's identity.

Returns#

Promise<any>

removeOllamaModel()#

removeOllamaModel(name): Promise<{ removed: any; }>

Defined in: ui/server/src/ollama.mjs:59

Parameters#

name

any

Returns#

Promise<{ removed: any; }>

startOllamaPull()#

startOllamaPull(name): Promise<any>

Defined in: ui/server/src/ollama.mjs:47

Starts a model pull and returns the raw fetch Response so the caller (the /api/ollama/pull route) can stream Ollama's own newline-delimited JSON progress events straight through to the browser as they arrive, instead of buffering a multi-GB download in memory. No timeout here — a real pull can legitimately run for many minutes.

Parameters#

name

any

Returns#

Promise<any>