github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 08:59:13 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #7239] Add Tab-Enabled Autocomplete for Local Model Parameters in Ollama CLI #51109

New Issue
Closed
opened 2026-04-28 18:24:25 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-28 18:24:25 -05:00
Owner
Copy Link

Originally created by @lucianoayres on GitHub (Oct 17, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/7239

It would greatly enhance usability if the Ollama CLI supported tab-autocomplete for model names when using commands like run, show, list, etc. For example:

# This would autocomplete to something like `llama3.2`, based on the locally available models.
ollama run lla<TAB>

Implementing this across all commands where model names are parameters would improve efficiency and user experience.
Thank you!

Originally created by @lucianoayres on GitHub (Oct 17, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/7239 It would greatly enhance usability if the Ollama CLI supported tab-autocomplete for model names when using commands like `run`, `show`, `list`, etc. For example: ```bash # This would autocomplete to something like `llama3.2`, based on the locally available models. ollama run lla<TAB> ``` Implementing this across all commands where model names are parameters would improve efficiency and user experience. Thank you!
GiteaMirror added the feature request label 2026-04-28 18:24:25 -05:00
GiteaMirror commented 2026-04-28 18:24:27 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Oct 19, 2024):

#!/bin/bash

# TAB completion for ollama CLI.
#
# Install in /etc/bash_completion.d/, or in your home dir
# and `source <this file>` in your .bashrc.
#
# Needs `jq` and `curl`.
#
# Install `pup` [https://github.com/ericchiang/pup] to
# get ollama library completion.
#
# A few environment variables control processing:
#   _OLLAMA_MODEL_TTL    
#       Minimum number of seconds between model fetches from the
#       ollama server. Default 300s.
#   _OLLAMA_LIBRARY_TTL
#       Minimum number of seconds between model fetches from the
#       ollama library.  Default 3600s.
#   _OLLAMA_QUANTS_TTL
#       Minimum number of seconds between quant fetches from the
#       ollama library.  Default 300s.
#   _OLLAMA_LIBRARY_LIMIT
#       Maximum number of models to retrieve from the ollama
#       library.  Default 10.
#   _OLLAMA_LIBRARY_SORT
#       Sort order for models fetched from the ollama library.
#       One of [newest popular featured].  Default newest.
#   _OLLAMA_QUANTS
#       The list of quantizations used for completion for
#       `ollama create --quantize`.

if [ -z "$BASH_VERSION" ] ; then
  echo "$0: Need bash"
  return 1
fi
if ! command -v jq >&- ; then
  echo "$0: Need jq for model completion"
  return 1
fi
if ! command -v curl >&- ; then
  echo "$0: Need curl for model completion"
  return 1
fi

# restore post exec hook
PROMPT_COMMAND=${PROMPT_COMMAND//_ollama_post_exec ; /}

_OLLAMA_MODELS=""
_OLLAMA_LIBRARY=""
_OLLAMA_MODEL_TTL=${_OLLAMA_MODEL_TTL-300}
_OLLAMA_LIBRARY_TTL=${_OLLAMA_LIBRARY_TTL-3600}
_OLLAMA_QUANTS_TTL=${_OLLAMA_QUANTS_TTL-300}
_OLLAMA_LIBRARY_LIMIT=${_OLLAMA_LIBRARY_LIMIT-10}
_OLLAMA_LIBRARY_SORT=${_OLLAMA_LIBRARY_SORT-newest}     # newest popular featured
_OLLAMA_MODELS_TIMESTAMP=0
_OLLAMA_LIBRARY_TIMESTAMP=0
_OLLAMA_QUANTS_ALL=(
    F32 F16 BF16
    Q2_K Q2_K_S
    Q3_K_S Q3_K_M Q3_K_L
    Q4_0 Q4_1 Q4_1_F16 Q4_K_S Q4_K_M
    Q5_0 Q5_1 Q5_K_S Q5_K_M
    Q6_K
    Q8_0
    IQ1_S IQ1_M
    IQ2_XXS IQ2_XS IQ2_S IQ2_M
    IQ3_XXS IQ3_XS IQ3_S
    IQ4_NL IQ4_XS
)
# The above is all that is supported, but just use a select few for completion.
#_OLLAMA_QUANTS=${_OLLAMA_QUANTS-${_OLLAMA_QUANTS_ALL[@]}}
_OLLAMA_QUANTS=${_OLLAMA_QUANTS-Q4_0 Q4_1 Q4_1_F16 Q8_0 Q4_K_S Q4_K_M F16}

_ollama_fetch_models(){
  _OLLAMA_MODELS="$(curl -s ${OLLAMA_HOST-localhost:11434}/api/tags | jq -r '.models[].name')"
}

_ollama_maybe_fetch_models(){
  [ $[ $(date +%s) - ${_OLLAMA_MODELS_TIMESTAMP-0} ] -lt $_OLLAMA_MODEL_TTL ] && return 0
  _ollama_fetch_models
  _OLLAMA_MODELS_TIMESTAMP=$(date +%s)
}

_ollama_fetch_library_models(){
  ! command -v pup >&- && return 0
  _OLLAMA_LIBRARY=$(echo "{$(echo $(curl -s https://ollama.com/library?sort=${_OLLAMA_LIBRARY_SORT-newest} | 
      pup '#repo ul li a' | 
      sed -ne 's@^<a href="/library/\([^"]*\)".*@"\1:":{"base":"/library/\1"}@p' |
      head -${_OLLAMA_LIBRARY_LIMIT-10}) |
      tr ' ' ,)}")
}

_ollama_maybe_fetch_library_models(){
  [ $[ $(date +%s) - ${_OLLAMA_LIBRARY_TIMESTAMP-0} ] -lt $_OLLAMA_LIBRARY_TTL ] && return 0
  _ollama_fetch_library_models
  _OLLAMA_LIBRARY_TIMESTAMP=$(date +%s)
}

_ollama_maybe_fetch_library(){
  local model completions timestamp library quants
  ! command -v pup >&- && return 0
  [ -z "$_OLLAMA_LIBRARY" ] && {
    _ollama_maybe_fetch_library_models
    [ -z "$_OLLAMA_LIBRARY" ] && return 0
  }
  # we might have to fetch quants.  in the case where the user has typed
  # enough of the model name to uniquely identify it, we can do the fetch.
  model=${COMP_LINE:0:COMP_POINT}
  model=${model##* }
  [[ "$model" == *:* ]] && model=${model%%:*}: || model=${model%%:*}
  completions=( $(compgen -W "$(jq -rn "$_OLLAMA_LIBRARY|keys|.[]")" -- "$model") )
  [ ${#completions[*]} -ne 1 ] && return 0
  model="${completions[0]}"
  # skip if quants were retrieved recently
  timestamp="$(jq -rn "$_OLLAMA_LIBRARY|.\"$model\".timestamp")"
  [ -z "$timestamp" -o "$timestamp" == null ] && timestamp=0
  [ $[ $(date +%s) - $timestamp ] -lt $_OLLAMA_QUANTS_TTL ] && return 0
  library="$(jq -rn "$_OLLAMA_LIBRARY|.\"$model\".base")"
  [ -z "$library" -o "$library" == null ] && return 0
  quants="$(echo $(curl -s https://ollama.com${library}/tags | pup 'section div div div div div div text{}'))"
  [ -z "$quants" -o "$quants" == null ] && return 0
  # add the quants prefixed by the model
  quants=$(eval echo $model{${quants// /,}})
  _OLLAMA_LIBRARY=$(jq -cn "$_OLLAMA_LIBRARY*{\"$model\":{\"quants\":\"$quants\",\"timestamp\":$(date +%s)}}")
}

# ":" is a word break character by default, which messes
# completion attempts for models with tags.  so here we
# massage the potential matches so that bash presents the
# appropriate words.  expects an array of completion
# alternatives in $_OLLAMA_COMPLETE.
_ollama_massage_completions(){
  local comp model tmp
  comp=${COMP_LINE:0:COMP_POINT}
  comp=${comp##* }
  [[ "$comp" != *:* ]] && return 0
  model=${comp%:*}:
  tmp=()
  for i in ${_OLLAMA_COMPLETE[*]} ; do
    [ "${i:0:${#comp}}" == "$comp" ] && {
      t=${i:${#model}}
      [ -n "$t" ] && tmp+=($t)
    }
  done
  _OLLAMA_COMPLETE=("${tmp[@]}")
}

_ollama_complete_models(){
  [ -z "$_OLLAMA_MODELS" ] && return 0
  read -a _OLLAMA_COMPLETE <<< $(tr '\n' ' ' <<< $_OLLAMA_MODELS)
  _ollama_massage_completions
  compgen -W "${_OLLAMA_COMPLETE[*]}" -- "$1"
}

_ollama_complete_library(){
  local model
  [ -z "$_OLLAMA_LIBRARY" ] && return 0
  model=${COMP_LINE:0:COMP_POINT}
  model=${model##* }
  [[ -z "$model" || "$model" != *:* ]] && {
    compgen -W "$(jq -rn "$_OLLAMA_LIBRARY|keys|.[]")" -- "$model"
    return 0
  }
  model="${model%%:*}:"
  _OLLAMA_COMPLETE=( $(jq -rn "$_OLLAMA_LIBRARY|.\"$model\".quants") )
  _ollama_massage_completions
  compgen -W "${_OLLAMA_COMPLETE[*]}" -- "$1"
}

_ollama() {
  local word verb
  _ollama_maybe_fetch_models
  for word in ${COMP_WORDS[@]:1} ; do
    [[ "$word" != -* ]] && {
      verb="$word"
      break
    }
  done
  [ "$verb" == "$2" -a "${COMP_LINE:${#COMP_LINE}-1}" != " " ] && verb="*"
  case "$verb" in
    serve|start)
            COMPREPLY=( $(compgen -W "--help" -- "$2") ) ;
            ;;
    create)
            if [ "${COMP_WORDS[COMP_CWORD-1]}" == --quantize ] ; then
              COMPREPLY=( $(echo "$_OLLAMA_QUANTS" | tr ' ' '\n' | grep -i "^$2") ) ;
            elif [ "${COMP_WORDS[COMP_CWORD-1]}" == --file ] ; then
              COMPREPLY=( $(compgen -f -o filenames -- "$2") ) ;
            else
              COMPREPLY=( $(compgen -W "--help --file --quantize" -- "$2") ) ;
            fi
            _OLLAMA_FLUSH_MODELS_CACHE=1
            ;;
    show)
            if [ "${2:0:1}" == - ] ; then
              COMPREPLY=( $(compgen -W "--help --license --template --modelfile --parameters --system" -- "$2") ) ;
            else
              COMPREPLY=( $(_ollama_complete_models "$2") ) ;
            fi
            ;;
    run)
            if [ "${2:0:1}" == - ] ; then
              COMPREPLY=( $(compgen -W "--help --format --insecure --keepalive --nowordwrap --verbose" -- "$2") ) ;
            elif [ "${COMP_WORDS[COMP_CWORD-1]}" == --format ] ; then
              COMPREPLY=( $(compgen -W "json" -- "$2") ) ;
            else
              COMPREPLY=( $(_ollama_complete_models "$2") ) ;
            fi
            ;;
    pull)
            if [ "${2:0:1}" == - ] ; then
              COMPREPLY=( $(compgen -W "--help --insecure" -- "$2") ) ;
            else
              _ollama_maybe_fetch_library
              COMPREPLY=( $(_ollama_complete_library "$2") ) ;
            fi
            _OLLAMA_FLUSH_MODELS_CACHE=1
            ;;
    push)
            if [ "${2:0:1}" == - ] ; then
              COMPREPLY=( $(compgen -W "--help --insecure" -- "$2") ) ;
            else
              COMPREPLY=( $(_ollama_complete_models "$2") ) ;
            fi
            ;;
    list|ls)
            if [ "${2:0:1}" == - ] ; then
              COMPREPLY=( $(compgen -W "--help" -- "$2") ) ;
            else
              COMPREPLY=( $(_ollama_complete_models "$2") ) ;
            fi
            ;;
    ps)
            if [ "${2:0:1}" == - ] ; then
              COMPREPLY=( $(compgen -W "--help" -- "$2") ) ;
            else
              COMPREPLY=( $(_ollama_complete_models "$2") ) ;
            fi
            ;;
    cp)
            if [ "${2:0:1}" == - ] ; then
              COMPREPLY=( $(compgen -W "--help" -- "$2") ) ;
            else
              COMPREPLY=( $(_ollama_complete_models "$2") ) ;
            fi
            _OLLAMA_FLUSH_MODELS_CACHE=1
            ;;
    rm)
            if [ "${2:0:1}" == - ] ; then
              COMPREPLY=( $(compgen -W "--help" -- "$2") ) ;
            else
              COMPREPLY=( $(_ollama_complete_models "$2") ) ;
            fi
            _OLLAMA_FLUSH_MODELS_CACHE=1
            ;;
    help)
            COMPREPLY=( $(compgen -W "serve create show run pull push list ps cp rm" -- "$2") )
            ;;
    *)      COMPREPLY=( $(compgen -W "serve create show run pull push list ps cp rm help" -- "$2") )
            ;;
  esac
}

_ollama_post_exec(){
  [ "$_OLLAMA_FLUSH_MODELS_CACHE" == 1 ] && {
    _OLLAMA_MODELS=""
    _OLLAMA_MODELS_TIMESTAMP=0
    _OLLAMA_FLUSH_MODELS_CACHE=0
  }
  [ "$_OLLAMA_FLUSH_LIBRARY_CACHE" == 1 ] && {
    _OLLAMA_LIBRARY=""
    _OLLAMA_LIBRARY_TIMESTAMP=0
    _OLLAMA_FLUSH_LIBRARY_CACHE=0
  }
}

PROMPT_COMMAND="_ollama_post_exec ; $PROMPT_COMMAND"

complete -F _ollama ollama
<!-- gh-comment-id:2423530279 --> @rick-github commented on GitHub (Oct 19, 2024): ```bash #!/bin/bash # TAB completion for ollama CLI. # # Install in /etc/bash_completion.d/, or in your home dir # and `source <this file>` in your .bashrc. # # Needs `jq` and `curl`. # # Install `pup` [https://github.com/ericchiang/pup] to # get ollama library completion. # # A few environment variables control processing: # _OLLAMA_MODEL_TTL # Minimum number of seconds between model fetches from the # ollama server. Default 300s. # _OLLAMA_LIBRARY_TTL # Minimum number of seconds between model fetches from the # ollama library. Default 3600s. # _OLLAMA_QUANTS_TTL # Minimum number of seconds between quant fetches from the # ollama library. Default 300s. # _OLLAMA_LIBRARY_LIMIT # Maximum number of models to retrieve from the ollama # library. Default 10. # _OLLAMA_LIBRARY_SORT # Sort order for models fetched from the ollama library. # One of [newest popular featured]. Default newest. # _OLLAMA_QUANTS # The list of quantizations used for completion for # `ollama create --quantize`. if [ -z "$BASH_VERSION" ] ; then echo "$0: Need bash" return 1 fi if ! command -v jq >&- ; then echo "$0: Need jq for model completion" return 1 fi if ! command -v curl >&- ; then echo "$0: Need curl for model completion" return 1 fi # restore post exec hook PROMPT_COMMAND=${PROMPT_COMMAND//_ollama_post_exec ; /} _OLLAMA_MODELS="" _OLLAMA_LIBRARY="" _OLLAMA_MODEL_TTL=${_OLLAMA_MODEL_TTL-300} _OLLAMA_LIBRARY_TTL=${_OLLAMA_LIBRARY_TTL-3600} _OLLAMA_QUANTS_TTL=${_OLLAMA_QUANTS_TTL-300} _OLLAMA_LIBRARY_LIMIT=${_OLLAMA_LIBRARY_LIMIT-10} _OLLAMA_LIBRARY_SORT=${_OLLAMA_LIBRARY_SORT-newest} # newest popular featured _OLLAMA_MODELS_TIMESTAMP=0 _OLLAMA_LIBRARY_TIMESTAMP=0 _OLLAMA_QUANTS_ALL=( F32 F16 BF16 Q2_K Q2_K_S Q3_K_S Q3_K_M Q3_K_L Q4_0 Q4_1 Q4_1_F16 Q4_K_S Q4_K_M Q5_0 Q5_1 Q5_K_S Q5_K_M Q6_K Q8_0 IQ1_S IQ1_M IQ2_XXS IQ2_XS IQ2_S IQ2_M IQ3_XXS IQ3_XS IQ3_S IQ4_NL IQ4_XS ) # The above is all that is supported, but just use a select few for completion. #_OLLAMA_QUANTS=${_OLLAMA_QUANTS-${_OLLAMA_QUANTS_ALL[@]}} _OLLAMA_QUANTS=${_OLLAMA_QUANTS-Q4_0 Q4_1 Q4_1_F16 Q8_0 Q4_K_S Q4_K_M F16} _ollama_fetch_models(){ _OLLAMA_MODELS="$(curl -s ${OLLAMA_HOST-localhost:11434}/api/tags | jq -r '.models[].name')" } _ollama_maybe_fetch_models(){ [ $[ $(date +%s) - ${_OLLAMA_MODELS_TIMESTAMP-0} ] -lt $_OLLAMA_MODEL_TTL ] && return 0 _ollama_fetch_models _OLLAMA_MODELS_TIMESTAMP=$(date +%s) } _ollama_fetch_library_models(){ ! command -v pup >&- && return 0 _OLLAMA_LIBRARY=$(echo "{$(echo $(curl -s https://ollama.com/library?sort=${_OLLAMA_LIBRARY_SORT-newest} | pup '#repo ul li a' | sed -ne 's@^<a href="/library/\([^"]*\)".*@"\1:":{"base":"/library/\1"}@p' | head -${_OLLAMA_LIBRARY_LIMIT-10}) | tr ' ' ,)}") } _ollama_maybe_fetch_library_models(){ [ $[ $(date +%s) - ${_OLLAMA_LIBRARY_TIMESTAMP-0} ] -lt $_OLLAMA_LIBRARY_TTL ] && return 0 _ollama_fetch_library_models _OLLAMA_LIBRARY_TIMESTAMP=$(date +%s) } _ollama_maybe_fetch_library(){ local model completions timestamp library quants ! command -v pup >&- && return 0 [ -z "$_OLLAMA_LIBRARY" ] && { _ollama_maybe_fetch_library_models [ -z "$_OLLAMA_LIBRARY" ] && return 0 } # we might have to fetch quants. in the case where the user has typed # enough of the model name to uniquely identify it, we can do the fetch. model=${COMP_LINE:0:COMP_POINT} model=${model##* } [[ "$model" == *:* ]] && model=${model%%:*}: || model=${model%%:*} completions=( $(compgen -W "$(jq -rn "$_OLLAMA_LIBRARY|keys|.[]")" -- "$model") ) [ ${#completions[*]} -ne 1 ] && return 0 model="${completions[0]}" # skip if quants were retrieved recently timestamp="$(jq -rn "$_OLLAMA_LIBRARY|.\"$model\".timestamp")" [ -z "$timestamp" -o "$timestamp" == null ] && timestamp=0 [ $[ $(date +%s) - $timestamp ] -lt $_OLLAMA_QUANTS_TTL ] && return 0 library="$(jq -rn "$_OLLAMA_LIBRARY|.\"$model\".base")" [ -z "$library" -o "$library" == null ] && return 0 quants="$(echo $(curl -s https://ollama.com${library}/tags | pup 'section div div div div div div text{}'))" [ -z "$quants" -o "$quants" == null ] && return 0 # add the quants prefixed by the model quants=$(eval echo $model{${quants// /,}}) _OLLAMA_LIBRARY=$(jq -cn "$_OLLAMA_LIBRARY*{\"$model\":{\"quants\":\"$quants\",\"timestamp\":$(date +%s)}}") } # ":" is a word break character by default, which messes # completion attempts for models with tags. so here we # massage the potential matches so that bash presents the # appropriate words. expects an array of completion # alternatives in $_OLLAMA_COMPLETE. _ollama_massage_completions(){ local comp model tmp comp=${COMP_LINE:0:COMP_POINT} comp=${comp##* } [[ "$comp" != *:* ]] && return 0 model=${comp%:*}: tmp=() for i in ${_OLLAMA_COMPLETE[*]} ; do [ "${i:0:${#comp}}" == "$comp" ] && { t=${i:${#model}} [ -n "$t" ] && tmp+=($t) } done _OLLAMA_COMPLETE=("${tmp[@]}") } _ollama_complete_models(){ [ -z "$_OLLAMA_MODELS" ] && return 0 read -a _OLLAMA_COMPLETE <<< $(tr '\n' ' ' <<< $_OLLAMA_MODELS) _ollama_massage_completions compgen -W "${_OLLAMA_COMPLETE[*]}" -- "$1" } _ollama_complete_library(){ local model [ -z "$_OLLAMA_LIBRARY" ] && return 0 model=${COMP_LINE:0:COMP_POINT} model=${model##* } [[ -z "$model" || "$model" != *:* ]] && { compgen -W "$(jq -rn "$_OLLAMA_LIBRARY|keys|.[]")" -- "$model" return 0 } model="${model%%:*}:" _OLLAMA_COMPLETE=( $(jq -rn "$_OLLAMA_LIBRARY|.\"$model\".quants") ) _ollama_massage_completions compgen -W "${_OLLAMA_COMPLETE[*]}" -- "$1" } _ollama() { local word verb _ollama_maybe_fetch_models for word in ${COMP_WORDS[@]:1} ; do [[ "$word" != -* ]] && { verb="$word" break } done [ "$verb" == "$2" -a "${COMP_LINE:${#COMP_LINE}-1}" != " " ] && verb="*" case "$verb" in serve|start) COMPREPLY=( $(compgen -W "--help" -- "$2") ) ; ;; create) if [ "${COMP_WORDS[COMP_CWORD-1]}" == --quantize ] ; then COMPREPLY=( $(echo "$_OLLAMA_QUANTS" | tr ' ' '\n' | grep -i "^$2") ) ; elif [ "${COMP_WORDS[COMP_CWORD-1]}" == --file ] ; then COMPREPLY=( $(compgen -f -o filenames -- "$2") ) ; else COMPREPLY=( $(compgen -W "--help --file --quantize" -- "$2") ) ; fi _OLLAMA_FLUSH_MODELS_CACHE=1 ;; show) if [ "${2:0:1}" == - ] ; then COMPREPLY=( $(compgen -W "--help --license --template --modelfile --parameters --system" -- "$2") ) ; else COMPREPLY=( $(_ollama_complete_models "$2") ) ; fi ;; run) if [ "${2:0:1}" == - ] ; then COMPREPLY=( $(compgen -W "--help --format --insecure --keepalive --nowordwrap --verbose" -- "$2") ) ; elif [ "${COMP_WORDS[COMP_CWORD-1]}" == --format ] ; then COMPREPLY=( $(compgen -W "json" -- "$2") ) ; else COMPREPLY=( $(_ollama_complete_models "$2") ) ; fi ;; pull) if [ "${2:0:1}" == - ] ; then COMPREPLY=( $(compgen -W "--help --insecure" -- "$2") ) ; else _ollama_maybe_fetch_library COMPREPLY=( $(_ollama_complete_library "$2") ) ; fi _OLLAMA_FLUSH_MODELS_CACHE=1 ;; push) if [ "${2:0:1}" == - ] ; then COMPREPLY=( $(compgen -W "--help --insecure" -- "$2") ) ; else COMPREPLY=( $(_ollama_complete_models "$2") ) ; fi ;; list|ls) if [ "${2:0:1}" == - ] ; then COMPREPLY=( $(compgen -W "--help" -- "$2") ) ; else COMPREPLY=( $(_ollama_complete_models "$2") ) ; fi ;; ps) if [ "${2:0:1}" == - ] ; then COMPREPLY=( $(compgen -W "--help" -- "$2") ) ; else COMPREPLY=( $(_ollama_complete_models "$2") ) ; fi ;; cp) if [ "${2:0:1}" == - ] ; then COMPREPLY=( $(compgen -W "--help" -- "$2") ) ; else COMPREPLY=( $(_ollama_complete_models "$2") ) ; fi _OLLAMA_FLUSH_MODELS_CACHE=1 ;; rm) if [ "${2:0:1}" == - ] ; then COMPREPLY=( $(compgen -W "--help" -- "$2") ) ; else COMPREPLY=( $(_ollama_complete_models "$2") ) ; fi _OLLAMA_FLUSH_MODELS_CACHE=1 ;; help) COMPREPLY=( $(compgen -W "serve create show run pull push list ps cp rm" -- "$2") ) ;; *) COMPREPLY=( $(compgen -W "serve create show run pull push list ps cp rm help" -- "$2") ) ;; esac } _ollama_post_exec(){ [ "$_OLLAMA_FLUSH_MODELS_CACHE" == 1 ] && { _OLLAMA_MODELS="" _OLLAMA_MODELS_TIMESTAMP=0 _OLLAMA_FLUSH_MODELS_CACHE=0 } [ "$_OLLAMA_FLUSH_LIBRARY_CACHE" == 1 ] && { _OLLAMA_LIBRARY="" _OLLAMA_LIBRARY_TIMESTAMP=0 _OLLAMA_FLUSH_LIBRARY_CACHE=0 } } PROMPT_COMMAND="_ollama_post_exec ; $PROMPT_COMMAND" complete -F _ollama ollama ```
GiteaMirror commented 2026-04-28 18:24:27 -05:00
Author
Owner
Copy Link

@vt-alt commented on GitHub (Nov 21, 2024):

@rick-github Btw, you don't need to list built-in commands and options as Cobra provides a built-in hidden command for that:

$ ollama __complete ''
serve   Start ollama
create  Create a model from a Modelfile
show    Show information for a model
run     Run a model
stop    Stop a running model
pull    Pull a model from a registry
push    Push a model to a registry
list    List models
ps      List running models
cp      Copy a model
rm      Remove a model
help    Help about any command
:4
Completion ended with directive: ShellCompDirectiveNoFileComp

$ ollama __complete run --
--format        Response format (e.g. json)
--help  help for run
--insecure      Use an insecure registry
--keepalive     Duration to keep a model loaded (e.g. 5m)
--nowordwrap    Don't wrap words to the next line automatically
--verbose       Show timings for response
:4
Completion ended with directive: ShellCompDirectiveNoFileComp

But its output is designed to be parsed by their internal completion script which is explicitly disabled in ollama with DisableDefaultCmd: true.

<!-- gh-comment-id:2490009185 --> @vt-alt commented on GitHub (Nov 21, 2024): @rick-github Btw, you don't need to list built-in commands and options as Cobra provides a built-in hidden command for that: ``` $ ollama __complete '' serve Start ollama create Create a model from a Modelfile show Show information for a model run Run a model stop Stop a running model pull Pull a model from a registry push Push a model to a registry list List models ps List running models cp Copy a model rm Remove a model help Help about any command :4 Completion ended with directive: ShellCompDirectiveNoFileComp $ ollama __complete run -- --format Response format (e.g. json) --help help for run --insecure Use an insecure registry --keepalive Duration to keep a model loaded (e.g. 5m) --nowordwrap Don't wrap words to the next line automatically --verbose Show timings for response :4 Completion ended with directive: ShellCompDirectiveNoFileComp ``` But its output is designed to be parsed by their internal completion script which is explicitly disabled in ollama with `DisableDefaultCmd: true`.
GiteaMirror commented 2026-04-28 18:24:28 -05:00
Author
Owner
Copy Link

@vt-alt commented on GitHub (Nov 22, 2024):

My take, and I plan to package this for ALT Linux.

# ollama(1) completion                                     -*- shell-script -*-
# vi: ft=sh sw=4
# SPDX-License-Identifier: Apache-2.0
# shellcheck disable=SC2034

_comp_ollama()
{
    local cur prev words cword comp_args
    _comp_initialize -n =: -- "$@" || return
    words=("${words[@]:0:$cword+1}")

    if ((cword == 2)); then
        # Models listing commands.
        case "$prev" in
            pull)
                local models_list=/usr/share/ollama/models-list.txt
                local tags_list=/usr/share/ollama/tags-list.txt
                if [[ $cur == *:* ]]; then
                    _comp_compgen_split -- "$(LANG=C look "$cur" "$tags_list")"
                    _comp_ltrim_colon_completions "$cur"
                else
                    _comp_compgen_split -- "$(LANG=C look "$cur" "$models_list")"
                    if [[ ${COMPREPLY-} != "$cur" ]]; then
                        # Allow user to enter ':' after the model name or a space if
                        # another tab is pressed.
                        compopt -o nospace
                    fi
                fi
                ;;
            cp|push|rm|run|show)
                _comp_compgen_split -- "$(ollama list 2>/dev/null | tail -n+2 | grep -Eo '^\S+')"
                _comp_ltrim_colon_completions "$cur"
                ;;
            stop)
                _comp_compgen_split -- "$(ollama ps 2>/dev/null | tail -n+2 | grep -Eo '^\S+')"
                _comp_ltrim_colon_completions "$cur"
                ;;
        esac
        ((${#COMPREPLY[@]})) && return
    fi

    # Fallback to completing Cobra built-ins.
    local requestComp lastParam lastChar args
    args=("${words[@]:1}")
    requestComp="${words[0]} __completeNoDesc ${args[*]}"
    stParam=${words[$((${#words[@]}-1))]}
    lastChar=${lastParam:$((${#lastParam}-1)):1}
    if [[ -z $cur && $lastChar != = ]]; then
        requestComp="$requestComp ''"
    fi
    if [[ $cur == -*=* ]]; then
        cur="${cur#*=}"
    fi
    local out directive
    out=$(eval "$requestComp" 2>/dev/null)
    directive=${out##*:}
    out=${out%:*}
    if [[ $directive == "$out" ]]; then
        directive=0
    fi
    local shellCompDirectiveError=1
    local shellCompDirectiveNoSpace=2
    local shellCompDirectiveNoFileComp=4
    local shellCompDirectiveFilterFileExt=8
    local shellCompDirectiveFilterDirs=16
    local shellCompDirectiveKeepOrder=32
    if (((directive & shellCompDirectiveError) != 0)); then
        return
    else
        if (((directive & shellCompDirectiveNoSpace) != 0)); then
            compopt -o nospace
        fi
        if (((directive & shellCompDirectiveKeepOrder) != 0)); then
            compopt -o nosort
        fi
        if (((directive & shellCompDirectiveNoFileComp) != 0)); then
            compopt +o default
        fi
    fi
    _comp_compgen_split -- "$out"
} &&
    complete -F _comp_ollama ollama

For the script to work you will need to have /usr/share/ollama/{models,tags}-list.txt files, created with something like https://github.com/ollama/ollama/issues/7751#issuecomment-2488290937 but they shall be properly sorted with LANG=C sort — this is mandatory for look to work correctly.

(The second part with a fallback is based on bash_completionsV2.go from Cobra and is greatly shortened.)

UI/UX considerations:

  1. Only the pull shows remote models, not the run — this is to not overwhelm the user, perhaps, they will more frequently will want to run the local models.
  2. First it completes only for model names without tags, this is either to not overwhelm the user since currently there is 6235 tags but 135 models). Then user have chance to press : and complete for tags for the model.
<!-- gh-comment-id:2492895731 --> @vt-alt commented on GitHub (Nov 22, 2024): My take, and I plan to package this for ALT Linux. ```bash # ollama(1) completion -*- shell-script -*- # vi: ft=sh sw=4 # SPDX-License-Identifier: Apache-2.0 # shellcheck disable=SC2034 _comp_ollama() { local cur prev words cword comp_args _comp_initialize -n =: -- "$@" || return words=("${words[@]:0:$cword+1}") if ((cword == 2)); then # Models listing commands. case "$prev" in pull) local models_list=/usr/share/ollama/models-list.txt local tags_list=/usr/share/ollama/tags-list.txt if [[ $cur == *:* ]]; then _comp_compgen_split -- "$(LANG=C look "$cur" "$tags_list")" _comp_ltrim_colon_completions "$cur" else _comp_compgen_split -- "$(LANG=C look "$cur" "$models_list")" if [[ ${COMPREPLY-} != "$cur" ]]; then # Allow user to enter ':' after the model name or a space if # another tab is pressed. compopt -o nospace fi fi ;; cp|push|rm|run|show) _comp_compgen_split -- "$(ollama list 2>/dev/null | tail -n+2 | grep -Eo '^\S+')" _comp_ltrim_colon_completions "$cur" ;; stop) _comp_compgen_split -- "$(ollama ps 2>/dev/null | tail -n+2 | grep -Eo '^\S+')" _comp_ltrim_colon_completions "$cur" ;; esac ((${#COMPREPLY[@]})) && return fi # Fallback to completing Cobra built-ins. local requestComp lastParam lastChar args args=("${words[@]:1}") requestComp="${words[0]} __completeNoDesc ${args[*]}" stParam=${words[$((${#words[@]}-1))]} lastChar=${lastParam:$((${#lastParam}-1)):1} if [[ -z $cur && $lastChar != = ]]; then requestComp="$requestComp ''" fi if [[ $cur == -*=* ]]; then cur="${cur#*=}" fi local out directive out=$(eval "$requestComp" 2>/dev/null) directive=${out##*:} out=${out%:*} if [[ $directive == "$out" ]]; then directive=0 fi local shellCompDirectiveError=1 local shellCompDirectiveNoSpace=2 local shellCompDirectiveNoFileComp=4 local shellCompDirectiveFilterFileExt=8 local shellCompDirectiveFilterDirs=16 local shellCompDirectiveKeepOrder=32 if (((directive & shellCompDirectiveError) != 0)); then return else if (((directive & shellCompDirectiveNoSpace) != 0)); then compopt -o nospace fi if (((directive & shellCompDirectiveKeepOrder) != 0)); then compopt -o nosort fi if (((directive & shellCompDirectiveNoFileComp) != 0)); then compopt +o default fi fi _comp_compgen_split -- "$out" } && complete -F _comp_ollama ollama ``` For the script to work you will need to have `/usr/share/ollama/{models,tags}-list.txt` files, created with something like https://github.com/ollama/ollama/issues/7751#issuecomment-2488290937 but they shall be properly sorted with `LANG=C sort` — this is mandatory for `look` to work correctly. (The second part with a fallback is based on [bash_completionsV2.go](https://github.com/spf13/cobra/blob/main/bash_completionsV2.go) from Cobra and is greatly shortened.) UI/UX considerations: 1. Only the `pull` shows remote models, not the `run` — this is to not overwhelm the user, perhaps, they will more frequently will want to run the local models. 2. First it completes only for model names without tags, this is either to not overwhelm the user since currently there is 6235 tags but 135 models). Then user have chance to press `:` and complete for tags for the model.
GiteaMirror commented 2026-04-28 18:24:28 -05:00
Author
Owner
Copy Link

@hey-august commented on GitHub (Feb 1, 2025):

Why was this closed as completed? The feature requested by @lucianoayres has not been implemented.

<!-- gh-comment-id:2628638261 --> @hey-august commented on GitHub (Feb 1, 2025): Why was this closed as completed? The feature requested by @lucianoayres has not been implemented.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label feature request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#51109
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?