gini

README

gini

CLI for simple i/o with Google Gemini AI model

disclaimer

The use of this tool does not guarantee security or usability for any particular purpose. Please review the code and use at your own risk.

Don't trust, verify

installation

This step assumes you have Go compiler toolchain installed on your system.

go install github.com/kubetrail/gini@latest

Get an API key from Google AI studio and setup an env variable GOOGLE_API_KEY for it.

Setup shell completion. See more info at:

gini completion -h

For instance, setup bash completion by adding following line to your .bashrc

source <(gini completion bash)

usage

gini chat [--auto-save]

--auto-save flag will save chat history to a randomly generated filename.

example chat history

gini chat --auto-save

please type prompt below
press enter twice to send prompt
just enter to quit
[1]>>> hi, could you please generate a floating point number between 0.115 and 0.117

      0.115997                     

[2]>>> could you please generate a list of 10 such numbers

      • 0.115023                   
      • 0.115193
      • 0.115316
      • 0.115408
      • 0.115628
      • 0.115746
      • 0.115859
      • 0.115997
      • 0.116031
      • 0.116173

      These are all floating point numbers between 0.115 and 0.117, generated
      randomly.

[3]>>> finally, could you please tell me the average of this list of numbers

      The average of the list of numbers
      • 0.115023
      • 0.115193
      • 0.115316
      • 0.115408
      • 0.115628
      • 0.115746
      • 0.115859
      • 0.115997
      • 0.116031
      • 0.116173

      is 0.115689.

      To find the average, we add up all the numbers and divide by the total
      number of numbers:

      (0.115023 + 0.115193 + 0.115316 + 0.115408 + 0.115628 + 0.115746 +
      0.115859 + 0.115997 + 0.116031 + 0.116173) / 10 = 1.15689 / 10 = 0.115689

      Therefore, the average of the list of numbers is 0.115689.

[4]>>> 
history saved to history-0d9d6887-ce12-4e89-824d-91b87b1a636f.txt

The prompt detects a blank line as termination, therefore, in order to send a prompt that has blank lines in it, start the prompt with double curly braces {{ and end with }} as shown below.

gini chat 
please type prompt below and press enter twice to send it
see more info in gini chat --help
just enter to quit
[1]>>> {{
hi, can you explain to me this code below

package main

func main() {
  fmt.Println("hello world")
}

}}
      The code below is a simple "Hello, world!" program written in Go. It
      prints the string "hello world" to the standard output.

      ┃ package main
      ┃ 
      ┃ import "fmt"
      ┃ 
      ┃ func main() {
      ┃     fmt.Println("hello world")
      ┃ }

      Here's a breakdown of the code:
      • package main: This line declares the package name for the program. The
        main package is the entry point for the program.
      • import "fmt": This line imports the fmt package, which provides
        functions for formatted I/O.
      • func main() { ... }: This is the main function of the program. It's the
        entry point for the program, and it's where the program execution
        begins.
      • fmt.Println("hello world"): This line prints the string "hello world" to
        the standard output (usually the console or terminal window where the
        program is running).

      When you run the program, you should see the output:

      ┃ hello world

image analysis

Images can be analyzed using a combination of raw image data and associated text prompt. Below is an example:

gini analyze image \
  --file seagull-on-a-rock.jpg \
  --format image/jpeg \
  could you please detect objects in the provided image

 A seagull standing on a rock in front of a blurred background of green plants.
 The seagull is gray and white with a yellow beak. The rock is brown and the background is green.

Please note that image analysis is conducted using gemini-pro-vision model by default. Furthermore, please make sure --formats match corresponding image format, or leave it blank if all images are jpeg images.

list models

Following models can be selected when performing a task. Select model by via --model flag using its name. For example gini chat --model=models/gemini-pro-vision etc.

gini list models

name: models/gemini-pro
basemodeid: ""
version: "001"
displayname: Gemini Pro
description: The best model for scaling across a wide range of tasks
inputtokenlimit: 30720
outputtokenlimit: 2048
supportedgenerationmethods:
    - generateContent
    - countTokens
temperature: 0.9
topp: 1
topk: 1

name: models/gemini-pro-vision
basemodeid: ""
version: "001"
displayname: Gemini Pro Vision
description: The best image understanding model to handle a broad range of applications
inputtokenlimit: 12288
outputtokenlimit: 4096
supportedgenerationmethods:
    - generateContent
    - countTokens
temperature: 0.4
topp: 1
topk: 32

name: models/embedding-001
basemodeid: ""
version: "001"
displayname: Embedding 001
description: Obtain a distributed representation of a text.
inputtokenlimit: 2048
outputtokenlimit: 1
supportedgenerationmethods:
    - embedContent
    - countTextTokens
temperature: 0
topp: 0
topk: 0

safety

--allow-harm-probability flag is set to negligible to prevent output from displaying content that could be harmful. Change it at your own risk, for example,

gini chat --allow-harm-probability=medium --auto-save

advanced config

Model config params such as --top-p, --top-k, --temperature, --candiate-count and --max-output-tokens can be supplied for fine tuning

Documentation

Overview

Copyright © 2023 kubetrail.io authors

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.