chat-cli-anything 0.1.8

Chat with anything on cli.

# Chat-Cli-Anything

Interact with GPT-like services from the command line, supporting local RAG, and pipe input.

Why:
① Developers spend a significant amount of time using the terminal to interact with the system, and when encountering issues in the terminal, they expect not to have to switch to a search engine or elsewhere.
② By leveraging the language model's excellent text capabilities and text processing abilities, it helps us achieve code generation, explanation, and translation beyond simple searches, as well as document Q&A.

## Tutorial

1. Installation

```shell
pip install "chat-cli-anything"

To simplify command length, you can set command aliases (optional):

# put this in your ~/.bashrc or ~/.zshrc
alias config="cc-config"
alias code="cc-code"
alias chat="cc-chat"
alias db="cc-db"
alias service="cc-service"

2. Add LLM provider

cc-config add "openai" "https://api.openai.com/v1"  --api-key "your-api-key"

If command aliases are set:

config add "openai" "https://api.openai.com/v1"  --api-key "your-api-key"

3. Start Using

example 1 Normal question

cc-ask "Who is the author of the science fiction novel 'Three Body'?"

cc-ask "How to get the memory size occupied by a process in Linux?"

example 2 Using pipe

# Find the filename with the maximum size in the current directory
ls -lah | cc-ask "filename with max size"

# For debugging error
gcc -o a.out a.cpp 2>&1 | cc-ask "what is the error" -s

example 3 Interactive questioning

cc-chat "What is the capital of France?"

example 4 Local document Q&A

cc-ask "Which school did the candidate graduate from?" -f resume.pdf

example 5 Local document collection

Step 1: Digest text to build a local database

# d2light code
cc-db ingest "/path/to/detectron2-light" -n d2light

Step 2: Ask a question

# For open questions, the '-a/--advance' option would be helpful
cc-ask -d d2light "DETR implementation" -a

example 6 Code explanation

# Explain line by line
cc-code explain /path/to/your/code.py -l

# If the code is lengthy, add the -l parameter to continue generating
cc-code explain /path/to/your/code.py -l -c

Notice

It depends on torch version 2.2, so it may override the current environment version.

Command Explanation

cc-config

Configure LLM provider

1. Add LLM provider

cc-config add [OPTIONS] NAME BASE_URL

Options:
  --model TEXT
  --api-key TEXT
  --proxy TEXT
  --help          Show this message and exit.

Example:

cc-config add "openai" "https://api.openai.com/v1"  --model "gpt-3.5-turbo-1106" --api-key "sk-xxxxxxxxxxx"

2. Test provider

cc-config ping [OPTIONS] NAME

Ping provider.

Options:
  --help  Show this message and exit.

3. List all providers

cc-config list [OPTIONS]

List all configurations.

Options:
  -s, --show-api-key
  --help              Show this message and exit.

The default api-key is hidden; use -s/--show-api-key to display the key.

4. Remove provider

cc-config remove [OPTIONS] [NAME]

Remove a configuration.

Options:
  --help  Show this message and exit.

5. Switch active provider

ask.py cc-config switch [OPTIONS] NAME

Switch to a different configuration and save the change.

Options:
  --help  Show this message and exit.

6. Load configuration file

Import configurations.

Options:
  -o, --override  Whether to override the original config.
  --help          Show this message and exit.

7. Export configuration file

ask.py cc-config dump [OPTIONS] PATH

Export current configurations.

Options:
  -o, --override  Whether to override the existing file.
  --help          Show this message and exit.

cc-ask

Perform a question-answering task

cc-ask [OPTIONS] QUERY

Start a chat session with the given query.

Options:
  -d, --db TEXT        Name of database.
  -f, --filename TEXT  Name of file.
  -r, --rerank         Whether to rerank the results.
  -s, --show-chunks    Whether to show the related chunks retrieved from
                       the database.
  -a, --advance        Whether to use advance RAG.
  --help               Show this message and exit.

cc-chat

Interactive Q&A

cc-chat [OPTIONS] [QUERY]

Interactive chat. Enter '/quit' to exit.

Options:
  -d, --db TEXT          Name of the database.
  -f, --filename TEXT    Name of the file.
  -n, --not-interactive
  -s, --show-history
  -c, --clear
  --help                 Show this message and exit.

cc-code

Some common commands for code based on cc-ask.

Code Explanation

cc-code explain [OPTIONS] [FILENAME]

Explain code.

Options:
  -o, --object-name TEXT
  -l, --line               Line by line.
  -c, --continue-generate
  --help                   Show this message and exit.

Example 1: Specify filename

cc-code explain some_complex_scripts.py

Example 2: Specify a particular function

cc-code explain some_complex_scripts.py -o SomeClass::some_function

Example 3: Use pipe

cat some_complex_scripts.py | cc-code explain

Fix Issues

cc-code fix [OPTIONS] [FILENAME]

Fix code.

Options:
  -o, --object-name TEXT
  -c, --continue-generate
  --help                   Show this message and exit.

Code Refactoring

cc-code refactor [OPTIONS] [FILENAME]

Refactor code.

Options:
  -o, --object-name TEXT
  -c, --continue-generate
  --help                   Show this message and exit.

Code Review

cc-code review [OPTIONS] [FILENAME]

Review code.

Options:
  -o, --object-name TEXT
  -c, --continue-generate
  --help                   Show this message and exit.

Code Translation

cc-code translate [OPTIONS] LANGUAGE [FILENAME]

Translate code from one language to another. Supported languages include c++, cpp,
c, rust, typescript, javascript, markdown, html.

Options:
  -o, --object-name TEXT
  -c, --continue-generate
  --help                   Show this message and exit.

Select Code from Generated Results

Used to select code snippets from the generated responses of the above commands

cc-code select [OPTIONS] [INDEX]

Select code snippet from the last output.

Argument:     index: code snippet index

Options:
  -c, --count  get the number of code snippets
  --help       Show this message and exit.

-c/--count Get the number of candidate code blocks in the last response

Example:

>>> cc-code select -c
2

# Select the second code block from the last output
# For macOS
>>> cc-code select 1 | pbcopy
# For Linux
>>> cc-code select 1 | xclip -selection clipboard

cc-db

list: List all text collections

cc-db list [OPTIONS] [NAME]

List all document databases.

Options:
  -s, --short  List in short format.
  --help       Show this message and exit.

ingest: Digest documents

cc-db ingest [OPTIONS] [FILES]...

Read documents and convert them into a searchable database.

Options:
  -n, --name TEXT     The name of the knowledge base.
  -m, --comment TEXT  Add comment to info.
  --help              Show this message and exit.

remove: Remove document collection

cc-db remove [OPTIONS] NAME

Remove database with the given name.

Options:
  -d, --remove-documents  Remove documents if data.
  --help                  Show this message and exit.

search: Search within a document collection

cc-db search [OPTIONS] DB QUERY

Options:
  -k, --topk INTEGER  Number of top results to return.
  --help              Show this message and exit.

cc-service

Used to manage local text-to-index services.

If you want to run a text-to-index service on your local machine, execute the command "pip install chat-cli-anything[all]".

start:

cc-service start [OPTIONS]

Options:

--help  Show this message and exit.

This process may take some time (~1min) to load the model, and you can check if it has started with cc-service status.

stop

cc-service stop [OPTIONS]

Options:
  --help  Show this message and exit.

status

cc-service status [OPTIONS]

Options:
  --help  Show this message and exit.