ggml-org/llama.cpp

llama : add CLI assistant

Open

#10,688 opened on Dec 6, 2024

View on GitHub
 (4 comments) (3 reactions) (0 assignees)C++ (110,169 stars) (18,202 forks)batch import
enhancementgood first issueroadmap

Description

The https://github.com/AnswerDotAI/shell_sage project seems quite fun and useful. GitHub Copilot CLI was also a rather useful utility.

We should build a fully-local alternative leveraging llama-server's advanced context reuse techniques in the same spirit as llama.vim. Speculative decoding should be very effective in this scenario as well. Likely the Qwen 2.5 family would be ideal for this use case.

Contributor guide

Tech stack
cppshell
Domain
clibackend
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
4
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
over 1 week
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
fresh
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
needs investigation
Prerequisites
C++ knowledgeshell scriptingfamiliarity with llama.cpp
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
20
Research direction
The issue proposes building a CLI assistant similar to shell sage and GitHub Copilot CLI, leveraging llama server's context reuse and speculative decoding. To start, examine the existing llama.vim implementation and the shell sage project for design patterns. Investigate llama server's API for context reuse and how to integrate speculative decoding. The Qwen 2.5 family is suggested as a model choice. The exact CLI interface and command structure need to be defined, possibly through a community discussion.
llama : add CLI assistant · ggml-org/llama.cpp#10688 | Good First Issue