Build Your Own AI Coding Assistant in 5 Minutes: A Hands‑On Guide

Current programming pain points

Repeatedly entering background information (project stack, coding standards, architecture) for each new AI chat.

Prompt length is long and complex; users must write dozens of lines to obtain acceptable code.

Context is lost after switching sessions; the model forgets previous preferences.

Answers are overly generic and often do not follow team‑specific conventions.

AI Agent concept

An AI Agent (AI Assistant) is a preset AI helper with a fixed role, knowledge background and behavior rules. It acts as a personal programming consultant that remembers the developer’s technology‑stack preferences, coding style, common frameworks, project background and development habits.

Core value of an AI Agent

Prompt length reduced to 20‑100 characters (instead of 200+).

Automatic context retention; no need to re‑enter background each time.

Domain‑specific expertise (e.g., ElasticStack) versus generic solutions.

Higher efficiency in daily coding queries.

Implementation overview (Coco AI platform)

Option A – DeepSeek API (recommended for beginners)

Open the Coco Server management UI and select the "Model Provider" menu.

Add the DeepSeek model, enable it (blue switch), and confirm the configuration.

Configure the assistant’s basic information: name, description, icon, tags (e.g., elasticsearch, kibana, search engine).

Select "Simple Mode" (sufficient for a coding assistant).

Choose the response model "deepseek-chat" from the dropdown.

Write the role prompt that defines the assistant as an ElasticStack expert (see code block below).

Option B – Local Ollama model (advanced users, optional)

# Install Ollama (Mac/Linux)
curl -fsSL https://ollama.com/install.sh | sh

# Pull the model
ollama pull qwen2.5:7b

# Configure Ollama address in Coco Server
http://localhost:11434

Creating the AI Assistant

Navigate to the "AI Assistant" menu and click "+ New".

Enter the assistant name "ElasticStack智能助手" and a brief description.

Set the icon and tags (e.g., elasticsearch, kibana, search engine).

Choose "Simple Mode" and select the DeepSeek model.

Paste the following role prompt:

# ElasticStack智能体

你是一位专业的ElasticStack运维专家，精通Elasticsearch、Logstash、Kibana和Beats组件的部署、配置与故障排查。你的核心能力包括帮助用户解决数据索引、搜索优化、日志分析和可视化等问题。

## 请基于官方文档和最佳实践，提供准确、可操作的指导，例如：
- 编写查询DSL
- 设计索引映射
- 调试Pipeline配置
- 构建Logstash过滤器
- 优化集群性能

## 请遵守以下规则：
1. 先抓住问题要点，再分步骤解答
2. 结果可附参考建议
3. 响应格式：先确认问题要点，再分步解答，结尾可附参考建议

Panel configuration

Enter the "Settings" menu, switch to the "Application Settings" tab, and enable the start‑page switch (blue).

In the "Chat Settings" section, enable the start page: 起始页：✅ 已启用（开关为蓝色） Enable the predefined assistant prompts (senior programmer, Python expert, Java expert, DBA/SQL performance, etc.):

全屏组件-摘要：✅
资深程序员：✅
Python专家：✅
Java专家：✅
DBA/SQL性能调优：✅

Using the assistant

After configuration, the assistant appears on the client home page. Example usage: 我的 ES 查询很慢，怎么优化？ The AI replies as an ElasticStack expert, providing concrete optimization steps.

Benefits summary

One‑time configuration, permanent effect – eliminates repeated prompts.

Domain‑expert responses – AI always answers with professional knowledge.

Efficiency boost – a 10‑character question replaces a 200‑character prompt.

Knowledge accumulation – prompts become documentation that can be continuously refined.

Future outlook – Retrieval‑Augmented Generation (RAG)

Integrating RAG technology can further improve the assistant by enabling real‑time document retrieval, code‑based recommendations, persistent context memory, and personalized learning of user habits.