Show HN:基于Ollama的开源文档AI
Show HN: Open-Source DocumentAI with Ollama

原始链接: https://rlama.dev/

RLAMA 是一款功能强大的本地文档问答工具,允许你使用本地 Ollama 模型创建和管理检索增强生成 (RAG) 系统。它可以索引各种格式的文档文件夹(文本、代码、PDF、DOCX 等)以构建知识库,确保所有处理都在本地进行,从而保护隐私和安全。 只需简单的命令,你就可以创建 RAG 系统(例如 `rlama rag llama3 documentation ./docs`),启动交互式查询会话 (`rlama run documentation`),列出可用的 RAG (`rlama list`),以及删除它们 (`rlama delete documentation`)。该工具还支持自我更新 (`rlama update`) 和显示版本号 (`rlama --version`)。RLAMA 专为开发者和技术用户设计,方便他们查询文档、进行研究或管理私有知识库。它还包含故障排除提示和支持的文件格式列表。

这个Hacker News帖子讨论了rlama.dev,一个使用Ollama进行本地LLM处理的开源DocumentAI工具。最初的反馈强调了一个关键限制:它缺乏文档分块功能,这阻碍了它有效处理大型文档(如书籍)的能力。用户指出,目前的实现加载整个文档,超过了嵌入模型的token限制,影响了检索精度。 开发者Dontizi回应说,目前正在测试带有重叠分块的实现,以提高大型文本的RAG性能。讨论还涉及到本地模型中较长的上下文提示需要大量资源的挑战,以及关于分层搜索方法和元数据使用的建议。 用户请求了一些功能,例如用于集成的API,以及架构文档,以及围绕文件系统访问的安全考虑。开发者透露了一个基于Go的技术栈,利用Cobra进行CLI,Ollama API进行LLM集成,以及本地文件系统进行存储,同时计划根据社区反馈进行未来的改进。

原文

🛠️ New

Introducing rlama-CLI

RLAMA

A powerful document question-answering tool that connects to your local Ollama models. Create, manage, and interact with RAG systems for all your document needs.

Available for macOS, Linux, and Windows

Document Indexing

Index any document folder for intelligent retrieval and querying.

Multi-Format Support

Support for text, code, PDF, DOCX, and many other document formats.

Local Processing

Process everything locally with Ollama models. No data leaves your machine.

Interactive Sessions

Create interactive RAG sessions to query your document knowledge base.

Easy Management

Simple commands to create, list, and delete your RAG systems.

Developer Friendly

Built with Go and designed for developers and technical users.

# Create a new RAG system named "documentation" using the llama3 model
# and indexing all documents in the ./docs folder
rlama rag llama3 documentation ./docs

# You'll see progress as documents are processed
Processing file: docs/installation.md
Processing file: docs/commands.md
Processing file: docs/troubleshooting.pdf
...
RAG system "documentation" created successfully!
PDF
DOCX
MD
Question about docs...

Technical Documentation

Query your project documentation, manuals, and specifications with ease.

Local
Private
Secure
100% local processing
No data leaves your computer

Private Knowledge Base

Create secure RAG systems for sensitive documents with full privacy.

Study
Research
Learn
Summarize key concepts...
Interactive learning sessions

Research Assistant

Query research papers, textbooks, and study materials for faster learning.

Command Reference

rag

Create a new RAG system from documents

rlama rag [model] [rag-name] [folder-path]
Example: rlama rag llama3 documentation ./docs
run

Start an interactive session with a RAG system

rlama run [rag-name]
Example: rlama run documentation
list

List all available RAG systems

rlama list
Example: rlama list
delete

Delete a RAG system

rlama delete [rag-name] [--force/-f]
Example: rlama delete old-project --force
update

Update RLAMA to the latest version

rlama update [--force/-f]
Example: rlama update
version

Display RLAMA version

rlama --version
Example: rlama -v

Troubleshooting

Common issues and their solutions

Supported File Formats

Text

.txt
.md
.html
.json
.csv
.yaml
.yml
.xml

Code

.go
.py
.js
.java
.c
.cpp
.h
.rb
.php
.rs
.swift
.kt

Documents

.pdf
.docx
.doc
.rtf
.odt
.pptx
.ppt
.xlsx
.xls
.epub

Ready to streamline your document question-answering?

联系我们 contact @ memedata.com