跳到正文
DEV Community· Krishna Kumar·· 3 小時前AI 評分31

OpenDesk AI:以開源 AI 讓機器中的幽靈重現

OpenDesk AI: Bringing the Ghost in the Machine to Life with Open-Source AI

AI 導讀

OpenDesk AI 是一款開源 AI 助手,可搜尋用戶的文件、以自然語言回答問題,並從個人知識庫擷取相關段落,生成有文件內容根據的答案。計劃採用 Gemma、文件檢索及本機推理;所需內容安裝完成後,設定妥當的版本可離線運作。本機推理有助讓資料留在用戶的電腦上,但仍取決於硬件,且離線使用前須先下載模型。

正文

I built OpenDesk AI, an open-source AI assistant that brings forgotten knowledge back to life.

Think of all the PDFs, project documentation, notes, and technical knowledge sitting untouched on your computer. They may look like ghosts from the past, but they still have answers to give.

OpenDesk AI helps users search their documents, ask questions in natural language, retrieve relevant passages, and get answers grounded in their own knowledge base.

For this week's Ghost in the Machine theme, I wanted to explore a simple idea: what if the knowledge we leave behind could speak again?

Code

GitHub: https://github.com/Krishnakumar7255/OpenDesk-AI-

How I Built It

My planned technology stack includes:

React + Tailwind CSS: Frontend interface.

FastAPI: Backend and API endpoints.

Gemma: Open-weight language model for generating responses.

Document retrieval: Extracting relevant passages from project documents and PDFs before generating answers.

Local inference: An opportunity to run the model on local hardware instead of depending entirely on a hosted AI API.

The retrieval pipeline helps connect answers to the user's documents rather than relying only on the model's general knowledge.

Why Does Open Innovation Matter?

For me, open innovation is about ownership, transparency, and freedom to experiment.

A document assistant can contain private notes, unpublished projects, internal documentation, and personal knowledge. An architecture that supports local inference can help keep this information on the user's own machine instead of sending every query to a third-party AI service.

Open-weight models also offer flexibility. Developers can experiment with different models, adapt their pipelines, inspect their behavior, and build without being tied to a single proprietary API.

A closed API can be convenient, but it may introduce usage costs, rate limits, and dependency on an external provider. An open approach gives developers more control over how their tools evolve.

Local inference still depends on hardware, and the model must be downloaded before offline use. But once everything required is installed, a properly configured version can work without an internet connection.

That's the kind of freedom I want to explore: AI that helps people rediscover their knowledge without taking control away from them.

The Spirit of It 👻

Every document contains a little piece of the past: an old idea, a forgotten solution, or a project someone thought nobody would revisit.

OpenDesk AI is my attempt to give that knowledge a voice again.

Because sometimes, the most interesting ghost in the machine isn't something supernatural. It's an idea we forgot we had.

Prize Categories

Open-Source AI / Open-Weight Models.

來源:DEV Community · dev.to