# AI use cases for Project Maya (mw00/project-maya) Repo: [mw00/project-maya](https://github.com/mw00/project-maya) | Use cases: 3 | Page: https://dedupedai.com/use-cases/mw00-project-maya/ Project Maya is a free, local engine and server that runs large GLM-5.3-Flash and DeepSeek-V4.1-Flash-1017 models privately on your own NVIDIA or AMD GPUs. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Run a private coding-agent model on your own GPU: https://dedupedai.com/use-cases/mw00-project-maya/private-local-coding-api/ - Ask questions about confidential files without cloud upload: https://dedupedai.com/use-cases/mw00-project-maya/analyze-sensitive-documents-offline/ - Run DeepSeek-V4.1-Flash locally with picture input: https://dedupedai.com/use-cases/mw00-project-maya/run-deepseek-v4-with-images/ ## How to run a private coding-agent model on your own GPU with Project Maya Repo: [mw00/project-maya](https://github.com/mw00/project-maya) | Difficulty: Advanced | Time: 3 hours | Basis: From the README Serve GLM-5.3-Flash locally so coding tools call it through an OpenAI- or Anthropic-compatible API. **Who it is for:** A developer whose team cannot send source code to a hosted model provider, and who already has a capable GPU PC. **The problem:** Hosted model APIs send your code and prompts off the machine, and your coding tools expect a standard API. **The outcome:** A local server that answers API requests from your existing tools, with tool calls, and no prompts leaving your PC. ### Steps 1. Clone the repo and run `./setup.sh --setup`, then pick the model the setup recommends for your PC (Maya-S is the default). 2. Run `./maya.sh --bench` to confirm the model runs and see your decode and prefill speeds. 3. Start the local server through the setup or maya.sh and note the address it reports. (inferred, check the README) 4. Point your OpenAI- or Anthropic-compatible client at that address and send one test request with a tool call. (inferred, check the README) ### Prompt to paste into your agent ```text Use the GitHub repo mw00/project-maya. Set up a local GLM-5.3-Flash server on using ./setup.sh --setup and ./maya.sh --bench, then send one OpenAI-compatible chat request with a tool definition to the local server. Expected output: the benchmark results and a valid chat response containing a tool call. Verify that the request was answered by the local server and not a remote API. ``` ### Caveat The README lists Maya-S as 96.5 GB and says the setup recommends a model for your PC, so check your RAM and VRAM before you start the download. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/mw00-project-maya/private-local-coding-api/ ## How to ask questions about confidential files without cloud upload with Project Maya Repo: [mw00/project-maya](https://github.com/mw00/project-maya) | Difficulty: Intermediate | Time: 2 hours 30 minutes | Basis: From the README Chat with a large model in the browser while your documents stay on your own machine. **Who it is for:** A researcher, lawyer or analyst holding confidential material who needs a strong model but cannot upload it to a third party. **The problem:** Strong large models are usually hosted, so sensitive text has to leave your control. **The outcome:** A browser chat with a local 300B-class model that works through your questions on documents that never leave the PC. ### Steps 1. Run `./setup.sh --setup` and pick a model that fits your memory (Maya-S24 suits cards of 24 GB or less). 2. Open the browser chat the setup starts and confirm it loads the local model. (inferred, check the README) 3. Paste or attach the document text in the chat and ask a question about it. 4. Turn thinking levels up for harder questions if the chat offers them for your model. ### Prompt to paste into your agent ```text Use the GitHub repo mw00/project-maya. Run ./setup.sh --setup on , choose Maya-S24, then open the local browser chat and ask a summary question about . Expected output: a summary generated locally. Verify that the network traffic during the answer stays on the local machine. ``` ### Caveat The README describes the browser chat but not its address or how to attach files, so the exact steps may differ on your version. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/mw00-project-maya/analyze-sensitive-documents-offline/ ## How to run DeepSeek-V4.1-Flash locally with picture input with Project Maya Repo: [mw00/project-maya](https://github.com/mw00/project-maya) | Difficulty: Advanced | Time: 4 hours | Basis: From the README Set up DeepSeek-V4.1-Flash-1017 on your GPUs and include vision support for images. **Who it is for:** An engineer with a multi-GPU or large-memory machine who wants a second frontier-class model and needs image understanding. **The problem:** Few open models of this size run on personal hardware, and many do not handle pictures. **The outcome:** A local DeepSeek-V4.1-Flash-1017 server that answers text and picture prompts, with its own numbers in the monitor. ### Steps 1. Run `./setup.sh --model DeepSeek-V4.1-Flash-1017` to set up the DeepSeek engine (maya-ds4). 2. Leave vision on for the GGUF files, since the setup fetches DeepSeek's vision tower; use `--no-vision` only if you want to skip it. 3. Run `./maya.sh --bench` to measure speed on your GPUs. 4. Send one request that includes an image and check the answer describes the picture. (inferred, check the README) ### Prompt to paste into your agent ```text Use the GitHub repo mw00/project-maya. On , run ./setup.sh --model DeepSeek-V4.1-Flash-1017 and then ./maya.sh --bench. Expected output: benchmark numbers for the DeepSeek model and a successful image-plus-text answer. Verify that the vision component loaded (the setup should report it) before counting the test as passed. ``` ### Caveat The README says the GGUF files total 340 GB or more, so this setup needs a large NVMe SSD and enough RAM plus VRAM before the first run. Risk, reverse engineering: The DeepSeek-V4.1-Flash-1017 GGUF files and the optional DSpark drafter are large third-party or repo-hosted model weights, so verify checksums and licenses before relying on them. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/mw00-project-maya/run-deepseek-v4-with-images/