# AI use cases for Atria Dawn (atria-asi/Atria-Dawn)

Repo: [atria-asi/Atria-Dawn](https://github.com/atria-asi/Atria-Dawn) | Use cases: 4 | Page: https://dedupedai.com/use-cases/atria-asi-atria-dawn/

Atria Dawn is a 744B-parameter MoE agentic model from the Shanghai Artificial Intelligence Laboratory, with hosted API access and config recipes for Codex, Claude Code, and Kimi Code.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Run Codex CLI on Atria Dawn as a text-only agent: https://dedupedai.com/use-cases/atria-asi-atria-dawn/run-codex-text-only-atria-dawn/
- Block Claude Code from reading PDFs and images: https://dedupedai.com/use-cases/atria-asi-atria-dawn/block-pdf-image-reads-in-claude-code/
- Connect Kimi Code to Atria Dawn over Chat Completions: https://dedupedai.com/use-cases/atria-asi-atria-dawn/connect-kimi-code-to-atria-dawn/
- Serve Atria Dawn on your own hardware with SGLang or vLLM: https://dedupedai.com/use-cases/atria-asi-atria-dawn/serve-atria-dawn-locally-with-sglang-or-vllm/

## How to run Codex CLI on Atria Dawn as a text-only agent with Atria Dawn

Repo: [atria-asi/Atria-Dawn](https://github.com/atria-asi/Atria-Dawn) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Point Codex at the Atria endpoint and stop it from sending images to the model.

**Who it is for:** A developer who uses Codex CLI for coding tasks and wants Atria Dawn as the model without image input errors.

**The problem:** By default Codex attaches images, and the Atria endpoint rejects them with a 400 error because Atria Dawn accepts text only.

**The outcome:** Codex runs against Atria Dawn over the Responses API with image input disabled on the client side.

### Steps

1. Create ~/.codex/atria-catalog.json with the model entry from the README: slug Atria Dawn, context_window 256000, max_context_window 256000, and input_modalities ["text"].
2. Add model_provider = "atria", base_url = "https://api.atria-asi.ai/v1", env_key = "ATRIA_API_KEY", and wire_api = "responses" to ~/.codex/config.toml.
3. Add model_catalog_json = "~/.codex/atria-catalog.json" and set features.view_image = false in the same config file.
4. Set the ATRIA_API_KEY environment variable to your Atria key.
5. Run `codex -m Atria-Dawn` and send a plain text prompt. (inferred, check the README)

### Prompt to paste into your agent

```text
Using the atria-asi/Atria-Dawn README, configure Codex CLI to use the model Atria-Dawn through the provider at https://api.atria-asi.ai/v1 with text-only input. Write ~/.codex/atria-catalog.json and update ~/.codex/config.toml as described. Expected output: a Codex session that answers a text prompt without a 400 error. Verify that the catalog lists Atria-Dawn with input_modalities set to text.
```

### Caveat

model_catalog_json replaces the whole model catalog, so any other model you switch to must also be listed in the same file or it will assume multimodal input.

Risk, credentials: Your ATRIA_API_KEY grants access to the hosted endpoint, so keep it in an environment variable and out of shared config files.

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/atria-asi-atria-dawn/run-codex-text-only-atria-dawn/

## How to block Claude Code from reading PDFs and images with Atria Dawn

Repo: [atria-asi/Atria-Dawn](https://github.com/atria-asi/Atria-Dawn) | Difficulty: Intermediate | Time: 15 minutes | Basis: From the README

Add a PreToolUse hook that denies the Read tool on PDF and image files.

**Who it is for:** A developer running Claude Code with Atria Dawn who needs the agent to stay text only.

**The problem:** Claude Code's Read tool can open PDFs and images, which Atria Dawn cannot process.

**The outcome:** Read calls on PDF and common image files are denied before they run, and the agent is told why.

### Steps

1. Save the README's PreToolUse Python script to ${Target_dir}/block_pdf_image_read.py, using an absolute path for ${Target_dir}.
2. Add the "hooks" block from the README to ~/claude_dir/settings.json, with matcher "Read" and the command python3 ${Target_dir}/block_pdf_image_read.py.
3. Start a Claude Code session in a project and ask it to read a PDF file. (inferred, check the README)
4. Confirm the hook returns a deny decision with the reason about PDF files. (inferred, check the README)

### Prompt to paste into your agent

```text
Using the atria-asi/Atria-Dawn README, install the Claude Code PreToolUse hook that blocks Read on PDF and image files. Save the script to <target-dir>/block_pdf_image_read.py and register it in ~/claude_dir/settings.json. Expected output: a denied Read call for a PDF path. Verify by asking Claude Code to read a .png file and checking that the denial reason appears.
```

### Caveat

The hook only checks file extensions on the Read tool, so it will not catch images or PDFs read through other tools.

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/atria-asi-atria-dawn/block-pdf-image-reads-in-claude-code/

## How to connect Kimi Code to Atria Dawn over Chat Completions with Atria Dawn

Repo: [atria-asi/Atria-Dawn](https://github.com/atria-asi/Atria-Dawn) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Add an Atria provider and model entry to Kimi Code's user config.

**Who it is for:** A developer who uses Kimi Code and wants Atria Dawn as an OpenAI-compatible model.

**The problem:** Kimi Code needs a provider entry with the right base URL, model name, and context setting before it will load the model.

**The outcome:** Kimi Code lists atria/Atria-Dawn as its default model and sends requests to the Atria endpoint.

### Steps

1. Run `kimi --version` and confirm you have version 0.43.1 or later.
2. Back up your config with `cp -a ~/.kimi-code/config.toml ~/.kimi-code/config.toml.bak`.
3. Merge the README's default_model, [providers.atria], and [models."atria/Atria-Dawn"] blocks into ~/.kimi-code/config.toml, with base_url https://api.atria-asi.ai/v1.
4. Put your atr_ key on the api_key line as a literal value.
5. Start Kimi Code and send a short text prompt to confirm it responds. (inferred, check the README)

### Prompt to paste into your agent

```text
Using the atria-asi/Atria-Dawn README, configure Kimi Code 0.43.1 or later to use atria/Atria-Dawn through https://api.atria-asi.ai/v1. Back up ~/.kimi-code/config.toml first, then merge the provider and model blocks. Expected output: a working text reply from Atria Dawn. Verify that default_model is set above the tables and that the context setting is present.
```

### Caveat

The README says the api_key value is read literally, with no ${VAR} expansion and no OPENAI_API_KEY fallback.

Risk, credentials: The api_key line holds your Atria key as a literal value in a plain config file, so restrict file permissions on that config.

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/atria-asi-atria-dawn/connect-kimi-code-to-atria-dawn/

## How to serve Atria Dawn on your own hardware with SGLang or vLLM with Atria Dawn

Repo: [atria-asi/Atria-Dawn](https://github.com/atria-asi/Atria-Dawn) | Difficulty: Advanced | Time: 2 hours | Basis: From the README

Run the model locally using a supported SGLang or vLLM version.

**Who it is for:** A team or researcher that needs to host Atria Dawn in its own environment instead of calling the hosted API.

**The problem:** Hosted access may not fit data handling or cost needs, so the model has to run locally on a supported inference framework.

**The outcome:** An inference server runs Atria Dawn locally and accepts requests.

### Steps

1. Choose SGLang at v0.5.13.post1 or later, or vLLM at v0.23.0 or later.
2. Open the SGLang cookbook link for GLM-5.2 or the vLLM recipes link for zai-org/GLM-5.2 from the README.
3. Install the chosen framework at the minimum version and follow its GLM-5.2 guide to load the model. (inferred, check the README)
4. Check the Download table: Atria Dawn Preview and Preview FP8 are on Hugging Face and ModelScope, while the full release says Coming soon.

### Prompt to paste into your agent

```text
Using the atria-asi/Atria-Dawn README, set up a local inference server for Atria Dawn with SGLang v0.5.13.post1 or later, or vLLM v0.23.0 or later, following the linked guide. Choose a model from the download table that is currently available. Expected output: a running local server that answers one test request. Verify the framework version before loading the model.
```

### Caveat

The README gives no hardware requirements, and the full Atria Dawn and FP8 releases are listed as Coming soon, so only the Preview models can be downloaded now.

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/atria-asi-atria-dawn/serve-atria-dawn-locally-with-sglang-or-vllm/