chatsong, developed by Jingangdidi, is a lightweight portable AI chat client for prompt testing and private sessions. It functions as a standalone gateway to large language model APIs, forwarding prompts to configured endpoints and letting users compare responses across providers. The app emphasises single-file portability and local-first conversation storage. Designed for developers, prompt engineers, and privacy-conscious power users, it provides a compact environment for rapid model experimentation while keeping data on the user's machine.

What tasks can you actually use it for?

The app acts as a client gateway to multiple LLM endpoints, making it suitable for prompt testing, model comparison, and iterative prompt engineering. It supports document-assisted queries by parsing ZIP archives and ingesting HTML, PDF, and plain text files so you can query file contents directly. Integrated web search augments responses with recent information, Markdown rendering preserves code and structured output, and export to HTML produces clean Q&A records for sharing.

How accurate are the outputs compared to doing it manually?

The client forwards prompts to the endpoints you configure, so generated responses reflect the behaviour of the selected model and provider rather than the client itself. Because users supply API keys, model choice and verification remain the user's responsibility. Web search can supply up-to-date context, but factual claims from generated text require independent checking for technical or high-stakes tasks. Incognito mode creates temporary sessions that do not persist to local history.

What file formats does it support and what are the input limits?

The tool parses ZIP archives and extracts member files, and it accepts HTML, PDF, and plain text documents for contextual queries. It runs as a standalone binary across Windows, macOS, and Linux, and it requires an internet connection to reach external model APIs and perform web searches. Updating typically involves downloading the latest executable from the project's repository and replacing the existing file in the working folder.

Does it fit into a developer or privacy-focused workflow?

Configuration uses a plain text settings file for API keys and endpoint management, which lets developers script or version-control connection details. The client uses a much smaller system footprint compared with browser or Electron alternatives, making it practical for quick tests on modest machines. The project is hosted on GitHub and has positive reception among developer users, reflecting adoption by prompt engineers and power users who prioritise control over configuration and data handling.

Practical choice for developers who accept model responsibility

The app is a pragmatic option for developers and prompt engineers who need a compact client and accept that output quality depends on their chosen model and supplied API keys. For rigorous use, pair sessions with manual verification or the integrated web search to confirm facts. This tool suits workflows that prioritise local configuration control and rapid switching between external endpoints rather than embedded model services.