[Side Project] FlexiAI Toolsmith + 2 Quick Demos – Seeking Feedback & Testers

API
, , , ,
1

Hi everyone, I’m Razvan. I’ve built FlexiAI Toolsmith as a side project: a multi-channel Python framework for AI chat assistants with built-in tools like web forms, spreadsheets/CSV operations, YouTube search, security audits, and more. It’s still in early bootstrap and needs plenty of refactoring, so I’d love your feedback before creating a full tutorial series.

Demo Videos:

Note: at the moment I’ve only exposed the Python implementations of each tool—assistant-side instructions and JSON-function specs are coming soon.

If you’re interested in testing the repo (SavinRazvan/flexiai-toolsmith) and trying out agents that use forms, spreadsheets/CSV, OCR (coming soon), etc., please reply here or DM me. Your input will help me decide if it’s worth investing time in detailed video tutorials.

Thanks in advance for your thoughts!

2

I took a look at this and I noticed a few things:

  • The optional .env variables will only be optional if you set them as empty strings.
  • The web interface seems to be calling asst_SbnReaK3gT3h2V19X1cJKXqk but we don’t have access to your org.
  • The Assistants API isn’t technically deprecated, but OpenAI publicly stated that it’s about to be, so it’s still effectively deprecated and it’s inadvisable to rely on it.

Otherwise, it looks like a pretty cool project! The UI homepage looks very nice.

1 Like
3

  • You’re correct. My immediate next steps are:

    • Moving Assistant ID, Assistant Name, and User Name into the .env file to enhance configurability.

    • Creating a detailed video tutorial showing how to craft an Assistant, including:

      • Defining effective assistant instructions,
      • Designing advanced tool specifications,
      • Demonstrating the setup of flexible JSON-defined tools.

    My goal here is to make it easier for everyone to experiment and later transition smoothly to a more robust Dynamic Multi-Agent System rather than relying on the current static setup.

Regarding the Assistants API deprecation (OpenAI FAQ), I agree it’s problematic because:

  • Other competitors like DeepSeek and Qwen are rapidly building similar capabilities. OpenAI’s decision may inadvertently push developers toward these alternatives.
  • It might be a strategic move to control the flexibility we currently enjoy with multi-agent alignment and experimentation, or perhaps just a business pivot—though it certainly feels restrictive.

However, even if OpenAI sunsets the Assistants API, it won’t fundamentally block development, as we can pivot to open-source alternatives. I’m already exploring wrappers around local models (like those in LM Studio) to achieve similar tool-calling functionality and create a comparable or even superior Assistants framework independently.

1 Like
4

Incorrect—Pydantic only makes a field optional if you declare it Optional[...] with a default (and, if you want to ignore KEY= entries, enable env_ignore_empty=True); merely writing KEY= in .env doesn’t grant optionality.

5

I tried the default value and also tried commenting it out. Both times got errors. Don’t know what else to tell ya. ¯_(ツ)_/¯