ROCm/FastFlowLM v0.1.2

🔧 FastFlowLM v0.1.2 — New API Support and Stability Improvements

on GitHub

🔧 FastFlowLM v0.1.2 — API Support and Stability Improvements

We’re pleased to announce v0.1.2 of FastFlowLM, focused on extending API compatibility and improving output stability in server mode.

✨ What's New

1. Initial Support for OpenAI-Compatible API

   • Added basic support for the /v1/api/completions endpoint.
     This opens the door for broader integration with OpenAI-compatible tools and clients.

2. Unicode Streaming Fix

   • Resolved a critical bug where Unicode characters could break the output stream in serve mode.

3. Minor Cleanup

   • Fixed several typos across logs, prompts, and documentation.

✅ Reminder: Usage Examples

Interactive Mode:

flm run llama3.2:1B

Server Mode (default model is llama3.2:1B):

flm serve

Now supports /v1/api/completions in addition to native REST format.