Final update before the deadline
Since my last update, something much bigger has been taking shape behind TenderGPT Edge.
I originally set out to build practical AI that could run on ordinary laptop hardware. But while building TenderGPT Edge, I kept asking a bigger question:
What if the intelligence layer itself could become something we build, inspect, test, and improve locally?
That project became TenderLM.
Since the last update, TenderLM has grown into a tested intelligence system that can now connect directly with TenderGPT Edge.
What has changed:
TenderLM → TenderGPT Edge integration is now working
A real tender document can move through the TenderGPT Edge pipeline and into TenderLM for additional intelligence processing.
Fresh document analysis
This is not just replaying an old result. The system has demonstrated fresh document analysis with new run IDs and preserved evidence provenance.
Tender intelligence capabilities
TenderLM now includes infrastructure for:
• requirement extraction and classification • mandatory requirement analysis • expanded tender taxonomy • contradiction detection • evidence references and provenance • human review workflows • real-world accuracy measurement infrastructure • controlled knowledge onboarding • source hashing and duplicate detection • source verification and lifecycle management
And one principle has remained non-negotiable:
No fake accuracy. No invented knowledge. No silent training contamination.
If real-world accuracy has not yet been established through human-reviewed ground truth, the system says:
NOT_AVAILABLE.
It does not pretend.
The really exciting part for me is where this can go next.
TenderLM is now being built with the foundation to gradually absorb verified knowledge across tenders, procurement, government requirements, regulations, evaluation criteria, supplier information, and related domains — while keeping source provenance and verification separate from model training.
So while TenderGPT Edge is the application people can interact with, there is now a deeper intelligence engine evolving behind it.
We're not claiming to have solved everything.
We're showing the work.
Tested systems. Real document processing. Evidence trails. Human review. Controlled knowledge growth. And now, a live connection between TenderGPT Edge and TenderLM.
With the deadline approaching, I'm excited about where this could go.
Because this may be only the beginning.
TenderGPT Edge + TenderLM.
Practical tender intelligence. Built to run on ordinary hardware. Built transparently. Built to grow.
Log in or sign up for Devpost to join the conversation.