DLLO
DISTRIBUTED LLM OBSERVATORY
An open-source, evidence-first framework for designing agent configurations, evaluating real agent behavior, and comparing reproducible AI observations across time and regions.
Browser workflows run locally through the DLLO Agent Lab bridge.
THREE WORKFLOWS · ONE EVIDENCE MODEL
DLLO keeps requirements, evidence, evaluation, provenance and interpretation explicit throughout the workflow.
AGENT STARTER
Turn goals, constraints, hardware evidence and explicit preferences into an evidence-backed architecture and concrete stack recommendation.
TEST YOUR AGENT
Connect an agent through the Local SUT Protocol and evaluate observable behavior using observer-owned expectations and evidence.
OBSERVATORY
Compare sufficiently compatible observations across time and geographic regions while preserving provenance and rejected comparisons.
EVIDENCE FIRST
Missing evidence remains unknown instead of becoming a negative claim.
DLLO does not invent a global winner when the evidence supports multiple valid outcomes.
The system under test does not own the expectations used to certify its behavior.
Observations preserve what was measured, where, when and under which benchmark context.
PUBLIC PREVIEW