Local LLM / Research Automation

research-local-xiaomi

Local research workflow for turning web evidence into structured reports with planning, critique, usage tracking, and reproducible outputs.

Public 2026
TypeScript CLI local LLM web search structured reports

Shows research pipeline design around local models and agentic review loops.

Code

Context

research-local-xiaomi is a local research command-line workflow around Xiaomi MiMo and OpenCode-style web search. It is framed as a reproducible research pipeline, not as a chat UI.

Problem

Research tasks often fail because the search plan, evidence, critique, and final output live in different places. The project keeps those steps explicit so the output can be inspected and repeated.

Constraints

  • Local-first workflow with no hosted backend requirement.
  • Structured output instead of free-form notes.
  • Usage tracking and critique are part of the workflow, not afterthoughts.

Approach

The workflow separates planning, evidence gathering, critique, and report generation. That makes it easier to see where a claim came from and where the model needs a review pass.

Implementation highlights

  • CLI-oriented flow for repeatable runs.
  • Planning and critique stages before final report generation.
  • Output structure designed for technical review.
  • Local model workflow used as a practical constraint.

Result

The project demonstrates how Antonii designs research automation around local models, structured artifacts, and agentic review loops.

What it demonstrates

Local LLM workflow design, evidence-oriented prompting, CLI product thinking, and practical research automation.

01Task brief
02Search plan
03Evidence gathering
04Critique loop
05Usage tracking
06Structured report