The question behind the work
Different models can bring different strengths to the same question. The council workflow explores how to collect those perspectives and make the review process explicit.
MULTI-MODEL WORKFLOWS
Several perspectives. One synthesis.
An exploration of local model orchestration: generate independently, review anonymously, then synthesize a final response.
THE PROCESSINPUT → UNDERSTANDING
01Generate
Collect answers from several local models.
02Review
Ask models to rank anonymized responses.
03Aggregate
Combine the peer rankings.
04Synthesize
Let a chairman model assemble the final response.
THE PROJECT
Ollama Council
Source and context from my portfolio
How it comes together
Python, JavaScript and Ollama connect generation, anonymous peer review and synthesis. The repository contains the implementation and any upstream attribution.
What to keep in mind
A council is a workflow for comparing answers. Its final output still needs evaluation; agreement between models does not establish correctness.
Follow the conversation.
Step through an illustration of the council workflow. These are sample responses, not live model inference.
SAMPLE QUESTION How should a team evaluate a multilingual summarizer?