Step 1 — Domain Detection
Classify the query into domains using:
- explicit keywords
- semantic similarity
- Obsidian tag scanning (#financial,medical,garden,energy,security,general)
Step 2 — Retrieve Models
Scan the vault for notes tagged with the domain tag(s).
Step 3 — Evidence Weighting
Assign weights:
- Tier 1 (evidence-backed): 1.0
- Tier 2 (domain-relevant): 0.7
- Tier 3 (creative/exploratory): 0.4
Step 4 — Selection
Pick the top 3–5 models by score.
Step 5 — Application
For each model:
- name it
- explain why it applies
- apply it to the query
- extract actionable insight
Step 6 — Synthesis
Combine insights into a single recommendation.
Step 7 — Confidence
Use Gradient of Certainty to express confidence.