Iconsult's 7-step guided architecture review — applied to OpenAI's Financial Research Agent
Fetched manager.py, agents/*.py. Identified orchestrator pattern, .as_tool() delegation, silent except Exception: return None, and terminal verifier.
Embedded project description (text-embedding-3-small) and ranked all 141 concepts by cosine similarity against pre-computed KG embeddings. Same input → same ranking — no LLM.
Assessed complexity as complex (86/100). Controls depth: simple (3 concepts, 1 hop) → moderate (5, 2 hops, optional critique) → complex (8, parallel subagents, 2nd traversal, mandatory critique).
4 parallel subagents, 2 traversal rounds (39 nodes, 45 edges). Logged 20 pattern assessments (7 implemented, 3 partial, 7 missing, 3 N/A).
Scoped to discovered concepts — returned exact chapter numbers, page ranges, and quotes. Follow-up questions generated deterministically from graph edge templates.
4 coverage dimensions (concept, relationship type, passage diversity, critical edges). 7-category maturity scorecard. 5 failure walkthroughs for missing patterns.
No LLM — 7 rule-based checks against fixed thresholds (workflow completeness, traversal ≥ 3, assessments ≥ 5, coverage ≥ 50%, critical edges). Flagged 2 issues; backfilled 6 concepts.
Server-side HTML — scores, scenarios, and coverage from DB merged with Claude-provided narrative (~1700 tokens). Full CSS/JS/zoom/tooltips baked in.
Phased checklist classifying each step as mechanical (concrete code change) or design decision (architectural choice). Persisted in DB for cross-session tracking.