SUMA Memory
Graph-Based Memory with Weighted Relationships. The brain of QUAD.
What Makes SUMA Different
Every AI memory system stores text and finds “similar” text.
SUMA stores relationships with weights and finds the RIGHT context — not just similar context.
Everyone Else (Vector Search)
User Question
→ Convert to numbers (embedding)
→ Find similar numbers in database
→ Return similar text
Fails for: relationships, importance, paths, explainability
SUMA (Weighted Graph)
User Question
→ Search graph nodes
→ Follow weighted edges
→ Calculate virtual weights across hops
→ Return ranked context with full path
Handles: relationships, importance, paths, explainability
The “Who is Cookie?” Test
Mem0 / LangChain
- 1. Convert “Cookie” to embedding
- 2. Find similar embeddings
- 3. Returns: “Cookie recipe” (0.89), “Cookie Monster” (0.87), “HTTP Cookie” (0.85)
- All wrong. Cost: $0.005
SUMA
- 1. Exact match on tags → no match
- 2. Fuzzy match (Levenshtein ≤2) → no match
- 3. LLM fallback: “Cookie = Kuvi, Lokesh's daughter”
- 4. Graph: Kuvi →(0.9)→ Lokesh →(0.95)→ Suman
- 5. Virtual weight: 0.9 × 0.95 × 0.85 = 0.727
- Correct answer. Cost: $0.001
SUMA vs The Competition
| Feature | Mem0 | Zep | LangChain | SUMA | |
|---|---|---|---|---|---|
| Graph structure | Labels | Temporal | No | No | Weighted |
| Weighted relationships | No | No | No | No | Yes (0.0-1.0) |
| Virtual weight (multi-hop) | No | No | No | No | Yes (decay^d) |
| Path explainability | No | No | No | No | Yes |
| Hierarchical spheres | No | No | No | Topics | Yes (recursive) |
| LLM-free search | No | Partial | No | No | Yes (most) |
| Cloud-agnostic | Yes | Partial | Yes | No (GCP) | Yes |
| Cost per 1K queries | ~$5 | ~$3 | ~$2 | ~$8 | ~$0.10 |
The Patent Moat
Adaptive Weight-Based Context Depth
Uses numerical weights to determine how deep to search the graph. Higher weights = deeper traversal.
Harmonic Mean Path Weighting
Calculates virtual weights across multi-hop paths: virtual_weight = max(path_weight × 0.85^distance).
Sphere-Aware Traversal
Hierarchical sphere structure scopes searches. Search relevant sphere first, expand only if needed.
Total: 50 patent claims across US + India. No competitor can replicate without licensing.
SUMA is a REST API
Any application can use SUMA. Nodes, relationships, weights — all accessible via HTTP. No vendor lock-in. No embedded SDK required.
The SUMMA Algorithm
5-step search algorithm. 4 steps are free. 1 costs $0.001. Most queries never reach step 5.
Search node tags and content for exact keyword match
Levenshtein distance ≤ 2. Catches typos, nicknames, variations
BFS from matched nodes. Follow weighted edges. Apply weight decay per hop
Calculate multi-hop strength: harmonic mean × 0.85^distance
Only if steps 1-4 return nothing. Ask AI to interpret the query
95% of queries resolved without any LLM call. Zero cost for algorithmic search.