Predictika · Cloud and edge server quotes

An AI sales agent's server quote, checked before it goes out

An AI sales agent prepares a quote for cloud and edge servers, with the model, the number of chassis and nodes, the CPUs, memory modules and their size, persistent memory, storage drives and extension modules, PCIe and network cards, the memory each node needs and the total memory the quote states. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the server rules and returns a verdict, every reason and the total memory it computed.

Scope: Cloud and edge server quotes: SV7100G4, SV5100G4, SV310G4, ES200G2, SV7220G3 and EP100, with nodes per chassis, CPUs, memory and persistent memory, storage drives and extension modules, PCIe and network cards, and the total memory the quote states. See everything this validator covers

This validator models Wiwynn cloud and edge server quotes, not just the scenarios shown below.

  • Nodes and sockets — the nodes each chassis holds, the CPUs each node supports and the memory slots, with at least one memory module for each CPU.
  • Memory and storage — the largest memory module, persistent memory on the SV310G4, the drives the base model and the extension modules support, and the memory each node needs against what it is quoted.
  • Cards and totals — PCIe and network cards for each model, the total memory stated against the total computed, and a person's review for persistent memory and orders above 96 nodes.

Any server quote that touches these rules is in scope — not only the scenarios pictured here. Not covered: rack-scale AI systems, power supplies, firmware, services and prices, because Wiwynn sells by contract and this validator models no prices. The model names and the node, slot, drive and card counts that Wiwynn publishes follow its public pages; the other limits, the 128 GB module maximum, the extension module capacity, the review thresholds and all rules are modeled from comparable servers and are not published by Wiwynn.

The AI sales agentBuilds the server quote from the conversation and says it is ready.
Predictika validatorChecks every rule at once and returns the verdict, each finding, and the computed figures.

6 scenarios in 3 groups · each group runs from the agent’s first attempt to the outcome · click one to switch

What the agent proposes

The AI sales agent says

Illustrative agent statement. The validator receives the structured request below.

Proposing agent: AI sales agent (scripted)

Before sending, the agent must call Recorded run

Predictika validator


      

        
        
        
      

      
Waiting for the call.

      
    

  

  
  

    
Where rules-based CPQ gets it wrong
Click to expand
    
Our assessment of what a typical rules-based CPQ (product rules, price rules and validation rules, as in Salesforce CPQ) does with the same cases. It is a judgment about effort and reliability, not a claim that nothing can be built.

    
What the validator caughtTypical rules-based CPQPredictika
Limits that differ for every modelOnly with care
Nodes, sockets, slots, module size, drives and cards each have a different limit on each of six models, so every rule needs a lookup by model.		The model and every quantity are fields of one request and are checked together on every call.
Persistent memory beside ordinary memoryOnly with workarounds
Persistent memory is allowed on one model only and may not outnumber ordinary memory, and both share the slots, so three rules must read the same counts.		Both counts are compared with the model on every call, whatever order they were chosen in.
Storage that grows with extension modulesOnly with care
The drive limit is the base count plus a number for each extension module, and the module count has its own limit.		The modules and drives are checked together, and each problem is its own finding.
A stated total that must match nodes and memoryOnly with workarounds
The total multiplies chassis, nodes and the memory in each node, so a separate rule must recompute it and compare it with the stated figure.		The stated total is a field of the request and is compared with the computed total on every call.
Referrals for persistent memory and large ordersOnly with workarounds
A review depends on the persistent memory and on the total node count, on top of the blocking rules.		The referrals are separate findings returned together with the blocking ones.

  

  

    
What this validator catches: findings from its test cases
Click to expand
    
Every rule is checked on every call, in any order the choices arrive; one problem never hides another. Blocking findings stop the quote; approval findings need a person to sign off; everything else is computed.

    
1The quote has 13 nodes per chassis; this model holds at most 12.
2The quote has 2 CPUs per node; this model has 1 socket(s) per node.
3The quote has 7 DRAM DIMMs per node; with persistent memory modules the node has only 6 DIMM slots.
4The quote has 33 DRAM DIMMs per node; with persistent memory modules the node has only 32 DIMM slots.
5The quote uses persistent memory modules; a person should confirm the customer software supports them.
6The quote has 17 DRAM DIMMs per node; with persistent memory modules the node has only 32 DIMM slots.
7The quote has 17 persistent memory modules per node; this model supports at most 16.
8Persistent memory modules (10) cannot outnumber DRAM DIMMs (8) in a node.
9The quote has 15 storage drives per node; this configuration supports 14.
10The quote has 3 storage extension modules; this model supports at most 2.
11The quote has 1 storage extension modules; this model supports at most 0.
12The quote has 64 GB DIMMs; this model takes DIMMs up to 32 GB.
13The quote has 7 PCIe cards per node; this model has 6 slots.
14The quote has 3 network cards per node; this model supports 2.
15Every node needs at least one network card.
16The customer needs 1,100 GB per node; the quoted memory gives 1,024 GB.
17The quote states 1,024 GB of memory in total; the quoted nodes give 2048 GB.
18The order has more than 96 nodes; a person should confirm volume terms with the account team.
19Every CPU needs at least one DRAM DIMM; the quote has 1 DIMMs for 2 CPUs.

  

  

    
Scenarios and results
Click to expand
    
The recorded result of the validator for each scenario on this page: the request the AI agent sent, the verdict, and every finding, in the order they appear above.

    
1. Eight SV7100G4 chassis: Eight SV7100G4 chassis with 13 nodes each, two CPUs per node and extra cards

    
The agent's request. Please quote 8 SV7100G4 chassis with 13 nodes each, 2 CPUs per node, 7 memory modules of 64 GB per node, 1 storage drive, 3 PCIe cards and 2 network cards per node. Each node needs 512 GB of memory. The total memory stated on the quote is 50,000 GB. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Too many nodes. The quote has 13 nodes per chassis; this model holds at most 12.
    • 
      
  • Too many CPUs. The quote has 2 CPUs per node; this model has 1 socket(s) per node.
    • 
      
  • Too many DIMMs. The quote has 7 DRAM DIMMs per node; with persistent memory modules the node has only 6 DIMM slots.
    • 
      
  • Memory needed exceeds configuration. The customer needs 512 GB per node; the quoted memory gives 448 GB.
    • 
      
  • Too many PCIe cards. The quote has 3 PCIe cards per node; this model has 2 slots.
    • 
      
  • Too many network cards. The quote has 2 network cards per node; this model supports 1.
    • 
      
  • Stated memory does not match. The quote states 50,000 GB of memory in total; the quoted nodes give 46,592 GB.
    • 
      
  • Large volume order. The order has more than 96 nodes; a person should confirm volume terms with the account team.
    • 
    

    
2. Eight SV7100G4 chassis: The agent resubmits after fixing the errors

    
The agent's request. Please quote 8 SV7100G4 chassis with 12 nodes each, 1 CPU per node, 6 memory modules of 128 GB per node, 1 storage drive, 2 PCIe cards and 1 network card per node. Each node needs 512 GB of memory. The total memory stated on the quote is 73,728 GB. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. 96 nodes in total, 73,728 GB of memory in total.
    • 
    

    
3. Six SV5100G4 servers: Six SV5100G4 servers with 256 GB memory modules, three extension modules and 20 drives

    
The agent's request. Please quote 6 SV5100G4 chassis with 1 node each, 2 CPUs per node, 16 memory modules of 256 GB per node, 20 storage drives, 1 PCIe card and 1 network card per node. Each node needs 1,024 GB of memory. It also has 3 storage extension modules. The total memory stated on the quote is 12,288 GB. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • DIMM size above the model limit. The quote has 256 GB DIMMs; this model takes DIMMs up to 128 GB.
    • 
      
  • Too many storage extension modules. The quote has 3 storage extension modules; this model supports at most 2.
    • 
      
  • Too many storage drives. The quote has 20 storage drives per node; this configuration supports 18.
    • 
      
  • Stated memory does not match. The quote states 12,288 GB of memory in total; the quoted nodes give 24,576 GB.
    • 
    

    
4. Six SV5100G4 servers: The agent resubmits after fixing the errors

    
The agent's request. Please quote 6 SV5100G4 chassis with 1 node each, 2 CPUs per node, 16 memory modules of 128 GB per node, 14 storage drives, 1 PCIe card and 1 network card per node. Each node needs 1,024 GB of memory. It also has 2 storage extension modules. The total memory stated on the quote is 12,288 GB. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. 6 nodes in total, 12,288 GB of memory in total.
    • 
    

    
5. Four SV310G4 servers with persistent memory: Four SV310G4 servers with 17 persistent memory modules and three network cards

    
The agent's request. Please quote 4 SV310G4 chassis with 1 node each, 2 CPUs per node, 16 memory modules of 64 GB per node, 2 storage drives, 1 PCIe card and 3 network cards per node. Each node needs 1,500 GB of memory. It also has 17 persistent memory modules. The total memory stated on the quote is 12,288 GB. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Too many DIMMs. The quote has 16 DRAM DIMMs per node; with persistent memory modules the node has only 32 DIMM slots.
    • 
      
  • Too many persistent memory modules. The quote has 17 persistent memory modules per node; this model supports at most 16.
    • 
      
  • Persistent memory without enough DRAM. Persistent memory modules (17) cannot outnumber DRAM DIMMs (16) in a node.
    • 
      
  • Too many network cards. The quote has 3 network cards per node; this model supports 2.
    • 
      
  • Stated memory does not match. The quote states 12,288 GB of memory in total; the quoted nodes give 12,800 GB.
    • 
      
  • Persistent memory to confirm. The quote uses persistent memory modules; a person should confirm the customer software supports them.
    • 
    

    
6. Four SV310G4 servers with persistent memory: The agent resubmits after fixing the errors

    
The agent's request. Please quote 4 SV310G4 chassis with 1 node each, 2 CPUs per node, 16 memory modules of 64 GB per node, 2 storage drives, 1 PCIe card and 2 network cards per node. Each node needs 1,500 GB of memory. It also has 16 persistent memory modules. The total memory stated on the quote is 12,288 GB. It is ready to send.

    
Verdict: Needs approval.

    
    
      
  • Persistent memory to confirm. The quote uses persistent memory modules; a person should confirm the customer software supports them.
    • 
      
  • Order summary. 4 nodes in total, 12,288 GB of memory in total.
    • 
    

  

  

    
Narration transcript
Click to expand
    
The text of the narrated walkthrough, step by step.

    
    
      
  1. Welcome to a demo of Predictika's Validator. An AI sales agent prepares a quote for cloud and edge servers, with the model, the number of chassis and nodes. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the server rules and returns a verdict, every reason and the total memory it computed.
    2. 
      
  2. I will start with the first request by the agent.
    3. 
      
  3. The agent says. Please quote 8 SV7100G4 chassis with 13 nodes each, 2 CPUs per node, 7 memory modules of 64 GB per node. Each node needs 512 GB of memory. The rest of the request is on screen. The total memory stated on the quote is 50,000 GB.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with seven findings and one approval, all in one answer. The quote has 13 nodes per chassis; this model holds at most 12. The quote has 2 CPUs per node; this model has 1 socket(s) per node. The quote has 7 DRAM DIMMs per node; with persistent memory modules the node has only 6 DIMM slots. The other four are on screen.
    6. 
      
  6. The validator returned seven blocking findings and one approval in one answer. A check that reads one field at a time would pass most of this; it shows only when the choices in the quote are checked against each other.
    7. 
      
  7. What this shows: for a stated total that must match nodes and memory, a rules-based CPQ can do this only with workarounds.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the nodes per chassis to 12, the CPUs per node to 1, the memory modules per node to 6, the memory module size to 128 GB and the total memory stated on the quote to 73,728 GB. Two more changes are on screen.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. All eight earlier findings are cleared.
    13. 
      
  13. Eight problems caught in one call, fixed in one pass, and a clean quote. That is the value of Predictika's validator: catching mistakes before they reach the customer. Now you can try the other scenarios yourself.
    14.