Zero-Node Static Engine
Request Pilot
Full-scale static contract extraction across production Java Spring Boot and TypeScript/React codebases. Inspect real-world endpoint resolution, wire schema unwrapping, and zero-AST verification times measured directly from native Rust.
1. Verified Production Benchmark (ThingsBoard): The 20 REST API measurements reported in Section 4 of the Whitepaper reflect native Rust instrumentation (measure_exact_tokens.rs) executed against the official ThingsBoard repository (multi-million line Java backend + Angular frontend). These represent reported internal test engineering results, not third-party audited figures.
2. Monorepo Mining Explorer (1,000 Repositories): The dataset below represents Sajimora's™ target evaluation index of publicly accessible full-stack enterprise repositories. Repository inventory filters include Java Spring Boot backends paired with TypeScript/React client consumers. Synthetic and sample projects with non-standard RPC (gRPC, GraphQL) or unparseable dynamic wrappers are excluded.
3. Test Conditions: Benchmarks executed locally on AMD Ryzen / Intel Core x86_64, NVMe SSD, single-threaded execution, measuring resident set size (RSS) memory and parser pass duration. Individual project execution times and token counts will vary based on hardware, monorepo depth, and AST nesting.
| Repository Name | Language Domain | Framework Stack | Endpoints | DTOs Reconciled | Extraction Time | Contract Status | Action |
|---|
Simulate real-time DTO schema mutations and see how Sajimora computes consumer impact without compiler rebuilds.
frontend/src/api/dashboardApi.ts:24res.data.lowStock (Required → Undefined)sajimora heal to auto-insert Jackson alias
See how replacing brute-force file context dumping with Sajimora's native MCP server eliminates 99.7% of token waste and saves AI companies millions in LLM API bills.
sajimora_endpoint_impact → B-tree SQLite point-query → Returns 65 exact tokens.
Estimate the exact dollar savings for autonomous AI coding agents and enterprise engineering teams.
{
"tool": "sajimora_what_if",
"status": "ANALYZED",
"verdict": "VIOLATED",
"mutation": {
"action": "RENAME_FIELD",
"targetDto": "DashboardDTO",
"field": "lowStock",
"renameTo": "lowStockAlerts",
"jsonPropertyRetained": false
},
"breaking_consumers": [
"frontend/src/api/dashboardApi.ts:24 (calls res.data.lowStock)"
],
"telemetry": {
"execution_latency_ms": 2.41,
"context_tokens_saved": 82400,
"api_dollars_saved_usd": 0.2472,
"prevented_failed_agent_loop": true
}
}
Derived directly from the live Rust engine test suite (measure_exact_tokens.rs). Comparing raw Grep line-dumping against Sajimora's B-Tree type closure.
| # | Target DTO & Field | Grep Tokens Ingested | Grep Cached | Sajimora Exact Tokens | Sajimora Cached | Call Sites Flagged | Reduction |
|---|---|---|---|---|---|---|---|
| 1 | RuleChain.name | 46,025 (1,097 l) | 15,000 | 464 (1,786 c) | 15,000 | 8 call sites | -98.99% |
| 2 | TbResourceInfo.customerId | 10,933 (287 l) | 61,175 | 353 (1,359 c) | 15,544 | 5 call sites | -96.77% |
| 3 | AgentApplication.agentId | 1,218 (34 l) | 72,258 | 372 (1,433 c) | 15,977 | 5 call sites | -69.46% |
| 4 | JwtPair.token | 3,114 (66 l) | 73,626 | 192 (739 c) | 16,429 | 0 call sites | -93.83% |
| 5 | Alarm.customerId | 10,933 (287 l) | 76,890 | 259 (998 c) | 16,701 | 2 call sites | -97.63% |
| 6 | Device.customerId | 10,933 (287 l) | 87,973 | 291 (1,121 c) | 17,040 | 3 call sites | -97.34% |
| 7 | OtaPackageInfo.deviceProfileId | 2,035 (53 l) | 99,056 | 251 (966 c) | 17,411 | 1 call site | -87.67% |
| 8 | AgentAppProfile.name | 46,025 (1,097 l) | 101,241 | 338 (1,301 c) | 17,742 | 4 call sites | -99.27% |
| 9 | Edge.customerId | 10,933 (287 l) | 147,416 | 366 (1,409 c) | 18,160 | 5 call sites | -96.65% |
| 10 | Integration.configuration | 24,550 (523 l) | 158,499 | 312 (1,200 c) | 18,606 | 3 call sites | -98.73% |
| 11 | Job.customerId | 10,933 (287 l) | 183,199 | 322 (1,241 c) | 18,998 | 3 call sites | -97.05% |
| 12 | Queue.name | 46,025 (1,097 l) | 194,282 | 187 (719 c) | 19,400 | 0 call sites | -99.59% |
| 13 | WhiteLabelingParams.logoImageUrl | 268 (9 l) | 240,457 | 396 (1,526 c) | 19,667 | 5 call sites | Grep Missed* |
| 14 | DeviceProfile.name | 46,025 (1,097 l) | 240,875 | 302 (1,164 c) | 20,143 | 3 call sites | -99.34% |
| 15 | EntityInfo.name | 46,025 (1,097 l) | 287,050 | 189 (728 c) | 20,525 | 0 call sites | -99.59% |
| 16 | EntityView.entityId | 13,989 (378 l) | 333,225 | 266 (1,025 c) | 20,794 | 2 call sites | -98.10% |
| 17 | HomeDashboardInfo.dashboardId | 2,382 (62 l) | 347,364 | 281 (1,082 c) | 21,140 | 2 call sites | -88.20% |
| 18 | TenantProfile.name | 46,025 (1,097 l) | 349,896 | 302 (1,161 c) | 21,501 | 3 call sites | -99.34% |
| 19 | AiModel.version | 12,436 (321 l) | 396,071 | 260 (1,001 c) | 21,883 | 2 call sites | -97.91% |
| 20 | Asset.customerId | 10,933 (287 l) | 408,657 | 259 (998 c) | 22,223 | 2 call sites | -97.63% |
| TOTAL ACROSS 20 APIS | 445,785 tokens | 3,874,210 | 5,963 tokens | 369,453 | 54 Call Sites | -98.66% | |
* Note: On WhiteLabelingParams.logoImageUrl, raw string Grep returned 268 tokens but completely missed 5 dynamic Angular template call sites due to variable binding. Sajimora accurately resolved all 5 call sites using B-Tree contract closures.