@relayplane/proxy 1.5.22 → 1.5.24

package/README.md

@@ -48,6 +48,31 @@ A minimal config file:
 
 All configuration is optional — sensible defaults are applied for every field. The proxy merges your config with its defaults via deep merge, so you only need to specify what you want to change.
 
+## Architecture (Current)
+
+```text
+Client (Claude Code / Aider / Cursor)
+        |
+        |  OpenAI/Anthropic-compatible request
+        v
++-----------------------------------------------+
+| RelayPlane Proxy (local)                       |
+|-----------------------------------------------|
+| 1) Parse request                               |
+| 2) Infer task/complexity (pre-request)         |
+| 3) Select route/model                          |
+|    - explicit model / passthrough             |
+|    - relayplane:auto/cost/fast/quality        |
+|    - configured complexity/cascade rules       |
+| 4) Forward request to provider                 |
+| 5) Return provider response                    |
+| 6) (Optional) record telemetry metadata        |
++-----------------------------------------------+
+        |
+        v
+Provider APIs (Anthropic/OpenAI/Gemini/xAI/Moonshot/...)
+```
+
 ## How It Works
 
 RelayPlane is a local HTTP proxy. You point your agent at `localhost:4801` by setting `ANTHROPIC_BASE_URL` or `OPENAI_BASE_URL`. The proxy:

package/dist/standalone-proxy.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"standalone-proxy.d.ts","sourceRoot":"","sources":["../src/standalone-proxy.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AAEH,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAKlC,OAAO,KAAK,EAAE,QAAQ,EAAY,MAAM,kBAAkB,CAAC;AAG3D,OAAO,EAAE,cAAc,EAAE,MAAM,YAAY,CAAC;AAU5C,2DAA2D;AAC3D,eAAO,MAAM,mBAAmB,gBAAuB,CAAC;AAExD;;GAEG;AACH,MAAM,WAAW,gBAAgB;IAC/B,OAAO,EAAE,MAAM,CAAC;IAChB,SAAS,EAAE,MAAM,CAAC;CACnB;AAED;;GAEG;AACH,eAAO,MAAM,iBAAiB,EAAE,MAAM,CAAC,MAAM,EAAE,gBAAgB,CA6C9D,CAAC;AAEF;;GAEG;AACH,eAAO,MAAM,aAAa,EAAE,MAAM,CAAC,MAAM,EAAE;IAAE,QAAQ,EAAE,QAAQ,CAAC;IAAC,KAAK,EAAE,MAAM,CAAA;CAAE,CAa/E,CAAC;AAEF;;;GAGG;AACH,eAAO,MAAM,kBAAkB,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAGrD,CAAC;AAEF;;;GAGG;AACH,eAAO,MAAM,aAAa,EAAE,MAAM,CAAC,MAAM,EAAE;IAAE,QAAQ,EAAE,QAAQ,CAAC;IAAC,KAAK,EAAE,MAAM,CAAA;CAAE,CAQ/E,CAAC;AAiCF;;GAEG;AACH,wBAAgB,sBAAsB,IAAI,MAAM,EAAE,CAWjD;AAED;;;GAGG;AACH,wBAAgB,iBAAiB,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,CAMvD;AAkBD,KAAK,aAAa,GAAG,MAAM,GAAG,MAAM,GAAG,SAAS,CAAC;AAEjD,UAAU,WAAW;IACnB,SAAS,EAAE,MAAM,CAAC;IAClB,MAAM,EAAE,aAAa,GAAG,IAAI,CAAC;CAC9B;AAcD,UAAU,aAAa;IACrB,OAAO,EAAE,OAAO,CAAC;IACjB,MAAM,EAAE,MAAM,EAAE,CAAC;IACjB,UAAU,EAAE,aAAa,GAAG,SAAS,GAAG,OAAO,CAAC;IAChD,cAAc,EAAE,MAAM,CAAC;CACxB;AAmBD,KAAK,UAAU,GAAG,QAAQ,GAAG,UAAU,GAAG,SAAS,CAAC;AA6EpD;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,OAAO,CAAC;IAClB;;;;;OAKG;IACH,aAAa,CAAC,EAAE,aAAa,GAAG,KAAK,GAAG,MAAM,CAAC;CAChD;AAmZD,wBAAgB,gBAAgB,CAAC,KAAK,EAAE,MAAM,GAAG,WAAW,CAe3D;AAuDD,wBAAgB,kBAAkB,CAAC,QAAQ,EAAE,KAAK,CAAC;IAAE,OAAO,CAAC,EAAE,OAAO,CAAA;CAAE,CAAC,GAAG,UAAU,CAiBrF;AAED,wBAAgB,cAAc,CAAC,YAAY,EAAE,MAAM,EAAE,OAAO,EAAE,aAAa,CAAC,YAAY,CAAC,GAAG,OAAO,CAIlG;AAq9CD;;GAEG;AACH,wBAAsB,UAAU,CAAC,MAAM,GAAE,WAAgB,GAAG,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,CAkrC/E"}
+{"version":3,"file":"standalone-proxy.d.ts","sourceRoot":"","sources":["../src/standalone-proxy.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AAEH,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAKlC,OAAO,KAAK,EAAE,QAAQ,EAAY,MAAM,kBAAkB,CAAC;AAG3D,OAAO,EAAE,cAAc,EAAE,MAAM,YAAY,CAAC;AAU5C,2DAA2D;AAC3D,eAAO,MAAM,mBAAmB,gBAAuB,CAAC;AAExD;;GAEG;AACH,MAAM,WAAW,gBAAgB;IAC/B,OAAO,EAAE,MAAM,CAAC;IAChB,SAAS,EAAE,MAAM,CAAC;CACnB;AAED;;GAEG;AACH,eAAO,MAAM,iBAAiB,EAAE,MAAM,CAAC,MAAM,EAAE,gBAAgB,CA6C9D,CAAC;AAEF;;GAEG;AACH,eAAO,MAAM,aAAa,EAAE,MAAM,CAAC,MAAM,EAAE;IAAE,QAAQ,EAAE,QAAQ,CAAC;IAAC,KAAK,EAAE,MAAM,CAAA;CAAE,CAa/E,CAAC;AAEF;;;GAGG;AACH,eAAO,MAAM,kBAAkB,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAGrD,CAAC;AAEF;;;GAGG;AACH,eAAO,MAAM,aAAa,EAAE,MAAM,CAAC,MAAM,EAAE;IAAE,QAAQ,EAAE,QAAQ,CAAC;IAAC,KAAK,EAAE,MAAM,CAAA;CAAE,CAQ/E,CAAC;AAiCF;;GAEG;AACH,wBAAgB,sBAAsB,IAAI,MAAM,EAAE,CAWjD;AAED;;;GAGG;AACH,wBAAgB,iBAAiB,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,CAMvD;AAkBD,KAAK,aAAa,GAAG,MAAM,GAAG,MAAM,GAAG,SAAS,CAAC;AAEjD,UAAU,WAAW;IACnB,SAAS,EAAE,MAAM,CAAC;IAClB,MAAM,EAAE,aAAa,GAAG,IAAI,CAAC;CAC9B;AAcD,UAAU,aAAa;IACrB,OAAO,EAAE,OAAO,CAAC;IACjB,MAAM,EAAE,MAAM,EAAE,CAAC;IACjB,UAAU,EAAE,aAAa,GAAG,SAAS,GAAG,OAAO,CAAC;IAChD,cAAc,EAAE,MAAM,CAAC;CACxB;AAmBD,KAAK,UAAU,GAAG,QAAQ,GAAG,UAAU,GAAG,SAAS,CAAC;AA6EpD;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,OAAO,CAAC;IAClB;;;;;OAKG;IACH,aAAa,CAAC,EAAE,aAAa,GAAG,KAAK,GAAG,MAAM,CAAC;CAChD;AAmZD,wBAAgB,gBAAgB,CAAC,KAAK,EAAE,MAAM,GAAG,WAAW,CAe3D;AAuDD,wBAAgB,kBAAkB,CAAC,QAAQ,EAAE,KAAK,CAAC;IAAE,OAAO,CAAC,EAAE,OAAO,CAAA;CAAE,CAAC,GAAG,UAAU,CAiBrF;AAED,wBAAgB,cAAc,CAAC,YAAY,EAAE,MAAM,EAAE,OAAO,EAAE,aAAa,CAAC,YAAY,CAAC,GAAG,OAAO,CAIlG;AAu9CD;;GAEG;AACH,wBAAsB,UAAU,CAAC,MAAM,GAAE,WAAgB,GAAG,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,CA+sC/E"}

package/dist/standalone-proxy.js

@@ -1823,7 +1823,7 @@ table{width:100%;border-collapse:collapse;font-size:.85rem}
 th{text-align:left;color:#64748b;font-weight:500;padding:8px 12px;border-bottom:1px solid #1e293b;font-size:.75rem;text-transform:uppercase;letter-spacing:.04em}
 td{padding:8px 12px;border-bottom:1px solid #111318}
 .section{margin-bottom:32px}.section h2{font-size:1rem;font-weight:600;margin-bottom:12px;color:#94a3b8}
-.dot{display:inline-block;width:8px;height:8px;border-radius:50%;margin-right:6px}.dot.up{background:#34d399}.dot.down{background:#ef4444}
+.dot{display:inline-block;width:8px;height:8px;border-radius:50%;margin-right:6px}.dot.up{background:#34d399}.dot.warn{background:#fbbf24}.dot.down{background:#ef4444}
 .badge{display:inline-block;padding:2px 8px;border-radius:6px;font-size:.75rem;font-weight:500}
 .badge.ok{background:#052e1633;color:#34d399}.badge.err{background:#2d0a0a;color:#ef4444}
 .badge.tt-code{background:#1e3a5f;color:#60a5fa}.badge.tt-analysis{background:#3b1f6e;color:#a78bfa}.badge.tt-summarization{background:#1a3a2a;color:#6ee7b7}.badge.tt-qa{background:#3a2f1e;color:#fbbf24}.badge.tt-general{background:#1e293b;color:#94a3b8}
@@ -1872,9 +1872,11 @@ async function load(){
     $('runs').innerHTML=(runsR.runs||[]).map(r=>
       '<tr><td>'+fmtTime(r.started_at)+'</td><td>'+r.model+'</td><td class="col-tt"><span class="badge '+ttCls(r.taskType)+'">'+(r.taskType||'general').replace(/_/g,' ')+'</span></td><td class="col-cx"><span class="badge '+cxCls(r.complexity)+'">'+(r.complexity||'simple')+'</span></td><td>'+(r.tokensIn||0)+'</td><td>'+(r.tokensOut||0)+'</td><td>$'+fmt(r.costUsd,4)+'</td><td>'+r.latencyMs+'ms</td><td><span class="badge '+(r.status==='success'?'ok':'err')+'">'+r.status+'</span></td></tr>'
     ).join('')||'<tr><td colspan=9 style="color:#64748b">No runs yet</td></tr>';
-    $('providers').innerHTML=(provH.providers||[]).map(p=>
-      '<div class="prov-item"><span class="dot '+(p.status==='healthy'?'up':'down')+'"></span>'+p.provider+'</div>'
-    ).join('');
+    $('providers').innerHTML=(provH.providers||[]).map(p=>{
+      const dotClass = p.status==='healthy'?'up':(p.status==='degraded'?'warn':'down');
+      const rate = p.successRate!==undefined?(' '+Math.round(p.successRate*100)+'%'):'';
+      return '<div class="prov-item"><span class="dot '+dotClass+'"></span>'+p.provider+rate+'</div>';
+    }).join('');
   }catch(e){console.error(e)}
 }
 load();setInterval(load,5000);
@@ -2088,7 +2090,7 @@ async function startProxy(config = {}) {
                 const offset = parseInt(params.get('offset') || '0', 10);
                 const sorted = [...requestHistory].reverse();
                 const runs = sorted.slice(offset, offset + limit).map(r => {
-                    const origCost = (0, telemetry_js_1.estimateCost)(r.originalModel, r.tokensIn, r.tokensOut);
+                    const origCost = (0, telemetry_js_1.estimateCost)('claude-opus-4-20250514', r.tokensIn, r.tokensOut);
                     const perRunSavings = Math.max(0, origCost - r.costUsd);
                     return {
                         id: r.id,
@@ -2117,15 +2119,15 @@ async function startProxy(config = {}) {
                 return;
             }
             if (req.method === 'GET' && telemetryPath === 'savings') {
-                // Savings = cost(original requested model) - cost(actual routed-to model)
-                // If no routing happened (same model) → savings = 0
-                // Uses actual token counts from the response, so zeroed tokens produce zero savings.
+                // Savings = cost if everything ran on Opus - actual cost
+                // Always compare against Opus as the baseline
+                const OPUS_BASELINE = 'claude-opus-4-20250514';
                 let totalOriginalCost = 0;
                 let totalActualCost = 0;
                 let totalSavedAmount = 0;
                 const byDayMap = new Map();
                 for (const r of requestHistory) {
-                    const origCost = (0, telemetry_js_1.estimateCost)(r.originalModel, r.tokensIn, r.tokensOut);
+                    const origCost = (0, telemetry_js_1.estimateCost)(OPUS_BASELINE, r.tokensIn, r.tokensOut);
                     const actualCost = r.costUsd;
                     const saved = Math.max(0, origCost - actualCost);
                     totalOriginalCost += origCost;
@@ -2161,14 +2163,39 @@ async function startProxy(config = {}) {
                 return;
             }
             if (req.method === 'GET' && telemetryPath === 'health') {
+                // Calculate per-provider success rates from recent history (last 50 requests per provider)
+                const providerStats = {};
+                const recentHistory = requestHistory.slice(-200); // Look at last 200 requests
+                for (const r of recentHistory) {
+                    const provider = r.provider || 'unknown';
+                    if (!providerStats[provider]) {
+                        providerStats[provider] = { success: 0, total: 0 };
+                    }
+                    providerStats[provider].total++;
+                    if (r.success) {
+                        providerStats[provider].success++;
+                    }
+                }
+                // Debug: log provider stats
+                console.log('[RelayPlane Health] Provider stats:', JSON.stringify(providerStats));
                 const providers = [];
                 for (const [name, ep] of Object.entries(exports.DEFAULT_ENDPOINTS)) {
                     const hasKey = !!process.env[ep.apiKeyEnv];
+                    const stats = providerStats[name.toLowerCase()];
+                    const successRate = stats && stats.total > 0 ? stats.success / stats.total : (hasKey ? 1 : 0);
+                    // Mark as unhealthy if success rate < 80% and has had requests
+                    let status = 'healthy';
+                    if (!hasKey) {
+                        status = 'down';
+                    }
+                    else if (stats && stats.total >= 5 && successRate < 0.8) {
+                        status = 'degraded';
+                    }
                     providers.push({
                         provider: name,
-                        status: hasKey ? 'healthy' : 'down',
+                        status,
                         latency: 0,
-                        successRate: hasKey ? 1 : 0,
+                        successRate: Math.round(successRate * 100) / 100,
                         lastChecked: new Date().toISOString(),
                     });
                 }