@relayplane/proxy 0.1.6 → 0.1.8

package/README.md

@@ -176,13 +176,85 @@ Options:
   -h, --help         Show help
 ```
 
+## REST API
+
+The proxy exposes endpoints for stats and monitoring:
+
+### `GET /health`
+
+Server health and version info.
+
+```bash
+curl http://localhost:3001/health
+```
+
+```json
+{
+  "status": "ok",
+  "version": "0.1.7",
+  "uptime": "2h 15m 30s",
+  "providers": { "anthropic": true, "openai": true, "google": false },
+  "totalRuns": 142
+}
+```
+
+### `GET /stats`
+
+Aggregated statistics and cost savings.
+
+```bash
+curl http://localhost:3001/stats
+```
+
+```json
+{
+  "totalRuns": 142,
+  "savings": {
+    "estimatedSavingsPercent": "73.2%",
+    "actualCostUsd": "0.0234",
+    "baselineCostUsd": "0.0873",
+    "savedUsd": "0.0639"
+  },
+  "modelDistribution": {
+    "anthropic/claude-3-5-haiku-latest": { "count": 98, "percentage": "69.0%" },
+    "anthropic/claude-sonnet-4-20250514": { "count": 44, "percentage": "31.0%" }
+  }
+}
+```
+
+### `GET /runs`
+
+Recent routing decisions.
+
+```bash
+curl "http://localhost:3001/runs?limit=10"
+```
+
+```json
+{
+  "runs": [
+    {
+      "runId": "abc123",
+      "timestamp": "2026-02-03T13:26:03Z",
+      "model": "anthropic/claude-3-5-haiku-latest",
+      "taskType": "code_generation",
+      "confidence": 0.92,
+      "mode": "auto",
+      "durationMs": 1203,
+      "promptPreview": "Write a function that..."
+    }
+  ],
+  "total": 142
+}
+```
+
 ## Data Storage
 
 All data stored locally at `~/.relayplane/data.db` (SQLite).
 
 ```bash
 # View recent runs
-sqlite3 ~/.relayplane/data.db "SELECT * FROM runs ORDER BY timestamp DESC LIMIT 10"
+sqlite3 ~/.relayplane/data.db "SELECT * FROM runs ORDER BY created_at DESC LIMIT 10"
 
 # Check routing rules
 sqlite3 ~/.relayplane/data.db "SELECT * FROM routing_rules"

package/dist/cli.js

@@ -25,6 +25,7 @@ var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__ge
 
 // src/proxy.ts
 var http = __toESM(require("http"));
+var url = __toESM(require("url"));
 
 // src/storage/store.ts
 var import_better_sqlite3 = __toESM(require("better-sqlite3"));
@@ -1590,7 +1591,107 @@ ${input.prompt}` : input.prompt;
   }
 };
 
+// src/config.ts
+var fs2 = __toESM(require("fs"));
+var path2 = __toESM(require("path"));
+var os2 = __toESM(require("os"));
+var import_zod = require("zod");
+var StrategySchema = import_zod.z.object({
+  model: import_zod.z.string(),
+  minConfidence: import_zod.z.number().min(0).max(1).optional(),
+  fallback: import_zod.z.string().optional()
+});
+var ConfigSchema = import_zod.z.object({
+  strategies: import_zod.z.record(import_zod.z.string(), StrategySchema).optional(),
+  defaults: import_zod.z.object({
+    qualityModel: import_zod.z.string().optional(),
+    costModel: import_zod.z.string().optional()
+  }).optional()
+});
+var DEFAULT_CONFIG = {
+  strategies: {
+    code_review: { model: "anthropic:claude-sonnet-4-20250514" },
+    code_generation: { model: "anthropic:claude-3-5-haiku-latest" },
+    analysis: { model: "anthropic:claude-sonnet-4-20250514" },
+    summarization: { model: "anthropic:claude-3-5-haiku-latest" },
+    creative_writing: { model: "anthropic:claude-sonnet-4-20250514" },
+    data_extraction: { model: "anthropic:claude-3-5-haiku-latest" },
+    translation: { model: "anthropic:claude-3-5-haiku-latest" },
+    question_answering: { model: "anthropic:claude-3-5-haiku-latest" },
+    general: { model: "anthropic:claude-3-5-haiku-latest" }
+  },
+  defaults: {
+    qualityModel: "claude-sonnet-4-20250514",
+    costModel: "claude-3-5-haiku-latest"
+  }
+};
+function getConfigPath() {
+  return path2.join(os2.homedir(), ".relayplane", "config.json");
+}
+function writeDefaultConfig() {
+  const configPath = getConfigPath();
+  const dir = path2.dirname(configPath);
+  if (!fs2.existsSync(dir)) {
+    fs2.mkdirSync(dir, { recursive: true });
+  }
+  if (!fs2.existsSync(configPath)) {
+    fs2.writeFileSync(
+      configPath,
+      JSON.stringify(DEFAULT_CONFIG, null, 2) + "\n",
+      "utf-8"
+    );
+    console.log(`[relayplane] Created default config at ${configPath}`);
+  }
+}
+function loadConfig() {
+  const configPath = getConfigPath();
+  writeDefaultConfig();
+  try {
+    const raw = fs2.readFileSync(configPath, "utf-8");
+    const parsed = JSON.parse(raw);
+    const validated = ConfigSchema.parse(parsed);
+    return validated;
+  } catch (err) {
+    if (err instanceof import_zod.z.ZodError) {
+      console.error(`[relayplane] Invalid config: ${err.message}`);
+    } else if (err instanceof SyntaxError) {
+      console.error(`[relayplane] Config JSON parse error: ${err.message}`);
+    } else {
+      console.error(`[relayplane] Failed to load config: ${err}`);
+    }
+    console.log("[relayplane] Using default config");
+    return DEFAULT_CONFIG;
+  }
+}
+function getStrategy(config, taskType) {
+  return config.strategies?.[taskType] ?? null;
+}
+function watchConfig(onChange) {
+  const configPath = getConfigPath();
+  const dir = path2.dirname(configPath);
+  if (!fs2.existsSync(dir)) {
+    fs2.mkdirSync(dir, { recursive: true });
+  }
+  let debounceTimer = null;
+  fs2.watch(dir, (eventType, filename) => {
+    if (filename === "config.json") {
+      if (debounceTimer) clearTimeout(debounceTimer);
+      debounceTimer = setTimeout(() => {
+        console.log("[relayplane] Config file changed, reloading...");
+        const newConfig = loadConfig();
+        onChange(newConfig);
+      }, 100);
+    }
+  });
+}
+
 // src/proxy.ts
+var VERSION = "0.1.8";
+var recentRuns = [];
+var MAX_RECENT_RUNS = 100;
+var modelCounts = {};
+var serverStartTime = 0;
+var currentConfig = loadConfig();
 var DEFAULT_ENDPOINTS = {
   anthropic: {
     baseUrl: "https://api.anthropic.com/v1",
@@ -1887,9 +1988,9 @@ function convertMessagesToGemini(messages) {
           return { text: p.text };
         }
         if (p.type === "image_url" && p.image_url?.url) {
-          const url = p.image_url.url;
-          if (url.startsWith("data:")) {
-            const match = url.match(/^data:([^;]+);base64,(.+)$/);
+          const url2 = p.image_url.url;
+          if (url2.startsWith("data:")) {
+            const match = url2.match(/^data:([^;]+);base64,(.+)$/);
             if (match) {
               return {
                 inline_data: {
@@ -1899,7 +2000,7 @@ function convertMessagesToGemini(messages) {
               };
             }
           }
-          return { text: `[Image: ${url}]` };
+          return { text: `[Image: ${url2}]` };
         }
         return { text: "" };
       });
@@ -2313,28 +2414,88 @@ async function startProxy(config = {}) {
   };
   const server = http.createServer(async (req, res) => {
     res.setHeader("Access-Control-Allow-Origin", "*");
-    res.setHeader("Access-Control-Allow-Methods", "POST, OPTIONS");
+    res.setHeader("Access-Control-Allow-Methods", "GET, POST, OPTIONS");
     res.setHeader("Access-Control-Allow-Headers", "Content-Type, Authorization");
     if (req.method === "OPTIONS") {
       res.writeHead(204);
       res.end();
       return;
     }
-    if (req.method !== "POST" || !req.url?.includes("/chat/completions")) {
-      if (req.method === "GET" && req.url?.includes("/models")) {
-        res.writeHead(200, { "Content-Type": "application/json" });
-        res.end(
-          JSON.stringify({
-            object: "list",
-            data: [
-              { id: "relayplane:auto", object: "model", owned_by: "relayplane" },
-              { id: "relayplane:cost", object: "model", owned_by: "relayplane" },
-              { id: "relayplane:quality", object: "model", owned_by: "relayplane" }
-            ]
-          })
-        );
-        return;
+    const parsedUrl = url.parse(req.url || "", true);
+    const pathname = parsedUrl.pathname || "";
+    if (req.method === "GET" && pathname === "/health") {
+      const uptimeMs = Date.now() - serverStartTime;
+      const uptimeSecs = Math.floor(uptimeMs / 1e3);
+      const hours = Math.floor(uptimeSecs / 3600);
+      const mins = Math.floor(uptimeSecs % 3600 / 60);
+      const secs = uptimeSecs % 60;
+      const providers = {};
+      for (const [name, config2] of Object.entries(DEFAULT_ENDPOINTS)) {
+        providers[name] = !!process.env[config2.apiKeyEnv];
       }
+      res.writeHead(200, { "Content-Type": "application/json" });
+      res.end(JSON.stringify({
+        status: "ok",
+        version: VERSION,
+        uptime: `${hours}h ${mins}m ${secs}s`,
+        uptimeMs,
+        providers,
+        totalRuns: recentRuns.length > 0 ? Object.values(modelCounts).reduce((a, b) => a + b, 0) : 0
+      }));
+      return;
+    }
+    if (req.method === "GET" && pathname === "/stats") {
+      const stats = relay.stats();
+      const savings = relay.savingsReport(30);
+      const totalRuns = Object.values(modelCounts).reduce((a, b) => a + b, 0);
+      const modelDistribution = {};
+      for (const [model, count] of Object.entries(modelCounts)) {
+        modelDistribution[model] = {
+          count,
+          percentage: totalRuns > 0 ? (count / totalRuns * 100).toFixed(1) + "%" : "0%"
+        };
+      }
+      res.writeHead(200, { "Content-Type": "application/json" });
+      res.end(JSON.stringify({
+        totalRuns,
+        savings: {
+          estimatedSavingsPercent: savings.savingsPercent.toFixed(1) + "%",
+          actualCostUsd: savings.actualCost.toFixed(4),
+          baselineCostUsd: savings.baselineCost.toFixed(4),
+          savedUsd: savings.savings.toFixed(4)
+        },
+        modelDistribution,
+        byTaskType: stats.byTaskType,
+        period: stats.period
+      }));
+      return;
+    }
+    if (req.method === "GET" && pathname === "/runs") {
+      const limitParam = parsedUrl.query["limit"];
+      const parsedLimit = limitParam ? parseInt(String(limitParam), 10) : 20;
+      const limit = Math.min(Number.isNaN(parsedLimit) ? 20 : parsedLimit, MAX_RECENT_RUNS);
+      res.writeHead(200, { "Content-Type": "application/json" });
+      res.end(JSON.stringify({
+        runs: recentRuns.slice(0, limit),
+        total: recentRuns.length
+      }));
+      return;
+    }
+    if (req.method === "GET" && pathname.includes("/models")) {
+      res.writeHead(200, { "Content-Type": "application/json" });
+      res.end(
+        JSON.stringify({
+          object: "list",
+          data: [
+            { id: "relayplane:auto", object: "model", owned_by: "relayplane" },
+            { id: "relayplane:cost", object: "model", owned_by: "relayplane" },
+            { id: "relayplane:quality", object: "model", owned_by: "relayplane" }
+          ]
+        })
+      );
+      return;
+    }
+    if (req.method !== "POST" || !pathname.includes("/chat/completions")) {
       res.writeHead(404, { "Content-Type": "application/json" });
       res.end(JSON.stringify({ error: "Not found" }));
       return;
@@ -2381,33 +2542,44 @@ async function startProxy(config = {}) {
     const confidence = getInferenceConfidence(promptText, taskType);
     log(`Inferred task: ${taskType} (confidence: ${confidence.toFixed(2)})`);
     if (routingMode !== "passthrough") {
-      const rule = relay.routing.get(taskType);
-      if (rule && rule.preferredModel) {
-        const parsed = parsePreferredModel(rule.preferredModel);
+      const configStrategy = getStrategy(currentConfig, taskType);
+      if (configStrategy) {
+        const parsed = parsePreferredModel(configStrategy.model);
         if (parsed) {
           targetProvider = parsed.provider;
           targetModel = parsed.model;
-          log(`Using learned rule: ${rule.preferredModel}`);
+          log(`Using config strategy: ${configStrategy.model}`);
+        }
+      }
+      if (!configStrategy) {
+        const rule = relay.routing.get(taskType);
+        if (rule && rule.preferredModel) {
+          const parsed = parsePreferredModel(rule.preferredModel);
+          if (parsed) {
+            targetProvider = parsed.provider;
+            targetModel = parsed.model;
+            log(`Using learned rule: ${rule.preferredModel}`);
+          } else {
+            const defaultRoute = DEFAULT_ROUTING[taskType];
+            targetProvider = defaultRoute.provider;
+            targetModel = defaultRoute.model;
+          }
         } else {
           const defaultRoute = DEFAULT_ROUTING[taskType];
           targetProvider = defaultRoute.provider;
           targetModel = defaultRoute.model;
         }
-      } else {
-        const defaultRoute = DEFAULT_ROUTING[taskType];
-        targetProvider = defaultRoute.provider;
-        targetModel = defaultRoute.model;
       }
       if (routingMode === "cost") {
-        const simpleTasks = ["summarization", "data_extraction", "translation", "question_answering"];
-        if (simpleTasks.includes(taskType)) {
-          targetModel = "claude-3-5-haiku-latest";
-          targetProvider = "anthropic";
-        }
+        const costModel = currentConfig.defaults?.costModel || "claude-3-5-haiku-latest";
+        targetModel = costModel;
+        targetProvider = "anthropic";
+        log(`Cost mode: using ${costModel}`);
       } else if (routingMode === "quality") {
-        const qualityModel = process.env["RELAYPLANE_QUALITY_MODEL"] || "claude-sonnet-4-20250514";
+        const qualityModel = currentConfig.defaults?.qualityModel || process.env["RELAYPLANE_QUALITY_MODEL"] || "claude-sonnet-4-20250514";
         targetModel = qualityModel;
         targetProvider = "anthropic";
+        log(`Quality mode: using ${qualityModel}`);
       }
     }
     log(`Routing to: ${targetProvider}/${targetModel}`);
@@ -2454,12 +2626,19 @@ async function startProxy(config = {}) {
       );
     }
   });
+  watchConfig((newConfig) => {
+    currentConfig = newConfig;
+    console.log("[relayplane] Config reloaded");
+  });
   return new Promise((resolve, reject) => {
     server.on("error", reject);
     server.listen(port, host, () => {
+      serverStartTime = Date.now();
       console.log(`RelayPlane proxy listening on http://${host}:${port}`);
       console.log(`  Models: relayplane:auto, relayplane:cost, relayplane:quality`);
       console.log(`  Endpoint: POST /v1/chat/completions`);
+      console.log(`  Stats: GET /stats, /runs, /health`);
+      console.log(`  Config: ~/.relayplane/config.json (hot-reload enabled)`);
       console.log(`  Streaming: \u2705 Enabled`);
       resolve(server);
     });
@@ -2522,11 +2701,26 @@ async function handleStreamingRequest(res, request, targetProvider, targetModel,
     log(`Streaming error: ${err}`);
   }
   const durationMs = Date.now() - startTime;
+  const modelKey = `${targetProvider}/${targetModel}`;
+  modelCounts[modelKey] = (modelCounts[modelKey] || 0) + 1;
   relay.run({
     prompt: promptText.slice(0, 500),
     taskType,
     model: `${targetProvider}:${targetModel}`
   }).then((runResult) => {
+    recentRuns.unshift({
+      runId: runResult.runId,
+      timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+      model: modelKey,
+      taskType,
+      confidence,
+      mode: routingMode,
+      durationMs,
+      promptPreview: promptText.slice(0, 100) + (promptText.length > 100 ? "..." : "")
+    });
+    if (recentRuns.length > MAX_RECENT_RUNS) {
+      recentRuns.pop();
+    }
     log(`Completed streaming in ${durationMs}ms, runId: ${runResult.runId}`);
   }).catch((err) => {
     log(`Failed to record run: ${err}`);
@@ -2597,15 +2791,30 @@ async function handleNonStreamingRequest(res, request, targetProvider, targetMod
     return;
   }
   const durationMs = Date.now() - startTime;
+  const modelKey = `${targetProvider}/${targetModel}`;
+  modelCounts[modelKey] = (modelCounts[modelKey] || 0) + 1;
   try {
     const runResult = await relay.run({
       prompt: promptText.slice(0, 500),
       taskType,
       model: `${targetProvider}:${targetModel}`
     });
+    recentRuns.unshift({
+      runId: runResult.runId,
+      timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+      model: modelKey,
+      taskType,
+      confidence,
+      mode: routingMode,
+      durationMs,
+      promptPreview: promptText.slice(0, 100) + (promptText.length > 100 ? "..." : "")
+    });
+    if (recentRuns.length > MAX_RECENT_RUNS) {
+      recentRuns.pop();
+    }
     responseData["_relayplane"] = {
       runId: runResult.runId,
-      routedTo: `${targetProvider}/${targetModel}`,
+      routedTo: modelKey,
       taskType,
       confidence,
       durationMs,