@infersec/conduit 1.41.0 → 1.42.0

package/dist/cli.js

@@ -6,7 +6,7 @@ const __dirname = __pathDirname(__filename);
 
 import { parseArgs } from 'node:util';
 import 'node:crypto';
-import { a as asError, s as startInferenceAgent } from './start-CPrgh7rN.js';
+import { a as asError, s as startInferenceAgent } from './start-DGdf6ycx.js';
 import 'argon2';
 import 'node:child_process';
 import 'node:stream';

package/dist/index.js

@@ -5,7 +5,7 @@ const __filename = __fileURLToPath(import.meta.url);
 const __dirname = __pathDirname(__filename);
 
 import 'node:crypto';
-import { s as startInferenceAgent, a as asError } from './start-CPrgh7rN.js';
+import { s as startInferenceAgent, a as asError } from './start-DGdf6ycx.js';
 import 'argon2';
 import 'node:child_process';
 import 'node:stream';

package/dist/requestHandlers/createConduitAnthropicAPIReferenceHandlers.d.ts

@@ -1,5 +1,5 @@
 import { API_CLIENT_CONDUIT_ANTHROPIC_REFERENCE } from "@infersec/definitions";
-import { implementAPIReference } from "@infersec/fetch";
+import { implementAPIReference, type APIRequest } from "@infersec/fetch";
 import { Logger } from "@infersec/logger";
 import { APIClient } from "../apiClient/index.js";
 import { Configuration } from "../configuration.js";
@@ -19,7 +19,7 @@ export declare function createPostMessagesHandler(options: {
     getModelManager: () => ModelManager;
     logger: Logger;
 }): (params: {
-    req: import("@infersec/fetch").APIRequest;
+    req: APIRequest;
     res: import("@infersec/fetch").APIResponse;
     parameters: Record<string, never>;
     query: Record<string, never>;

package/dist/requestHandlers/createConduitOpenAIAPIReferenceHandlers.d.ts

@@ -1,5 +1,5 @@
 import { API_CLIENT_CONDUIT_OPENAI_REFERENCE } from "@infersec/definitions";
-import { implementAPIReference } from "@infersec/fetch";
+import { implementAPIReference, type APIRequest } from "@infersec/fetch";
 import { Logger } from "@infersec/logger";
 import { APIClient } from "../apiClient/index.js";
 import { Configuration } from "../configuration.js";
@@ -21,7 +21,7 @@ export declare function createGetModelsHandler(options: {
     logger: Logger;
     startup: number;
 }): (params: {
-    req: import("@infersec/fetch").APIRequest;
+    req: APIRequest;
     res: import("@infersec/fetch").APIResponse;
     parameters: Record<string, never>;
     query: Record<string, never>;
@@ -66,7 +66,7 @@ export declare function createPostChatCompletionsHandler(options: {
     logger: Logger;
     startup: number;
 }): (params: {
-    req: import("@infersec/fetch").APIRequest;
+    req: APIRequest;
     res: import("@infersec/fetch").APIResponse;
     parameters: Record<string, never>;
     query: Record<string, never>;
@@ -172,7 +172,7 @@ export declare function createPostCompletionsHandler(options: {
     logger: Logger;
     startup: number;
 }): (params: {
-    req: import("@infersec/fetch").APIRequest;
+    req: APIRequest;
     res: import("@infersec/fetch").APIResponse;
     parameters: Record<string, never>;
     query: Record<string, never>;

package/dist/{start-CPrgh7rN.js → start-DGdf6ycx.js}

@@ -15622,21 +15622,30 @@ const ServerToClientAPIRequestSchema = APIRequestSchema.extend({
 
 _enum(["inf-end", "inf-src", "storage"]);
 
-const PlanSlugSchema = _enum(["basic", "business", "free-trial", "open", "team"]);
-object({
-    href: string$1(),
-    limits: object({
-        endpoints: number$1().int().nonnegative(),
-        sources: number$1().int().nonnegative(),
-        users: number$1().int().nonnegative()
-    }),
-    monthlyCostEUR: number$1().int().nonnegative(),
-    name: string$1(),
-    slug: PlanSlugSchema,
-    trialDays: number$1().int().positive().nullable(),
-    websiteFeatures: array(string$1()),
-    websiteIconPath: string$1()
+const ConnectedSourceMetaSchema = object({
+    chargeName: literal("connected_source"),
+    unitCost: number$1(),
+    units: number$1()
 });
+const TokenInputMetaSchema = object({
+    chargeName: literal("tokens_input"),
+    modelId: string$1().nullable(),
+    requestId: string$1().nullable(),
+    unitCost: number$1(),
+    units: number$1()
+});
+const TokenOutputMetaSchema = object({
+    chargeName: literal("tokens_output"),
+    modelId: string$1().nullable(),
+    requestId: string$1().nullable(),
+    unitCost: number$1(),
+    units: number$1()
+});
+discriminatedUnion("chargeName", [
+    ConnectedSourceMetaSchema,
+    TokenInputMetaSchema,
+    TokenOutputMetaSchema
+]);
 
 function readEnvIntegerOptional({ defaultValue, name }) {
     const str = readEnvStringOptional(name, `${defaultValue}`).trim();
@@ -109465,7 +109474,7 @@ function calculateTokensPerSecond$2({ durationMs, totalTokens }) {
     }
     return Math.round(tokensPerSecond);
 }
-async function proxyOpenAIStreamingRoute({ body, configuration, logger, modelID, modelManager, path, reportMetrics, signal }) {
+async function proxyOpenAIStreamingRoute({ body, configuration, endpointId, logger, modelID, modelManager, path, reportMetrics, signal }) {
     function normalizeTokenCount(value) {
         if (typeof value === "number" && Number.isFinite(value) && value >= 0) {
             return value;
@@ -109493,7 +109502,7 @@ async function proxyOpenAIStreamingRoute({ body, configuration, logger, modelID,
             bytes: requestBodyBytes + responseBytes,
             completionTokens,
             engine: configuration.agentEngineType,
-            endpointId: null,
+            endpointId: endpointId ?? null,
             latencyMs,
             modelId: modelID,
             promptTokens,
@@ -109536,7 +109545,7 @@ async function proxyOpenAIStreamingRoute({ body, configuration, logger, modelID,
             bytes: requestBodyBytes,
             completionTokens: 0,
             engine: configuration.agentEngineType,
-            endpointId: null,
+            endpointId: endpointId ?? null,
             latencyMs,
             modelId: modelID,
             promptTokens: 0,
@@ -109586,7 +109595,7 @@ async function proxyOpenAIStreamingRoute({ body, configuration, logger, modelID,
             bytes: requestBodyBytes,
             completionTokens: 0,
             engine: configuration.agentEngineType,
-            endpointId: null,
+            endpointId: endpointId ?? null,
             latencyMs,
             modelId: modelID,
             promptTokens: 0,
@@ -109637,10 +109646,15 @@ async function proxyOpenAIStreamingRoute({ body, configuration, logger, modelID,
     };
 }
 
+function extractEndpointId$1(req) {
+    const value = req.headers["x-endpoint-id"];
+    const raw = typeof value === "string" ? value : Array.isArray(value) ? value[0] : null;
+    return raw && isValid(raw) ? raw : null;
+}
 function createConduitOpenAIAPIReferenceHandlers({ apiClient, configuration, getModelID, getModelManager, logger, startup }) {
     return {
         "/v1/chat/completions": {
-            POST: async ({ body, res }) => {
+            POST: async ({ body, req, res }) => {
                 const modelID = getModelID();
                 const modelManager = getModelManager();
                 const abortController = new AbortController();
@@ -109650,6 +109664,7 @@ function createConduitOpenAIAPIReferenceHandlers({ apiClient, configuration, get
                 return proxyOpenAIStreamingRoute({
                     body,
                     configuration,
+                    endpointId: extractEndpointId$1(req),
                     logger,
                     modelID,
                     modelManager,
@@ -109660,7 +109675,7 @@ function createConduitOpenAIAPIReferenceHandlers({ apiClient, configuration, get
             }
         },
         "/v1/completions": {
-            POST: async ({ body, res }) => {
+            POST: async ({ body, req, res }) => {
                 const modelID = getModelID();
                 const modelManager = getModelManager();
                 const abortController = new AbortController();
@@ -109670,6 +109685,7 @@ function createConduitOpenAIAPIReferenceHandlers({ apiClient, configuration, get
                 return proxyOpenAIStreamingRoute({
                     body,
                     configuration,
+                    endpointId: extractEndpointId$1(req),
                     logger,
                     modelID,
                     modelManager,
@@ -109789,7 +109805,7 @@ function extractAnthropicNonStreamUsage(body) {
         return null;
     }
 }
-async function proxyAnthropicStreamingRoute({ body, configuration, logger, modelID, modelManager, reportMetrics, signal }) {
+async function proxyAnthropicStreamingRoute({ body, configuration, endpointId, logger, modelID, modelManager, reportMetrics, signal }) {
     function reportMetricsSafe(payload) {
         reportMetrics(payload).catch(error => {
             logger.warn("Failed to upload LLM prompt metrics", {
@@ -109811,7 +109827,7 @@ async function proxyAnthropicStreamingRoute({ body, configuration, logger, model
             bytes: requestBodyBytes + responseBytes,
             completionTokens,
             engine: configuration.agentEngineType,
-            endpointId: null,
+            endpointId: endpointId ?? null,
             latencyMs,
             modelId: modelID,
             promptTokens,
@@ -109854,7 +109870,7 @@ async function proxyAnthropicStreamingRoute({ body, configuration, logger, model
             bytes: requestBodyBytes,
             completionTokens: 0,
             engine: configuration.agentEngineType,
-            endpointId: null,
+            endpointId: endpointId ?? null,
             latencyMs,
             modelId: modelID,
             promptTokens: 0,
@@ -109899,7 +109915,7 @@ async function proxyAnthropicStreamingRoute({ body, configuration, logger, model
             bytes: requestBodyBytes,
             completionTokens: 0,
             engine: configuration.agentEngineType,
-            endpointId: null,
+            endpointId: endpointId ?? null,
             latencyMs,
             modelId: modelID,
             promptTokens: 0,
@@ -110077,10 +110093,15 @@ async function proxyAnthropicStreamingRoute({ body, configuration, logger, model
     };
 }
 
+function extractEndpointId(req) {
+    const value = req.headers["x-endpoint-id"];
+    const raw = typeof value === "string" ? value : Array.isArray(value) ? value[0] : null;
+    return raw && isValid(raw) ? raw : null;
+}
 function createConduitAnthropicAPIReferenceHandlers({ apiClient, configuration, getModelID, getModelManager, logger }) {
     return {
         "/v1/messages": {
-            POST: async ({ body, res }) => {
+            POST: async ({ body, req, res }) => {
                 const modelID = getModelID();
                 const modelManager = getModelManager();
                 const abortController = new AbortController();
@@ -110090,6 +110111,7 @@ function createConduitAnthropicAPIReferenceHandlers({ apiClient, configuration,
                 return proxyAnthropicStreamingRoute({
                     body,
                     configuration,
+                    endpointId: extractEndpointId(req),
                     logger,
                     modelID,
                     modelManager,
@@ -110146,6 +110168,9 @@ async function handleSSERequests({ apiURL, configuration, logger, modelID, onReq
                     const payload = ServerToClientAPIRequestSchema.parse(JSON.parse(message.data));
                     const perRequestController = new AbortController();
                     activeRequests.set(payload.requestID, perRequestController);
+                    const effectiveSignal = signal
+                        ? AbortSignal.any([perRequestController.signal, signal])
+                        : perRequestController.signal;
                     handleRequest({
                         activeRequests,
                         apiURL,
@@ -110157,7 +110182,7 @@ async function handleSSERequests({ apiURL, configuration, logger, modelID, onReq
                         onRequestStart,
                         reportMetrics,
                         request: payload,
-                        signal: perRequestController.signal
+                        signal: effectiveSignal
                     }).catch(error => {
                         logger.error("SSE request handler failed", {
                             error: asError(error),
@@ -110201,6 +110226,7 @@ async function handleRequest({ activeRequests, apiURL, configuration, logger, mo
             });
         });
     }
+    const endpointId = request.parameters?.endpointID ?? null;
     const requestStartedAt = Date.now();
     const requestBytes = calculateRequestBytes(request.body ?? null);
     try {
@@ -110225,7 +110251,7 @@ async function handleRequest({ activeRequests, apiURL, configuration, logger, mo
             bytes: requestBytes + responseMetrics.responseBytes,
             completionTokens: 0,
             engine: configuration.agentEngineType,
-            endpointId: null,
+            endpointId,
             latencyMs,
             modelId: modelID,
             promptTokens: 0,
@@ -110279,7 +110305,7 @@ async function handleRequest({ activeRequests, apiURL, configuration, logger, mo
             bytes: isCancelled ? requestBytes : requestBytes + failureBytes,
             completionTokens: 0,
             engine: configuration.agentEngineType,
-            endpointId: null,
+            endpointId,
             latencyMs,
             modelId: modelID,
             promptTokens: 0,
@@ -110508,7 +110534,12 @@ async function proxyRequest({ configuration, request, signal }) {
     }
     const fetchOptions = {
         method: request.method,
-        headers: request.headers
+        headers: {
+            ...request.headers,
+            ...(request.parameters?.endpointID
+                ? { "x-endpoint-id": request.parameters.endpointID }
+                : undefined)
+        }
     };
     if (signal) {
         fetchOptions.signal = signal;

package/dist/utils/anthropic.d.ts

@@ -3,9 +3,10 @@ import { InferenceAgentLLMMetricsPayload, type ULID } from "@infersec/definition
 import { Logger } from "@infersec/logger";
 import { Configuration } from "../configuration.js";
 import { ModelManager } from "../modelManagement/ModelManager.js";
-export declare function proxyAnthropicStreamingRoute({ body, configuration, logger, modelID, modelManager, reportMetrics, signal }: {
+export declare function proxyAnthropicStreamingRoute({ body, configuration, endpointId, logger, modelID, modelManager, reportMetrics, signal }: {
     body: unknown;
     configuration: Configuration;
+    endpointId?: ULID | null;
     logger: Logger;
     modelID: ULID;
     modelManager: ModelManager;

package/dist/utils/openai.d.ts

@@ -3,9 +3,10 @@ import { InferenceAgentLLMMetricsPayload, type ULID } from "@infersec/definition
 import { Logger } from "@infersec/logger";
 import { Configuration } from "../configuration.js";
 import { ModelManager } from "../modelManagement/ModelManager.js";
-export declare function proxyOpenAIStreamingRoute({ body, configuration, logger, modelID, modelManager, path, reportMetrics, signal }: {
+export declare function proxyOpenAIStreamingRoute({ body, configuration, endpointId, logger, modelID, modelManager, path, reportMetrics, signal }: {
     body: unknown;
     configuration: Configuration;
+    endpointId?: ULID | null;
     logger: Logger;
     modelID: ULID;
     modelManager: ModelManager;

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@infersec/conduit",
   "description": "End user conduit agent for connecting local LLMs to the cloud.",
-  "version": "1.41.0",
+  "version": "1.42.0",
   "bin": {
     "infersec-conduit": "./dist/cli.js"
   },